Patent Grant
11788136
The invention relates to methods of capturing and sequencing immune-associated nucleotide sequences, and more particularly to methods of determining clonality of immune cells.
The maturation of lymphocytes is a fascinating process that is marked not only by immunophenotypic changes, but also by discrete and regulated molecular events(1-3). As T-cells mature, an important part of the associated molecular “maturation” involves the somatic alteration of the germline configuration of the T-cell receptor (TR) genes to a semi-unique configuration in order to permit the development of a clone of T-cells with an extracellular receptor specific to a given antigen(1-3). B-cells undergo a similar maturation process involving different loci that encode the antibody-containing B-cell receptor (BR). These clones, when considered together as a population, produce a repertoire of antigen sensitivity orders of magnitude larger than would be possible by way of inherited immunological diversity alone(3). Indeed, the somatic rearrangement of the TR and BR genes is one of the key ontological events permitting the adaptive immune response(3).
When molecular carcinogenesis occurs in a lymphoid cell lineage, the result is the selective growth and expansion of the tumoural lymphocytes relative to their normal counterparts(2). The so-called precursor (historically termed “lymphoblastic”) lesions are believed to reflect molecular carcinogenesis in lymphoid cells at a relatively immature stage of maturation(2). In contrast, if molecular carcinogenesis occurs at a point during or after the process of T-cell receptor gene re-arrangement (TRGR), the result is a “mature” (often also termed “peripheral”) T-cell lymphoma in which the tumour contains a massively expanded population of malignant T-cells with an immunophenotype reminiscent of mature lymphocytes, most if not all bearing an identical TR gene configuration(4). It is this molecular “homogeneity” of the TR configuration within a T-cell neoplasm that defines the concept of clonality in T-cell neoplasia(1,2,4).
The T-cell receptor is a heteroduplex molecule anchored to the external surface of T lymphocytes(5,21); there the TR, in cooperation with numerous additional signalling and structural proteins, functions to recognize an antigen with a high degree of specificity. This specificity, and indeed the vast array of potential antigenic epitopes that may be recognized by the population of T-cells on the whole, is afforded by (1) the number of TR encoding regions of a given T-cell receptor's genes as present in the germline; and (2) the intrinsic capacity of the TR gene loci to undergo somatic re-arrangement(3). There are four TR gene loci, whose protein products combine selectively to form functional TRs: T-cell receptor alpha (TRA) and T-cell receptor beta (TRB) encode the α and β chains, respectively, whose protein products pair to form a functional α/β TR; T-cell receptor gamma (TRG) and T-cell receptor delta (TRD) encode the γ and δ chains, respectively, whose protein products pair to form a functional γ/δ TR. The vast majority (>95%) of circulating T-cells are of the α/β type(21,22); for reasons as yet not fully understood, γ/δ T-cells tend to home mainly to epithelial tissues (e.g. skin and mucosae) and appear to have a different function than the more common a/s type T-cells.
The TRA locus is found on the long arm of chromosome 14 in band 14q11.2 and spans a total of 1000 kilobases (kb)(23); interestingly, sandwiched between the TRA V and J domains, is the TRD locus (14q11.2), itself spanning only 60 kb(24). The TRB locus is found on the long arm of chromosome 7 in band 7q35 and spans a total of 620 kb(25). The TRG locus is found on the short arm of chromosome 7 in region 7p15-p14 and spans 160 kb(26).
Within each TR gene locus are a variable number of variable (V) and join (J) segments(23-26) additional diversity (D) segments are present within the TRB and TRD loci(24,25). These V, D and J segments are grouped into respective V, D and J regions (see
BIOMED-2(29) is a product of several years of collaborative expert study, resulting in a thoroughly studied consensus T-cell clonality assay. The BIOMED-2 assay includes multiplexed primer sets for both Immunoglobulin (IG) and TR clonality assessment and can be implemented with commercially available electrophoresis systems (e.g. Applied Biosystems fluorescence electrophoresis platforms)(29). These commercially available primer sets have the advantage of standardization and ease of implementation. In addition, by virtue of the extensive study performed by the BIOMED consortium, the BIOMED-2 assay has the well-documented advantage of capturing the mono-clonality of the vast majority of control lymphomas bearing productive T-cell receptors (i.e. flow-sorted positive for either α/β or γ/δ T-cell receptors) using the specified TRB and TRG primer sets(29). Of note, having been in use for over a decade, the BIOMED-2 has been globally accepted as the diagnostic assay primer set of choice.
The current approach to TRGR testing is subject to a number of technical and practical caveats that dilute the applicability of TRGR testing to the full breadth of real-world contexts.
Because the PCR-based techniques that are employed in TRGR assays are subject to amplicon size restrictions(29,34), the sheer size of the TRA locus prevents a complete assay of the TRA gene in clinical settings. Indeed, although of smaller size, the TRB locus as a whole is also prohibitively large to sequence in its germline configuration. It is therefore of no surprise that much of the published data pertaining to the utility and validity of TRGR assays has stemmed from assays specific to only subparts of TRB as well as TRG, a locus of size much more amenable to a single-assay. In addition, since the TRD locus is often deleted after TR gene rearrangement (since it is contained within the TRA locus and excised whenever the TRA locus is rearranged), assays for TRD have also not been as rigorously studied. For this reason, any BIOMED-2-based T-cell clonality assay aimed at directing immunotherapy, requiring a complete sequence-based understanding of the TR genes involved, would be insufficient.
The BIOMED-2 assay is subject to additional technical challenges. As part of the standard TRGR assay, most laboratories rely on the demonstration of electrophoretic migration patterns for the determination of TR clonality. Interpretation of the assay depends on the demonstration (or lack thereof) of a dominant amplicon of specific (albeit not pre-defined) molecular weight, rather than the normal Gaussian distribution of amplicons of variable size. This approach, as has been described previously(35-47), is subject to interpretative error and other technical problems. Also, given the large amounts of DNA required for the multitude of multiplex tubes making up the assay, the overall assay can very quickly deplete DNA supplies, especially when obtained from limited sample sources.
Finally, and arguably of greatest import, is the issue of diagnostic bias used in the study of TRGR assay performance. More precisely, when laboratories seek to validate a TRGR assay, the requirement of “standard” samples will typically require that the laboratory utilize previously established clonal samples or samples previously diagnosed and accepted to represent clonal entities (e.g. previously diagnosed cases of lymphoma); these samples are in turn compared to “normal” controls. In contrast, the demographics of subsequent “real-life” test samples are unlikely to be so decidedly parsed into “normal” and “abnormal” subsets.
Current T-Cell Receptor (TCR) rearrangement profiling assays rely on targeted PCR amplification of rearranged TCR genomic loci. The simplest method for assessing clonality of T-cells involves qualitative assessment through multiplexed amplification of the individual loci using defined primer sets and interpretation of fragment size distributions according to the BIOMED2 protocolA1,2. Next-generation sequencing can be used as a read-out to provide quantitative assessment of the TCR repertoire including detection of low abundance rearrangements from bulk immune cells, or even pairing of the heterodimeric chain sequences with single cell preparation methodsA3,4. Hybrid-capture based library subsetting is an alternative method to PCR-based amplification that can improve coverage uniformity and library complexity when sample is not limiting and allows for targeted enrichment of genetic loci of interest from individual genes to entire exomesA5. In hybrid-capture methods, the formation of probe-library fragment DNA duplexes are used to recover regions of interestA6 7,8.
Similar to T-cells, B-cells involved in adaptive immunity also undergo somatic rearrangement of germline DNA to encode a functional B-cell receptor (BR). Like TRs, these sequences comprise by discrete V, D, J segments that are rearranged and potentially altered during B-cell maturation to encode a diversity of unique immunoglobulin proteins. The clonal diversity of B-cell populations may have clinical utility and, similar to T-cell lymphomas, several cancers are characterized by clonal expansion of specific BR/Ig sequences.
There is described herein, the development of a novel NGS-based T-cell clonality assay, incorporating all four TR loci. The assay was both analytically and clinically validated. For the former, a series of idealized specimens was used, with combined PCR/Electrophoresis and Sanger Sequencing to confirm NGS-data. The latter validation compared NGS results to the current gold standard for clinical T-cell clonality testing (i.e. the BIOMED-2 primer PCR method) on an appropriately-sized minimally-biased sample of hematopathology specimens. In the latter dataset also, the patterns of T-cell clonality were also correlated with clinical, pathologic, and outcome data.
In an aspect, there is provided, a method of capturing a population of T-Cell receptor and/or immunoglobulin sequences with variable regions within a patient sample, said method comprising: extracting/preparing DNA fragments from the patient sample; ligating a nucleic acid adapter to the DNA fragments, the nucleic acid adapter suitable for recognition by a pre-selected nucleic acid probe; capturing DNA fragments existing in the patient sample using a collection of nucleic acid hybrid capture probes, wherein each capture probe is designed to hybridize to a known V gene segment and/or a J gene segment within the T cell receptor and/or immunoglobulin genomic loci.
In an aspect, there is provided, a method of immunologically classifying a population of T-Cell receptor and/or immunoglobulin sequences, the method comprising:
and optionally (h) and (i):
In an aspect, there is provided, a method of identifying CDR3 regions in T-Cell receptor and/or immunoglobulin sequences, the method comprising:
The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.
These and other features of the preferred embodiments of the invention will become more apparent in the following detailed description in which reference is made to the appended drawings wherein:
In the following description, numerous specific details are set forth to provide a thorough understanding of the invention. However, it is understood that the invention may be practiced without these specific details.
The advantages of high-throughput DNA sequencing technologies could potentially be applied to T-cell clonality testing. The nature of T-cell gene diversity, requiring the consideration of potential variability arising from four distinct gene loci, makes obvious the benefit of multiplexing; what has traditionally required multiple separate tests could be combined in a single reaction. The capacity of modem DNA sequencing technologies to query longer contiguous segments of DNA in greater quantities relative to traditional techniques also provides an opportunity to explore the potential meaning of TRA and TRB sequence rearrangements. Sequence-level data might afford a greater ease of assay result interpretation. Indeed, the generation of sequence-level data in a TRGR assay would likely be much more informative than gross estimates of DNA electrophoretic migration patterns when disease trends are being studied; the high-level analysis of such data might help the identification of heretofore hidden patterns of TR rearrangement in specific T-cell lymphoma subtypes. The issue of replicate numbers for establishing test sensitivity/specificity can be easily overcome by exploiting the high-throughput capacity of modem DNA sequencing platforms; for a comparable investment of time (and possibly cost), sequencing-based approach to TRGR could perform a greater number of individual tests, thereby potentially allowing a more statistically robust estimate of test performance.
Traditional sequencing uses PCR-based techniques to markedly amplify input template DNA, thus improving the sensitivity of detection during the sequencing step. Indeed, many sequencing-based technologies still perform directed library preparation using PCR-based techniques to isolate and sequence regions of interest(38). By this approach, one might employ specific primer sets to enrich for regions of interest in the library preparation step. In the context of TRGR, however, a primer-based approach to library preparation would be challenging: in order to provide the sufficient breath of coverage required to interrogate the status of the vast number of TR genes (especially in the TRA locus), a massive array of primers would be required. Although it is theoretically possible to prime multiple regions in tandem, previous data suggest that such an approach might open the door to the possibility of technical error (for a more thorough review of the details of these errors and the studies that have supported this evidence, see(38)). In the context of TRGR, furthermore, a primer-based approach to library preparation introduces the possibility of allele dropout when the assay attempts to prime a rearranged gene based on the known germline configuration (an easily digestible review to this effect may be found here(39)).
A paradigm shift away from PCR primer-directed amplification of genomic areas of interest was required for sequencing experiments aimed at large numbers of genes. Indeed most sequencing-based technologies rather employ the upfront production of vast libraries of template oligonucleotides followed by a series of template enrichment steps(38). These latter steps may simply involve the extraction of DNA of specific lengths or quality, or rather the focus may be to enrich DNA containing specific sequences of interest. In the latter scenario, when specific sequence motifs are enriched for during library preparation, the resulting sequencing data will be enriched for the sequences of interest. Additionally, using the above stepwise approach, library preparation may be generalized to permit the enrichment of specific sequences out of a mix of “all” sequences produced from the primary non-specific amplification step; it is easy to see how this approach may be used to permit multiple separate assays using different enrichment approaches applied to a single input library(40).
Hybrid capture is a form of library enrichment in which a library is probed for known sequences of interest using tagged nucleic acid probes followed by a subsequent “pull-down” of the tagged hybrids(38); for example, DNA probes tagged with biotin can be efficiently enriched when hybridization is followed by a streptavidin enrichment step(38,40-43). The biotin/streptavidin enrichment procedure is schematized in
Rather than restricting the assessment of test performance of the above DNA sequencing approaches to a pre-set (and potentially biased) sample of “malignant” and “benign” T-cell lymphoproliferative disorders, a more prudent sampling rubric might use a “real-world” series of consecutive samples taken from a population as similar to the “test population” as possible. In the context of TRGR validation, such a sample might consist of a series of consecutive tissue samples from patients being worked-up by a hematologist and submitted for molecular (i.e. T-cell clonality) assessment. The overall sample size could be established based on an estimate of the historical incidence of T-cell lymphomas in such a population, such that the total size of the sample is adequately large to include a sufficient “expected” number of clonal T-cell lymphoproliferative disorders.
In many validation studies, the final pathology diagnosis is used as the gold standard against which the novel test is measured(41). While not unreasonable, there are arguments against employing such an approach. Of foremost concern is the potential for diagnostic or interpretative error, by which “true positivity” of disease could be misappropriated(44). In the realm of T-cell lymphomas, given at least partly due to their rarity, the frequent lack of pathologist experience might make this problem more likely. Furthermore, evidence indicates that even when diagnoses are based on consensus or panel based interpretation, the possibility of diagnostic bias by dominant opinion should be considered(45).
When a single clearly-defined outcome measure does not exist (or is limited by bias), a composite gold-standard might be more appropriate(46). Composite gold-standards might include a number of individual test results or clinical observations logically combined to produce “positive” or “negative” composites(46); of key import is that (1) well-defined rules of composition be set out a priori and (2) the number of samples or subjects with each of the composite test results should be well-described(46). Ideally, all samples or subjects should be evaluated using each of the composite tests(46).
In order to best study a novel test of TLPDs, rather than limiting the reference test to the gold-standard BIOMED-2 T-cell clonality assay or to pathology diagnoses, a series of both individual and composite references might be considered. From the perspective of analytical validity, one might consider validating an sequencing-based TRGR assay using standard PCR techniques followed by Sanger sequence verification. Since the sequences of each of the TR V and J genes are known, forward and reverse primer sets for each V and J genes, respectively, identified by the capture and sequencing assay could be used to verify that the detected result is valid; this could be followed by Sanger sequencing to validate the result of the DNA sequencing result (with deference specifically to the CDR3 variability-defining region).
In another experiment, one might consider comparing a sequencing-based TRGR result to the BIOMED-2 result (with each test applied to all specimens under study). The primary limitation of this approach would be that the BIOMED-2 assay, as explained above, does not test for any TRA rearrangements; thus this comparison alone would be insufficient. Additional comparisons might involve assessment of the sensitivity and specificity of each of the BIOMED-2 and sequencing-based TRGR assays at identifying benign or malignant TLPDs. For this, a composite gold-standard including histologic features (i.e. pathology diagnosis), immunophenotypic features, additional molecular features (as available, e.g. cytogenetic changes), clinical observations (e.g. presence or absence of features of malignancy), and outcome results (e.g. significant deviation in individual patient survival from the median) might be considered. The clinical validity of the sequencing results could thus be assessed against the current diagnostic standard by means of a much more thorough evaluation.
T-cell lymphomas are cancers of immune cell development that result in clonal expansion of malignant clones that dominate the T-cell repertoire of affected patients. Therefore, clonality assessment of these cell populations is essential for the identification and monitoring of T-cell lymphomas. We have developed a hybrid-capture method that recovers rearranged sequences of T-cell receptor (TCR) chains from all four classes (alpha, beta, gamma, and delta loci) in a single reaction from an Illumina sequencing library. We use this method to describe the TCR V(D)J repertoire of monoclonal cancer cell lines, tumor-derived lymphocyte cultures, and peripheral blood mononuclear cells from a healthy donor, as well as a set of 63 clinical isolates sent for clinical clonality testing for suspected T-cell lymphoma. PCR amplification and Sanger sequencing confirmed cell line and tumor predominant rearrangements, individual beta locus V and J allele prevalence was well correlated with results from a commercial PCR-based DNA sequencing assay with an r2 value of 0.94, and BIOMED2 PCR fragment size beta and gamma locus clonotyping of clinical isolates showed 73% and 77% agreement respectively. Our method allows for rapid, high-throughput and low cost characterization of TCR repertoires that will enhance sensitivity of tumor surveillance as well as facilitate serial analysis of patient samples with a quantitative read-out during clinical immunotherapy interventions.
In an aspect, there is provided, a method of capturing a population of T-Cell receptor and/or immunoglobulin sequences with variable regions within a patient sample, said method comprising: extracting/preparing DNA fragments from the patient sample; ligating a nucleic acid adapter to the DNA fragments, the nucleic acid adapter suitable for recognition by a pre-selected nucleic acid probe; capturing DNA fragments existing in the patient sample using a collection of nucleic acid hybrid capture probes, wherein each capture probe is designed to hybridize to a known V gene segment and/or a J gene segment within the T cell receptor and/or immunoglobulin genomic loci.
As used herein, “T-Cell Receptor” or “TCR” means a molecule found on the surface of T lymphocytes (or T cells), preferably human, that is responsible for recognizing fragments of antigen as peptides bound to major histocompatibility complex (MHC) molecules. The TCR is a disulfide-linked membrane-anchored heterodimeric protein normally consisting of the highly variable alpha (α) and beta (β) chains expressed as part of a complex with the invariant CD3 chain molecules. T cells expressing this receptor are referred to as α:β (or αβ) T cells, though a minority of T cells express an alternate receptor, formed by variable gamma (γ) and delta (δ) chains, referred as γδ T cells. Each chain is composed of two extracellular domains: Variable (V) region and a Constant (C) region. The variable domain of both the TCR α-chain and β-chain each have three hypervariable or complementarity determining regions (CDRs). CDR3 is the main CDR responsible for recognizing processed antigen.
The terms “antibody” and “immunoglobulin”, as used herein, refer broadly to any immunological binding agent or molecule that comprises a human antigen binding domain, including polyclonal and monoclonal antibodies. Depending on the type of constant domain in the heavy chains, whole antibodies are assigned to one of five major classes: IgA, IgD, IgE, IgG, and IgM. Several of these are further divided into subclasses or isotypes, such as IgG1, IgG2, IgG3, IgG4, and the like. The heavy-chain constant domains that correspond to the difference classes of immunoglobulins are termed α, δ, ε, γ and μ, respectively. The subunit structures and three-dimensional configurations of different classes of immunoglobulins are well known. The “light chains” of mammalian antibodies are assigned to one of two clearly distinct types: kappa (κ) and lambda (λ), based on the amino acid sequences of their constant domains and some amino acids in the framework regions of their variable domains. The variable domains comprise the complementarity determining regions (CDRs). The methods described herein may be applied to immunoglobulin sequences, including B-cell immunoglobulin sequences.
“V gene segments”, “J gene segments” and “D gene segments” as used herein, refer to the variable (V), joining (J), and diversity (D) gene segments involved in V(D)J recombination, less commonly known as somatic recombination. V(D)J recombination is the mechanism of genetic recombination that occurs in developing lymphocytes during the early stages of T and B cell maturation. The process results in the highly diverse immune repertoire of antibodies/immunoglobulins (Igs) and T cell receptors (TCRs) found on B cells and T cells, respectively.
The term “nucleic acid” includes DNA and RNA and can be either double stranded or single stranded.
The term “probe” as used herein refers to a nucleic acid sequence that will hybridize to a nucleic acid target sequence. In one example, the probe hybridizes to the RNA biomarker or a nucleic acid sequence complementary thereof. The length of probe depends on the hybridization conditions and the sequences of the probe and nucleic acid target sequence. In one embodiment, the probe is at least 8, 10, 15, 20, 25, 50, 75, 100, 150, 200, 250, 400, 500 or more nucleotides in length.
The term “adapter” as used herein refers a moiety capable of conjugation to a nucleic acid sequence for a particular purpose. For example, the adapter may be used to identify or barcode the nucleic acid. Alternatively, the adapter may be a primer which can be used to amplify the nucleic acid sequence.
The term “hybridize” or “hybridizable” refers to the sequence specific non-covalent binding interaction with a complementary nucleic acid. In a preferred embodiment, the hybridization is under stringent conditions. Appropriate stringency conditions which promote hybridization are known to those skilled in the art, or can be found in Current Protocols in Molecular Biology, John Wiley & Sons, N.Y. (1989), 6.3.1 6.3.6. For example, 6.0× sodium chloride/sodium citrate (SSC) at about 45° C., followed by a wash of 2.0×SSC at 50° C. may be employed.
In some embodiments, the method further comprises sequencing the captured DNA fragments, wherein the sequencing can be used to determine clonotypes within the patient sample. Various sequencing techniques are known to the person skilled in the art, such as polymerase chain reaction (PCR) followed by Sanger sequencing. Also available are next-generation sequencing (NGS) techniques, also known as high-throughput sequencing, which includes various sequencing technologies including: Illumina (Solexa) sequencing, Roche 454 sequencing, Ion torrent Proton/PGM sequencing, SOLiD sequencing. NGS allow for the sequencing of DNA and RNA much more quickly and cheaply than the previously used Sanger sequencing. In some embodiments, said sequencing is optimized for short read sequencing.
In some embodiments, the method further comprises amplifying the population of sequences using nucleic acid amplification probes/oligonucleotides that recognize the adapter prior to said sequencing.
In some embodiments, the method further comprises fragmenting DNA extracted from the patient sample to generate the DNA fragments.
In some embodiments, the ligating step is performed before the capturing step.
In some embodiments, the capturing step is performed before the ligating step.
The term “patient” as used herein refers to any member of the animal kingdom, preferably a human being and most preferably a human being that has AML or that is suspected of having AML.
The term “sample” as used herein refers to any fluid, cell or tissue sample from a subject which can be assayed for nucleic acid sequences. In some embodiments, the patient sample comprises tissue, urine, cerebral spinal fluid, saliva, feces, ascities, pleural effusion, blood or blood plasma.
In some embodiments, the patient sample comprises cell-free nucleic acids in blood plasma.
In some embodiments, the clonality analyses described herein may be use to track clonality across samples types.
In some embodiments, the hybrid capture probes are at least 30 bp in length. In a further embodiment, the hybrid capture probes are between 60 bp and 150 bp in length. In a further embodiment, the hybrid capture probes are between 80 bp and 120 bp in length. In a further embodiment, the hybrid capture probes are about 100 bp in length.
In some embodiments, the hybrid capture probes hybridize to at least 30 bp, preferably 50 bp, more preferably 100 bp of the V gene segment and/or J gene segment.
In some embodiments, the hybrid capture probes hybridize to at least a portion of the V gene segment and/or J gene segment at either the 3′ end or the 5′ end of the V gene segment and/or J gene segment respectively.
In some embodiments, the screening probes hybridize to at least a portion of the V gene segment.
In some embodiments, the screening probes hybridize to at least a portion of the V gene segment at the 3′ end.
In some embodiments, hybridizing comprises hybridizing under stringent conditions, preferably very stringent conditions.
In some embodiments, the collection of nucleic acid hybrid capture probes comprise at least 2, 5, 10, 20, 30, 80, 100, 300, 400, 500, 600, 700, 800 or 900 unique hybrid capture probes.
In some embodiments, the collection of nucleic acid hybrid capture probes is sufficient to capture at least 50%, 60%, 70%, 80%, 90% or 99% of known T-Cell receptor and/or immunoglobulin loci clonotypes.
In some embodiments, the hybrid capture probes are immobilized on an array.
In some embodiments, the hybrid capture probes comprise a label. In a further embodiment, the label is used to distinguish between sequences bound to the screening probes and unbound double stranded fragments, and preferably the capture is performed in solution.
In some embodiments, preparing the DNA fragments comprises extracting RNA from the patient sample and preparing corresponding cDNA.
In some embodiments, the method further comprises a depletion step, comprising depleting the DNA fragments of non-rearranged sequences using probes that recognize nucleic acid sequences adjacent to V and/or J gene segments in the genome. In some embodiments, the capturing of DNA fragments using V gene segment and J gene segment hybrid capture probes is performed in separate steps, and in any order with the depletion step, preferably in the following order: J gene capture, depletion, then V gene capture.
In an aspect, there is provided, a method of immunologically classifying a population of T-Cell receptor and/or immunoglobulin sequences, the method comprising:
Alternatively, step (h) may be searching the 6 translated sequences for flanking invariable anchor sequences to define the intervening T-Cell receptor and/or B-cell receptor CDR3 sequences encoded by the DNA fragments.
In some embodiments, the method further comprises, prior to step (a), aligning left and right reads of overlapping initial DNA fragments to produce the DNA fragments on which step (a) is performed.
In some embodiments, steps (a), (c), (e) are performed with BLASTn and step (i) is performed using expression pattern matching to known sequences and IMGT annotated data.
In an aspect, there is provided, a method of identifying CDR3 regions in T-Cell receptor and/or immunoglobulin sequences, the method comprising:
Alternatively, step (c) may be if V and J gene segments are identified, then searching the immunoglobulin sequence for flanking invariable anchor sequences to define the intervening T-Cell receptor and/or immunoglobulin CDR3 sequences.
In some embodiments, wherein steps (a) and (b) are performed using the Burrows-Wheeler Alignment or other sequence alignment algorithm.
In some embodiments, wherein if a CDR3 region is identified in step (c), then the method further comprises determining whether the identified V and J gene segments could be rearranged in the same locus using a heuristic approach.
In some embodiments, wherein if a CDR3 region is not identified in step (c), then the method further comprises determining if a combination of V(D)J gene segments is present based on Smith Waterman Alignment scores.
In an aspect, there is provided, a method for characterizing the immune repertoire of a subject, the immune repertoire comprising the subject's T-Cell population, the method comprising any of the hybrid capture methods described herein, any of the algorithmic methods described herein, or any combination thereof.
Any of the methods described herein may be used to capture a population of T-Cell receptor sequences, for immunologically classifying a population of T-Cell receptor sequences or for identifying CDR3 regions in T-Cell receptor.
In an aspect, the methods described herein are for characterizing T-cell clonality for a disease in the subject.
In some embodiments, the T-Cell receptor sequences are from tumour infiltrating lymphocytes.
In an aspect, the methods described herein are for identifying therapeutic tumour infiltrating lymphocytes for the purposes of expansion and reinfusion into a patient and/or adoptive cell transfer immunotherapy.
In an aspect, the methods described herein are for monitoring T-cell populations/turnover in a subject, preferably a subject with cancer during cancer therapy, preferably immunotherapy.
In an aspect, the methods described herein are for characterizing the immune repertoire of a subject, the immune repertoire comprising the subject's B-Cell population.
In an aspect, the methods described herein are for capturing a population of B-Cell receptor sequences with variable regions within a patient sample, for immunologically classifying a population of B-Cell receptor sequences, or for identifying CDR3 regions in B-Cell receptor sequences.
In an aspect, the methods described herein are for characterizing B-cell clonality as a feature of a disease in the subject.
The present methods may be used in subjects who have cancer. Cancers include adrenal cancer, anal cancer, bile duct cancer, bladder cancer, bone cancer, brain/cns tumors, breast cancer, castleman disease, cervical cancer, colon/rectum cancer, endometrial cancer, esophagus cancer, ewing family of tumors, eye cancer, gallbladder cancer, gastrointestinal carcinoid tumors, gastrointestinal stromal tumor (gist), gestational trophoblastic disease, hodgkin disease, kaposi sarcoma, kidney cancer, laryngeal and hypopharyngeal cancer, leukemia (acute lymphocytic, acute myeloid, chronic lymphocytic, chronic myeloid, chronic myelomonocytic), liver cancer, lung cancer (non-small cell, small cell, lung carcinoid tumor), lymphoma, lymphoma of the skin, malignant mesothelioma, multiple myeloma, myelodysplastic syndrome, nasal cavity and paranasal sinus cancer, nasopharyngeal cancer, neuroblastoma, non-hodgkin lymphoma, oral cavity and oropharyngeal cancer, osteosarcoma, ovarian cancer, pancreatic cancer, penile cancer, pituitary tumors, prostate cancer, retinoblastoma, rhabdomyosarcoma, salivary gland cancer, sarcoma—adult soft tissue cancer, skin cancer (basal and squamous cell, melanoma, merkel cell), small intestine cancer, stomach cancer, testicular cancer, thymus cancer, thyroid cancer, uterine sarcoma, vaginal cancer, vulvar cancer, waldenstrom macroglobulinemia, and wilms tumor.
In embodiments relating to T-cells, the subject may have a T-cell related disease, such as a T-cell lymphoma.
T-cell lymphomas are types of lymphoma affecting T cells, and can include peripheral T-cell lymphoma not otherwise specified, extranodal T cell lymphoma, cutaneous T cell lymphoma, including Sézary syndrome and Mycosis fungoides, anaplastic large cell lymphoma, angioimmunoblastic T cell lymphoma, adult T-cell Leukemia/Lymphoma (ATLL), blastic NK-cell Lymphoma, enteropathy-type T-cell lymphoma, hematosplenic gamma-delta T-cell Lymphoma, lymphoblastic Lymphoma, nasal NK/T-cell Lymphomas, treatment-related T-cell lymphomas.
In other embodiments relating to B-cells, the subject may have a B-cell related disease, plasma cell disorder, preferably a B-cell lymphoma.
B-cell are types of lymphoma affecting B cells and can include, diffuse large B-cell lymphoma (DLBCL), follicular lymphoma, marginal zone B-cell lymphoma (MZL) or mucosa-associated lymphatic tissue lymphoma (MALT), small lymphocytic lymphoma (also known as chronic lymphocytic leukemia, CLL), mantle cell lymphoma (MCL), DLBCL variants or sub-types of primary mediastinal (thymic) large B cell lymphoma, T cell/histiocyte-rich large B-cell lymphoma, primary cutaneous diffuse large B-cell lymphoma, leg type (Primary cutaneous DLBCL, leg type), EBV positive diffuse large B-cell lymphoma of the elderly, diffuse large B-cell lymphoma associated with inflammation, Burkitt's lymphoma, lymphoplasmacytic lymphoma, which may manifest as Waldenström's macroglobulinemia, nodal marginal zone B cell lymphoma (NMZL), splenic marginal zone lymphoma (SMZL), intravascular large B-cell lymphoma, primary effusion lymphoma, lymphomatoid granulomatosis, primary central nervous system lymphoma, ALK-positive large B-cell lymphoma, plasmablastic lymphoma, large B-cell lymphoma arising in HHVB-associated multicentric Castleman's disease, B-cell lymphoma, unclassifiable with features intermediate between diffuse large B-cell lymphoma and Burkitt lymphoma, B-cell lymphoma, unclassifiable with features intermediate between diffuse large B-cell lymphoma and classical Hodgkin lymphoma, AIDS-related lymphoma, classic Hodgkin's lymphoma and nodular lymphocyte predominant Hodgkin's lymphoma.
In an aspect, the methods described herein are for identifying therapeutic B-cells for the purposes of expansion and reinfusion into a patient.
In an aspect, the methods described herein are for monitoring B-cell populations/turnover in a subject, preferably a subject with cancer during cancer therapy, preferably immunotherapy.
In an aspect, the methods described herein are for detecting minimal residual disease, whereby TCR or immunoglobulin rearrangements may be used as a marker of disease.
In an aspect, there is provided a library of probes comprising the depletion probes in Table D or at least one of the V-gene and J-gene probes set forth in any of Tables 2.1, 4, B1, or B2.
In some embodiments, the clonality analyses described herein may be performed serially.
In some embodiments, the clonality analyses described herein may be used to distinguish between samples.
The advantages of the present invention are further illustrated by the following examples. The examples and their particular details set forth herein are presented for illustration only and should not be construed as a limitation on the claims of the present invention.
Methods and Materials
Assay Development
Several important theoretical considerations were entertained during the design phase of our novel sequencing-based TRGR assay (heretofore referred to as the NTRA).
Unlike the current BIOMED approach, we wished to avoid a gene-specific primer-based approach to signal amplification. To accomplish this, we chose a “hybrid capture” target enrichment approach by which input genomic DNA containing the TR genes might be enriched (or “captured”) relative to other segments of the genome. Several methodological approaches to target enrichment already exist, with multiple commercially available and rigorously optimized kits capable of enriching nearly any well-defined gene target(s)(47-48).
The NTRA needed to be robust enough to accommodate sample types of variable DNA quality; this requirement reflects the clinical need to apply TRGR assays to a wide variety of specimens in a wide variety of contexts. Knowing that Formalin-fixed paraffin-embedded (FFPE) specimens typically contain degraded and often poor quality DNA (as such representing the “lowest common denominator” of specimen quality)(49), it was deemed necessary to specifically evaluate NTRA performance on FFPE specimens. Furthermore, the use of hybrid capture is also amenable to highly fragmented DNA specimens such as those from circulating cell-free DNA.
Likewise, the most useful NTRA should allow users to both accurately assess the “clonality” of an input sample (as can be done using BIOMED-2 based assays) but also fully characterize the clonotypes of constituent TRGR configurations. Thus it was essential that the NTRA not simply produce a binary “clonal” vs. “polyclonal” result but also provide a much more robust and quantitative data output, including the genes and CDR3 regions present within identified TRGR configurations.
We recognized that much of the utility of the NTRA would depend on the design of a robust bioinformatic analysis pipeline. Of note, at the time at which this project was undertaken, only a single widely-used pipeline existed (the International standard source for ImMunoGeneTics sequences & metadata (IMGT) V-QUEST system), mainly designed around 5′RACE PCR followed by Roche 454 sequencing(51). As outlined below, several methodological and logistic motivations demanded a novel pipeline of our own design.
Current sequencing-based applications generally require that resultant sequence data (i.e. reads) be mapped to a reference (typically the genome of the organism of interest) using some form of alignment algorithm. Once this alignment is complete, secondary and tertiary tools are used to search for and catalogue sequence deviation from the reference. For our purposes, however, using the entire human genome as a reference map would be unnecessarily cumbersome, especially since the presence of closely juxtaposed V(D)J sequence within a single short (i.e. <500 basepairs (bp)) fragment of DNA is tantamount to evidence of TRGR. Furthermore, aligning to a single reference genome raises the informatics challenge of detecting gene rearrangements from a single alignment step. As such, a strategy of mapping sequence reads to only the reference genes in a parallel fashion (i.e. one mapping procedure to the V genes, and one separate mapping procedure to the J genes) was selected, along with an integrated TRGR detection algorithm
This strategy required the theoretical consideration that short sequence read input might result in excessive false negatives (i.e. artificially low TRGR detection rates). This problem might be mitigated, in theory at least, by ensuring that input DNA fragment lengths (and the resulting sequencing read lengths) are carefully set to within a reasonable range of sensitivity for the detection of TRGR in a given sequence. Since all possible TRGRs are combinatorially vast, this process could only be simulated using, for our purposes, an artificial test set of simply-concatenated sequences of all catalogued V, D, and J genes (a test set numbering 197400). By evaluating k-mer subsequences over a range of lengths (k), centred (without loss of generality) about the median of each artificial junction, an estimate of the sensitivity of TRGR detection for variable sequencing windows can be produced. This sequencing window can then be used as an “evidence-based” DNA insert length.
Insert Length Simulation
Appendix 2.0 outlines a MATLAB script designed to estimate the optimal DNA insert length (a value also generalizable to optimal shearing length and minimal Paired-end rEAd mergeR (PEAR)-assembled sequencing length) for the purposes of the NTRA. This optimum is subject to an important restriction: for our purposes, using the Illumina NEXTSEQ platform, read lengths are limited to paired-ended reads of 150 bp each—this translates to <300 bp read lengths when paired-ends are joined by overlapping sequence (using, in our case, the PEAR algorithm(52)). Briefly, the code produces a simulation read set of all possible combinations of V-D-J sequences by way of simple concatenation (with the caveat that a much larger diversity of sequence is found in nature stemming from alterations of junctional sequence by way of splicing inconsistencies); next, the algorithm selects a k-mer (of length from k=32 to 302, in intervals of 30 bp) from within each simulation sequence; the resulting k-mer (centred, without loss of generality, at the junction median) is then subject to Burrows-Wheeler Alignment algorithm (BWA) alignment against the known reference V and J genes (as in the TRSeq pipeline) to evaluate how well the k-mers of each of the artificial reads can be mapped to both V and J genes (representing bioinformatic identification of TRGR within the sequence in question). A histogram of percent detection vs. read length was then produced; analysis of those artificial V-D-J read combinations that could be reliably detected was also performed.
DNA Probe Design
We began by reviewing the sequence and metadata of all reference TR genes obtained by way of a (FASTA-formatted) data download from the IMGT database. All sequences were subjected to a series of Clustal W(53) alignment analyses to verify that sequence alignment was limited to known reference motifs (i.e. the J-gene F/W-G-X-G motif and V-gene conserved Cysteine(54)) and to allele-to-allele overlap.
DNA probe design was then performed using the IMGT reference sequences (including all annotated V and J gene functional, pseudogene and open reading frame sequences) using the XGEN LOCKDOWN probe technology. Briefly, this technology is a hybrid-capture-based technology by which biotin-tagged DNA probes (complementary to known sequences/genomic regions set at a 1×depth of coverage) are allowed to hybridize with sample DNA, followed by a streptavidin elution procedure performed to enrich the target sequences(40-43).
In line with previous studies employing XGEN LOCKDOWN probes(40), each DNA probe was designed to a length as close to 100 bp as possible. Using the IMGT database, germline-configuration sequences were extracted for all alleles of all J-genes, with additional leading and trailing IMGT nucleotides added (as necessary) to obtain 100 bp probe lengths; for those instances in which the IMGT data was insufficient to prepare 100 bp probes, additional random nucleotides were added to the leading and trailing ends of the available sequences. Again using the IMGT database, germline-configuration sequences were extracted for all alleles of all V-genes, with additional leading and trailing IMGT nucleotides added to ensure that the 5′ and 3′ ends of the germline-configuration genes were covered by a given probe (this design, it was theorized, would be able to account for gene re-arrangement at either end of a V-gene, regardless of strandedness, while still covering the vast majority of the sequence of each gene/allele). With careful placement of the probes as outlined above, we hoped that this design would also limit any specific stoichiometric bias among the V-genes represented in the target pool.
Table 2.1 outlines the complete list of XGEN LOCKDOWN probe design sequences (with relevant associated metadata).
NTRA Work-Flow
The NTRA work-flow is summarized in
NTRA Data Analysis: The TRSeq Pipeline
The NTRA TRSeq pipeline was designed around three main algorithmic steps. The first performs local alignment indexed to the TR V and J genes implemented using the Burrows-Wheeler-Alignment (BWA) algorithm(55). From this algorithm, two important results are obtained: the first is a “reads-on-target” estimate (since the genes enriched for (i.e. the TR V and J genes) are those genes used as the index reference gene set); second, by way of the resulting Sequence Alignment Map (SAM) file output, the original input reads are filtered to exclude those unlikely to contain any of the TR V or J genes. This latter step reduces the informatic burden of input to the (relatively computationally slow) second algorithm step (using either heuristics or the Smith-Waterman Alignment (SWA)). Of note, the BWA algorithm could be implemented on a UNIX-based platform only(55).
The second algorithm step is designed to extract CDR3 sequences wherever present. This algorithm was implemented in MATLAB, guided by previous publications(56), and using a regular-expression (regexp) based search algorithm.
The third step combined the above alignment and CDR3 data (where present), to decide whether a given read contains a TRGR. To do this, one of two decision approaches is used: if a CDR3 is identified in a read, a heuristic approach is employed to decide if the BWA-alignment reference genes could be rearranged within the same locus; the second, in the event that a CDR3 is not detected, relies on the SWA-determined alignment scores to determine if a given combination of V(D)J genes is present.
Bioinformatic Target Enrichment (Burrows-Wheeler-Alignment Algorithm)
Much like the technical aspects of the NTRA function to enrich TR genes at the DNA level, so too can an informatics target-enrichment approach be employed. Using the BWA algorithm(55), a series of FASTQ-formatted reads are first mapped relative to a reference index of IMGT TR V and J genes. Any reads containing sequence mapping to any of the reference genes are flagged as such in the SAM-formatted output file as mapped, whereas those not containing any TR V or J gene mapped sequence are assigned the SAM Flag 4. In this context, unmapped reads are unlikely to contain any detectable TR V(D)J gene rearrangements; this predicate is logical inasmuch as sufficient residual germline sequence of a TR V and/or J gene are required in a read to permit TRGR detection.
Reads-on-target and gene-coverage estimates are also derived using the BWA algorithm, since NTRA input probes consist only of TR V and J genes; this measure is calculated as a percentage of the number of unique reads mapped to the IMGT reference TR V and J gene indices relative to the total number of reads in the input FASTQ-formatted file.
CDR3 Sequence Extraction and SWA Alignment
This part of the TRSeq algorithm was implemented in MATLAB using strategies similar to those employed by the IMGT56-58). The IMGTN-QUEST system utilizes a CDR3 sequence extraction algorithm(57,59) and an SWA(60) algorithm performed against the IMGT reference sequences; the IMGT algorithms are all implemented in JAVA and processing is performed on IMGT servers.
As highlighted previously, we were unable to rely solely on the IMGT system for informatics results for several reasons: (1) the export of patient sequence data to an external non-secured network can be risky if insufficiently censored identifying metadata are also included; (2) the IMGT/High V-Quest system has a 500,000 sequence input limit (which may be substantially less than the number of sequence reads that need to be analyzed in the run of even a single high-throughput sequencing run); and (3) the queueing used by the IMGT can be lengthy, requiring a wait of possibly several days for sequence interpretation to begin.
A MATLAB implementation was chosen for convenience, programming familiarity, and because of easy vectorization, parallel computation and object-oriented programming capabilities. In addition, the MATLAB programming and command-line environments are able to easily incorporate UNIX and PERL-based scripts, including the BWA(Li, 2009) and CIRCOS software(61) suites, respectively.
The full coding of the analysis algorithm is presented in Appendix 2.6.2. The MATLAB code was written to accommodate FASTQ-formatted data, align each read using BWA to the reference TR V and J gene germline sequences, index the resultant data, test each indexed read for (and extract if present) a CDR3 sequence (using the uniformly present C-X(5 . . . 21)-F/W-G-X-G amino acid motif, per the IMGT canonical sequence motif(62,63)), and perform either an heuristic or SWA alignment-based validation of the reads mapped by BWA as evidence of a rearrangement within the read in question.
The SWA algorithm produces an optimal local alignment(60,64) of two co-input sequences (in this case, a query sequence relative to an IMGT reference sequence), and provides an alignment score (a unit-less measure of the degree to which the alignment perfectly matches an input sequence to its co-input sequence). For the purpose of this instance of the algorithm, for any case in which multiple possible alignments were produced, the alphabetical highest-scoring alignment was selected as the “correct” alignment, provided that this score was at least greater than the minimum cut-off score.
The minimum SWA alignment cut-off score was empirically determined for each of the three V, D, and J-gene gene groups using a large set of confirmed-negative sequences evaluated using the IMGT/HighV-QUEST system(56,57). The MATLAB code required for implementation of this algorithm is outlined in Appendix 2.6.1. A “practice” set obtained from the IMGT database(65,66) was also employed to test the pipeline, consisting of IMGT PCR-confirmed TRGR sequences with known V-D-J combinations and CDR3 sequences (see Section 3.1.3 for results of this practice set analysis).
Analytical Validation
A selection of 10 “First-Run” samples formed the basis of the analytical validation. These samples included 6 de-identified actual patient samples, obtained from flow-sorted peripheral blood specimens, tumour-infiltrating lymphocyte populations or in vitro cultures of lymphocytes. These samples were each subjected to flow-cytometric evaluation and cell-counting for basic immunophenotyping and cell-input consistency. In addition, four cell lines with known and well-described TR gene rearrangements (based on references cited by the IMGT database(67)) were also included (i.e. Jurkat (Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ) ACC-282), SUPT1 (American Type Culture Collection (ATCC) CRL-1942), CEM (ATCC CCL-119) and MOLT4 (ATCC CRL-1582)).
A three-part analytical validation approach was employed. First, the results obtainable by analysis of the sequencing data using the IMGT/High V-Quest pipeline were directly compared with the results of the TRSeq pipeline. Next, a PCR & Gel Electrophoresis experiment was designed to confirm the presence of the upper 90th centile of rearrangement configurations. Finally, the predominant rearrangements with accompanying TRSeq-identified CDR3 sequences were further Sanger-sequenced to validate this latter component of the NTRA analysis.
Comparison with IMGT Results
Given the limited input size capacity of the IMGT/High V-Quest system, a read-by-read comparison of a 10% random subset of the NTRA sequencing data was performed. From the IMGT analysis, a read was assumed to contain evidence of a rearrangement when the IMGT pipeline Junction analysis yielded an in-frame result. In addition, a read-by-read comparison of the alignment results (by gene name, for all V, D and J genes) was also performed.
PCR & Gel Electrophoresis Validation
A PCR-based experiment was deemed a reasonable orthogonal validation approach, given the gold standard BIOMED-2 assay methodology. Knowing that the number of possible rearrangements detected by the NTRA might be substantially large, the PCR validation was arbitrarily limited to those TRSeq-detected rearrangements in the upper 90th centile (i.e. percent rearrangement of greater than 10% of total rearrangements). Given this restriction, however, to ensure an adequate denominator of reactions for comparative purposes, all PCR validation experiments were uniformly performed across all 10 first-run samples.
PCR validation primer sets were constructed modeling the standard V-D-J orientation of rearranged TR genes; specifically, the PCR forward primer was set in the V gene and the reverse primer set in the anti-sense strand of the J gene. For each TRSeq-identified rearrangement above 10% of total rearrangements, the V and J genes were identified and the IMGT primer set database searched for gene (not allele) specific primers. While the IMGT primer database did contain a number of suggested primers, many of the TR genes did not have an available appertaining primer. As a result, where necessary, the anticipated rearrangement sequence (containing the V gene sequence artificially positioned before the J gene) was used to derive custom primers using the NCBI Primer-Blast tool(68). Careful attention was paid to ensure that each resulting theoretical PCR product length was at least 100 bp (the lower limit of fragment size reliably detectable by standard gel electrophoresis) and that a sufficient amount of the anticipated CDR3 region sequence would be preserved in the PCR product. In addition, the theoretical product length was recorded as an approximate size reference for analysis of the resulting electrophoresis migration patterns.
All putative primer pairs were then re-submitted to Primer-Blast M to assess for the possibility of non-specific products; the final set of putative primers pairs was also evaluated using the UCSC in silico PCR algorithm6) to confirm that no germline configuration products of less than 4 kb might be produced. Primer set physicochemical characteristics were evaluated using the IDT OligoAnalyzer Tool (v 3.1); Clustal W(55) alignments were used to identify significant primer sequence overlaps (Clustal W alignments note significant overlap of the TRGJ1 and TRGJ2 primers. This overlap was considered acceptable in order to define which of the TRGJ1 and TRGJ2 genes were present (given the presence of 5′ end non-homology). Since the PCR/electrophoresis results suggested the presence of both TRGJ1 and TRGJ2 positive products, the dominant TRGJ1 primer was selected for subsequent analyses and the TRGJ2 results excluded). The final primer-set sequences are listed in Table 2.2.
Custom primer set production was performed commercially by IDT and the forward and reverse primers were then mixed according to the design outlined in Appendix 2.7.2. PCR was performed in a 384-well plate on an Applied Biosystems VERITI thermal cycler using the Thermo Scientific 2×REDDYMIX PCR Master Mix kit according to the manufacturer's instructions; several control reactions were included, as highlighted in Appendix 2.7.2. Gel electrophoresis was performed in a 96-well Bio-Rad SUB-CELL Agarose Gel Electrophoresis System (necessitating 4 separate runs); electrophoretic migration was referenced against an Invitrogen TRACKIT 1 kb DNA ladder and visualized using ethidium bromide fluorescence, photographed in an ALPHAIMAGER Gel Imaging System. Electropherograms were digitally rendered, adjusted and composited using ADOBE PHOTOSHOP CC 2014. The resulting electrophoretic results were used in Receiver-Operating Characteristic (ROC) curve analyses relative to the corresponding TRSeq normalized read counts.
Sanger Sequencing Validadon
Based on the results of the above PCR & Gel Electrophoresis experiment, rearrangement-positive PCR products were purified using a QIAQUICK Spin PCR purification kit (100 bp to 1 kb range) according to the manufacturer's instructions (Appendix 2.7.3). Purified PCR products were then quantified by QUBIT and 20 ng equivalent aliquots were taken (with an additional volume reduction step using a SPEEDVAC, as required, for large volumes). The corresponding primer of the original primer pair with the lowest melting point was then selected for the purposes of single-direction Sanger Sequencing (performed at the TCGA Sick Kids Hospital Sequencing Facility).
The resulting sequencing results were analyzed using the FinchTV v 1.4 software suite, with corrections to sequencing error and reverse-complement sequence corrections performed manually as required. The originating TRSeq CDR3 sequences were then compared to the “reference” Sanger Sequence result. This comparison was performed in two ways: first, a basic multi-alignment comparison was performed (using the multialign algorithm of the MATLAB Bioinformatics Toolbox); second, a k-mer based PHRED-quality adjusted comparison was performed.
For the k-mer based approach, for a given V and J gene configuration, the most frequently detected TRSeq CDR3 sequences were aligned to the corresponding Sanger Sequencing result. In this context the Sanger Sequencing results were taken to represent a “consensus” of sequence data produced over all possible V and J configuration CDR3 sequences for that V-J gene configuration (reflecting the possibility of variable TRGR subclones). As such, in order to adjust the Sanger sequencing results to account for the potential alignment of a non-dominant subclone, a quality-based alignment algorithm was employed, based on the methods of(70). Each input TRSeq CDR3 sequence was aligned along a progressive series of k-mers of the Sanger sequence using a custom quality-based alignment algorithm (code outlined in Appendix 2.8). For each alignment result, if the optimal alignment score occurred within the expected sequencing region (thereby representing an optimal alignment within a region of Sanger sequence expected to contain the actual CDR3 based on flanking primer sets), as outlined in Table 3.1A, the CDR3 sequence was classified as correct (and vice-versa). This classification was then used to perform ROC analysis to determine what number of TRSeq CDR3 sequence read counts might be considered a validated cut-off.
Coverage Analysis
In addition to the above validation results, more detailed assessment of NTRA technical performance was also performed. Specifically, given that the NTRA relies on target enrichment, an assessment of the gene coverage of the NTRA was required. In addition, given that much of the utility of the NTRA might relate to identifying clonal cell populations, it was necessary to assess the dynamic sensitivity of the NTRA to decreasing numbers of cells bearing specific TR gene rearrangement configurations and, conversely, assess how standardized read counts might correlate with approximate input cell numbers.
Coverage Dynamics by Specimen Clonality
Given the nature of TRGR, by which genomic components are excised upon rearrangement, we evaluated the coverage dynamics across the first-run specimens. This analysis served not only as a mean of qualitatively comparing how V and J gene coverage might be expected to vary in specific types of specimens, but also to evaluate which coverage metrics might be most predictive of specimen type (i.e. clonal vs not) and what specific cut-off criteria might be used to this effect. To do this, ROC-based analyses of mean overall and locus-specific coverage data for V and J genes was performed, as well as percent genes at least 100×for each of V and J gene types.
Negative Control Coverage Assessment
For the purposes of this project, a fully germline TR gene configuration was approximated using a cell lines of embryonic origin and a cell line that has been fully sequenced without any known/reported TR gene derangements. The former scenario was approximated using the HEK293 cell line (an embryonic kidney cell line; ATCC CRL-1573) and the latter using a Coriell cell line (whose genome has been well-characterized and is not known to contain TR rearrangements). Use of the latter cell line was incorporated given that, in our hands, this cell line had been previously and purposefully degraded by FFPE treatment, representing a scenario of TR gene coverage assessment in the context of degraded DNA.
Total genomic DNA was extracted from previously cultured HEK293 cells and FFPE treated Coriell cell cultures and subsequently subjected to the NTRA, as outlined in Appendices 2.1 to 2.5. Standard TRSeq analyses were performed for each sample, with special deference paid to the coverage results.
Dilution Series
A rigorous dilution series experiment, in the context of this project, might involve a flow-sort spike of cells with a known TR gene configuration into a population previously determined to be “polyclonal”; this might be approximated, for example, using a well-characterized cell line spiked into a population of lymphocytes obtained from normal blood. Rather than undertaking this more complex and expensive approach, an approximation of this dilution experiment was undertaken with DNA obtained from the Jurkat cell line spiked into a known-polyclonal lymphocyte population DNA isolate (the A037 sample; see Results section 3.2). Specifically, Jurkat DNA was spiked in at log-decrements (as outlined in Table 2.3) based on a lymphocyte total DNA complement assumed to be 0.7 pg, given the results of previous publications(71-73). The total DNA of each sample in the dilution series was verified (and compared to expected values) using a QUBIT assay; the samples were then subjected to the NTRA, as outlined in Appendices 2.1 to 2.5. Standard TRSeq analyses were performed, with special deference to changes in the raw read counts of Jurkat-specific TRGR configurations across the dilution series.
Alternative Method and Algorithm
Hybrid-Capture Protocol
For T cell receptor (TCR) diversity and clonality analyses we investigated genomic DNA isolated from flow sorted T cells isolated by affinity magnetic bead isolation, peripheral blood mononuclear cells (PBMC) isolated from blood by density gradient separation, cell-free plasma DNA extracted from blood, or scraped and pelleted immortalized cell lines.
Isolated DNA is sheared to ˜275 bp fragments by sonication in 130 uL volumes (Covaris). DNA libraries are generated for illumina platform sequencing from 100-1000 ng of sheared DNA by ligation of sequencing library adaptors (NEXTFLEX) using the KAPA library preparation kit with standard conditions. Libraries are visually assessed (Agilent TAPESTATION) and quantified (QUBIT) for quality.
Hybridization with probes specifically targeting the V and J genes is performed under standard SEQCAP (Roche) conditions with XGEN blocking oligos (IDT) and human cot-1 blocking DNA (Invitrogen). Hybridization is performed either at 65 C overnight. The target capture panel consists of 598 probes (IDT) targeting the 3′ and 5′ 100 bp of all TCR V gene regions, and 95 probes targeting the 5′ 100 bp of all TCR J gene regions as annotated by IMGT (four loci, 1.8 Mb, total targeted 36 kb). Hybridization and capture can be performed as a single step with a combined V/J panel, as a single step with only the V panel, or as a three step process when non-rearranged fragment depletion is desired consisting of a V capture, then depletion, then J capture.
For depletion of non-rearranged fragments 500 ng-1000 ng of library is depleted by hybridization with a panel of 137 probes (IDT) targeting the 5′ 120 bp of selected TCR V gene region 3′ untranslated regions as annotated by IMGT (four loci, 1.8 Mb, total targeted 16.5 kb) and 131 probes (IDT) targeting the 5′ 120 bp of selected Ig V gene region 3′ untranslated regions as annotated by IMGT (three loci, 3.1 Mb, total targeted 15.7 kb). A modified and truncated SEQCAP protocol is employed wherein following incubation with M-270 streptavidin linked magnetic beads (Invitrogen), the hybridization reaction is diluted with wash buffer I, beads are discarded and the supernatant is cleaned up by standard Agencourt AMPURE XP SPRI bead purification (Beckman).
Algorithm
A custom Bash/Python/R pipeline is employed for analysis of paired read sequencing data generated by Illumina NEXTSEQ 2500 instrument from the hybrid-capture products. Referring to
(1) 150 bp paired-end reads are merged using PEAR 0.9.6 with a 25 bp overlap parameter. This results in an approximate 275 bp sequence for each fragment and enhances the sensitivity of V,J,D gene detection using the subsequent search strategies.
(2) Individual BLAST databases are created using all annotated V, D, J gene segments from IMGT. These full-length gene sequences are the targets of the hybrid-capture probe panel. Individual merged reads are iteratively aligned using BLASTn with an e value cut-off of 1 to the V database, J database then D database with word size of 5 for D segment queries. Trimming of identified V or J segments in the query sequence is performed prior to subsequent alignment to reduce false positives and increase specificity, particularly for the D gene query.
(3) In order to identify CDR3 sequences, the V/J junction position is extracted from the previous search data for those fragments containing both a V and J search result. 80 bp of DNA sequence flanking this junction is translated to amino acid sequence in all six open reading frames and sequences lacking stop codons are searched for invariable anchor residues using regular expressions specific for each TCR class as determined by sequence alignments of polyclonal hybrid-captured data from a healthy patient as well as TCR polypeptides annotated by IMGT.
(4) Calculation of capture efficiency (on-target/off-target capture ratio) is performed by aligning all recovered, merged reads to the human genome (BWA) and dividing the number of reads aligning to the TCR loci by the total number of reads. The total number of unique TCR clones is determined by finding the unique minimum set of V/J combinations and the number of occurrences of each is tabulated. This data is visualized using R as stacked bar charts to generate figures that can be quickly visually assessed on a sample-by-sample basis for monoclonal or polyclonal signatures or clinically relevant enrichment of particular clones.
Application of the Algorithm to Existing Sequencing Data
The custom pipeline is not dependent on our hybrid-capture protocol and can be performed on non-target captured whole genome or RNA-seq data. In this situation, an in silico capture is performed by extracting reads aligning to the four TCR loci (7:38250000-38450000, 7:141950000-142550000, 14:22000000-23100000) or Ig loci (chr2:89,100,000-90,350,000, chr14:106,400,000-107,300,000, chr22:22,350,000-23,300,000) from DNA (BWA) or RNA (STAR) sequence data (SamTools), followed by paired-end nucleotide sequencing data extraction (PicardTools). These reads are then inserted in to the previously described computational pipeline.
Results and Discussion
Informatics
Insert Length Simulation
The DNA Insert Length Simulation results were generated (data not shown). The analysis suggested a plateau of sensitivity of greater than 99.1% reached after 182 bp. For convenience, an adequately “evidence-based” insert length and informatics read length goal of 200 bp was chosen for the NTRA.
After further analysis excluded extra-locus V-D-J gene combinations (i.e. combinations not likely to result from rearrangements within the same TR locus), the number of missed combinations was reduced from 1752 to 80.
From among the above 80 intra-locus combinations, missed rearrangements originated only from among the TRB and TRG loci, with particular enrichment of TRBV6-2*01 and TRBV6-3*01 within the former (65 of 80) and enrichment of the TRGJ1*02 within the latter (15 of 80).
Analysis by phylogenetic sequence alignment (using the SWA alignment algorithm) within the TRBV6 group showed significant cophenetic linkage between the TRBV6-2*01 and TRBV6-3*01 genes (data not shown). Similarly, analysis by phylogenetic sequence alignment within the TRGJ gene group suggested significant cophenetic linkage between TRGJ1*02 and TRGJ2*01 (data not shown). These results suggest that combinations within the artificial read set involving either of these TRBV genes were likely misaligned to another TRBV gene (likely the next closest cophenetic “cousin,” TRBV6-2*02) and that the TRGJ1*02 gene was likely misaligned to the TRGJ1*01 gene. Of note, the observation of closer cophenetic linkage between TRBV6-2*01 and TRBV6-3*01 rather than between TRBV6-2*01 and TRBV6-2*02 (as would be expected for two alleles of the same TR gene) and of closer cophenetic linkage between TRGJ1*02 and TRGJ2*01 rather than between TRGJ1*01 and TRGJ1*02, suggests error on the part of the IMGT classification.
MATLAB SWA Score Cut-Off Determination
The results of the empirical V, D and J-gene MATLAB alignment score cut-off score experiment were generated (data not shown). This experiment employed the code presented in Appendix 2.6.1 run on a test set of 91375 Illumina sequencing reads obtained from anonymized myeloid leukemia samples enriched for sequences outside of the IG/TR loci. These sequences were “confirmed” negative for V, D, and J gene sequences using the IMGT/High V-QUEST system (Brochet et al., 2008; Giudicelli et al., 2011). Given an experimental number of sequencing reads of at least 1 million, a 6-sigma cut-off score for MATLAB TRSeq analysis suggests 53.23 for the V genes; 19.02 for the D genes; and 34.43 for the J genes. It is easily observed that the cut-off values increase respectively from D, to J, to V genes; this observation parallels the mean length of the reference sequences from D to J to V genes.
TRSeq Analysis of IMGT-Produced TRGR Sample Sequence Reads
A sample of 268 short read sequences was downloaded from the IMGT website. These sequences consist of a variety of previously characterized TR and IG gene rearrangements available for download in FASTA format. After re-formatting into FASTQ format (using arbitrary quality scores), the dataset was analyzed using the TRSeq pipeline. Of the 268 short read sequences, 55 were identified by the IMGT as containing TR genes (either V or J genes); to these reads, there was perfect (100%) TRSeq alignment concordance, both in relation to gene name and allele. The TRSeq algorithm identified 50 of the 55 reads as containing evidence of TRGR; the 5 remaining reads were identified by the IMGT as containing rearrangements within the TRD locus, each with a TRSeq CDR3 region correctly identified. These results suggest that the 5 TRSeq “false-negatives” were informatically rejected by the TRSeq algorithm based on insufficient TRD D-gene SWA alignment score values; this form of error is not alarming given the more stringent means by which the TRSeq SWA alignment score cut-off values were determined relative to the IMGT/High V-QUEST pipeline(55,58).
First-Run Results Summary
Table 2.5 outlines the flow-cytometric features of the 6 patient lymphocyte samples. These immunophenotypic features were in keeping with the lymphocyte sample sources of origin (also documented in Table 2.5), varying from normal patient peripheral blood mononuclear cells to highly immune-sensitized lymphocyte cultures from tumour infiltrating lymphocyte specimens. Notably, the A037 sample served as a model of a “polyclonal” lymphocyte population whereas, for the purposes of qualitative assessment at least, the L2D8 sample could be immunophenotypically interpreted as highly “clonal” in nature.
In addition, model “clonal” samples were included, consisting of the Jurkat, CEM, SUPT1 and MOLT4 cell lines. Table 2.6 lists the previously documented rearrangements, as cited in the IMGT database(67).
Prior to target enrichment and sequencing, adequate quality control was assured, as documented by pre and post-library preparation TAPESTATION tracings (data not shown). Post-target enrichment quality control was assured in the same manner.
Illumina NEXTSEQ sequencing was then performed on TAPESTATION-normalized pooled input target-enriched DNA. The appertaining read-pair FASTQ-formatted zipped files were decompressed and the PEAR paired-end merging algorithm was run with a minimum strand sequence overlap of 25 bp. A breakdown of the PEAR results were generated (data not shown). The resulting PEAR-merged FASTQ-formatted read files were input to the TRSeq pipeline.
TRSeq metadata for the first-run sample series were generated (data not shown), including input reads, reads-on-target, summary coverage statistics, and a histogram of read counts for the proportion of each locus contributing to identified TRGR's, respectively.
One important highlight is the variation in coverage seen across the 10 specimens relating to the D locus. As described in the introduction, since the D locus genes are sandwiched within the larger A locus, the D locus genes are often deleted upon A locus rearrangement. The coverage profiles of the D locus therefore paralleled this phenomenon with lower D locus coverage identified in the clearly clonal or oligoclonal samples relative to the polyclonal samples (e.g. 1.2D8 and cell line samples vs. A037 peripheral blood sample).
Composites of the circos plots obtained from the 10 first-run samples were generated (data not shown). Much as the coverage profiles differed across the samples (data not shown), the resulting circos plots demonstrated a clear aesthetic difference from polyclonal to clonal/oligoclonal samples, with emphasis on the number and relative width of the composite circos links (i.e. fewer and broader in width in the more clonal cases and vice versa). Also of note, the color distributions were distinctly different with the more polyclonal cases, containing a larger number of smaller-quantity “subclones” involving a more disparate number of TR genes.
Analytical Validation
IMGT/High V-Quest Comparison
Comparison of the IMGT/High V-Quest pipeline analysis to the TRSeq results were generated (data not shown). The degree of concordance of read-to-read interpretation with respect to identifiable rearrangements (as present or not identified) is excellent (99%), as is the degree of concordance of named D genes (99%). A lower degree of concordance is noted for named V and J genes (68% and 84%, respectively). These results may relate to different initial alignment algorithms employed, as well as different gene-identity cut-off values employed in the SWA algorithms of the IMGT/High V-Quest and TRSeq pipelines. In light of the results seen in Section 3.1.1, the possibility of V and J gene phylogenetic sequence misclassification in the publically-available IMGT sequence databases should also be considered as a possible contributing factor.
The high D-gene concordance relative to the V and J-gene values may relate to both the shorter reference sequences of the D-genes relative to the V and J genes, as well as the lower number of reference D-genes available for rearrangement. It is important to point out the possibility of a theoretical bias against D-gene identification in input reads, given that TRGR reads containing D-genes require 3 rather than 2 composite genes, which could be more difficult to detect in the context of restricted average read lengths. This consideration was brought to bear during the NTRA assay design phase (as described in Section 3.1.1), with the conclusion that adequate flanking 5′ and 3′ sequence would be available on average in the scenario of read input length of 200 bp or more to reliably identify reads containing V-D-J rearrangements.
PCR & Gel Electrophoresis
PCR primers were mixed and the results by Agarose gel electrophoresis were generated (data not shown). Note that results obtained from PCR reactions using the TRGJ2 reverse primer are excluded, as noted in Section 2.2.2. Two classification approaches may then be entertained, one based on dark-staining PCR bands only, and the other based on any staining (assuming bands to be of appropriate molecular weights, as set out in Table 3.1A). When these classifiers are compared with the read-count-normalized results of the TRSeq algorithm (as set out in Table 3.1A), the ROC curves of are obtained, respectively (data not shown). In the former scenario, the ROC Area-Under-the-Curve (AUC)=0.91 and p-value <0.001, with a TRSeq normalized read count of 6.7 or more. Based on the results, a less stringent classification results in a reduced AUC=0.71 and p-value <0.001, with a TRSeq normalized read count of 1.7 or more.
Sanger Sequencing Results
PCR reactions that were post-PCR purified were submitted for Sanger Sequencing. Alignment of each corresponding TRSeq CDR3 sequence (and associated raw read count) in relation to the manually-verified/corrected Sanger Sequencing Result were generated (data not shown); only those Sanger Sequencing specimens containing TRSeq-identified CDR3 regions, those of sufficient quality for interpretation, and those not rejected based on use of the TRGJ2 reverse primer were further considered.
There appears to be a trend for each distinct primer configuration inasmuch as TRSeq-identified CDR3 sequence configurations having sufficient associated read counts (data not shown), as suggested from Section 3.3.2, show the best contiguous alignments to the corresponding “reference” Sanger Sequences.
To better quantify this relationship, we utilized a k-mer based quality-score adjusted alignment analysis. For each relevant primer configuration, the corresponding CDR3 was aligned using PHRED-based quality-score adjustment across the length of the Sanger “reference” sequence. If the optimal alignment from this process was present within the sequence window in which a CDR3 was theoretically predicted to exist, the CDR3 read configuration was classified as “compatible.” The resulting classification analysis is represented by the ROC curve (AUC=0.832, p-value=0.006) (data not shown). Based on this analysis, the optimal TRSeq normalized read count cut-off is 4.9.
Coverage Analysis
Coverage Dynamics by Specimen Clonality
Using the qualitative data of Table 2.5, specimens were classified as either “clonal” or “polyclonal.” The resulting ROC curves for the various coverage metrics were prepared (data not shown). Of note, a mean V-gene coverage assessment of the gamma locus appeared to suggest the highest non-unity AUC. Further, the ROC analysis suggested that a mean V-gene coverage of greater than/equal to 4366.4 showed optimal sensitivity and specificity (86% and 67%, respectively) for predicting whether a specimen was unlikely to be clonal. Care should be taken not to use these cut-off points without additional validation, however, given the low number of data points constituting the analysis. Rather, these data stand to suggest a need for further evaluation of the potential predictability of “clonal” status derived from coverage analysis within the gamma locus.
Negative Control Coverage Assessment
The NTRA was tested on samples of previously cultured HEK293 and Coriell cell lines; these analyses aimed mainly at estimating coverage ceilings for the NTRA, but also served as added negative control specimens (i.e. specimens known or expected not to contain any TRGRs).
Applying the PEAR algorithm(52) (with a minimum 25 bp forward-reverse read overlap) resulted in pairing of 83% of input reads in the HEK293 sample and 90% of input reads in the Coriell sample.
In both instances, the number of subsequently identified TRGR configurations did not meet the TRSeq cut-off criteria (TRGRs were identified in 0 of 5,729,205 total input reads in the HEK293 cell line and only 7 of 2,761,466 total input reads in the Coriell cell line). This was in keeping with the anticipated fully-germline configuration of each of these non-lymphoid origin cell types.
For the HEK293 cell line, the percent V and J genes at or above 100×coverage was 100%; the overall TR V gene coverage averaged 29960x; and the overall TR J gene coverage averaged 8789x.
For the Coriell cell line, the percent V and J genes at or above 100×coverage was 100%; the overall TR V gene coverage averaged 13379x; and the overall TR J gene coverage averaged 3925x.
Dilution Series
A dilution experiment was performed at log-reduction intervals, set up according to the design of Table 2.3, and adjusted according to Table 3.2 to account for Jurkat DNA concentration discrepancies. Three Jurkat cell line unique TRGR configurations were selected for inter-dilution comparison, namely the TRAV8-4-TRAJ3, TRGV11-TRGJ1 and TRGV8-TRGJ2 rearrangements identified & confirmed in Section 3.3. The above configurations were confirmed absent in the polyclonal (A037) sample. In addition, each of these configurations showed a specific dominant CDR3 sequence.
The mean of the raw read-counts (i.e. not normalized) across the three tracked V-J configurations (with error bars for standard deviation) vs. expected approximate Jurkat cell numbers (with adjustments for significant digits) from Table 3.2 were generated (data not shown). An exponential trend line could be applied, with R-squared=0.9996.
Of note, when the extremum of the first dilution is excluded, the dilution curve is remarkably linear (data not shown), but with a positive slope. This suggests a linear direct correspondence between read count and number of cells bearing a given V-J configuration at low levels.
In contrast to the reliable low-level detection by way of V-J configuration, detection narrowed to absolute clonotype (by including the CDR3 sequence) was limited to only the first three dilution specimens (i.e. sensitivity down to an approximated 1 in 125 cells; data not shown).
This limited sensitivity speaks to the sensitivity of the TRSeq junction finder to sequencing error. Indeed, if even a single base is changed relative to the canonical regular expression required for detection of a CDR3 sequence, the junction finder will not identify the sequence correctly: likewise, any non-triplicate base insertion will not be detected as an in-frame CDR3 sequence. In contrast, since the TRSeq V and J gene enumeration scheme uses alignment-based algorithms, the TRSeq results relating to V and J gene enumeration are much more forgiving of higher the higher likelihood of sequencing error in clonotypes with low read counts, thus substantially improving the assay sensitivity for characteristically unique V-J gene configurations.
Support for these suppositions is echoed in part by previous work pertaining to core clonotype analyses(27). Indeed, when the proposed criteria of Bolotin, et. al.(27) for gathering low-level reads of similar but error-prone sequence into common core clonotypes are applied to the dilution experiment (implemented in Appendix 3), it is possible to identify reads comparable to the clonotypes described above in even the most dilute samples.
For example, running the code of Appendix 3 with the input core clonotype of the TRGV8-TRGJ2 configuration, and allowing for a maximum of 3 sequence mismatches, 3 or more reads of satisfactory clonotype can be identified in dilutions 2-5. If the number of sequence mismatches is increased to 4, reads of satisfactory clonotype can be identified in all dilutions (i.e. down to an estimated sensitivity of 1 in 185646 cells).
The importance of these results stems from the applicability of this form of core clonotype analysis to a more accurate identification of minimal-residual disease, for example, at very low levels with remarkable sensitivity, even in the absence of traditional primer-directed sequence enrichment(77).
NTRA—BIOMED-2 Comparison
In keeping with the general approach used to assess BIOMED-2 results, the NTRA TRB and TRG clonotype tables were analyzed to compare the ratio of the dominant clonotype read count relative to the “background” read count. The largest read count not satisfying the normalized TRSeq read count according to the results of Section 3.3 was taken as the background read count value; alternatively, in the case where the dominant clonotype did not satisfy the normalized TRSeq read count cut-off of Section 3.3, the next largest clonotype read count was taken as “background”. From among each of the TRB and TRG loci, the largest dominant clonotype-to-background ratios were compared to the overall BIOMED-2 results using a ROC analysis.
The ROC analysis result could be classified as “good” (78) with AUC=0.82, p-value <0.001 (data not shown). Of note, this AUC value appears comparable to those observed in Section 3.3. Of even more impressive note is that the ROC-suggested dominant clonotype-to-background cut-off value was also comparable to that outlined in the current BIOMED-2 TRGR assay interpretation guidelines(79); indeed, the ROC analysis-suggested value of 3.4, which is effectively the median value of the “indeterminate” range of dominant peak-to-background ratios recommended for BIOMED-2 result interpretation(79).
Interestingly, when the above process was broken down into two separate comparisons of the TRB and TRG loci, the TRG locus was found to be the significant driver: the TRG locus comparison alone yielded a ROC AUC=0.81 (p-value <0.001) whereas the TRB locus comparison alone yielded a ROC AUC=0.60 (p-value=0.17).
NTRA Coverage Metrics—BIOMED-2 Comparison
As in Section 3.4, an analysis of coverage variation in relating to clonal status was undertaken. In contrast to the results of Section 3.4, a far less significant series of areas-under-the-curve were observed from this analysis. The greatest AUC was noted by analysis of mean V-gene coverage (i.e. mean V-gene coverage over all four loci) with AUC=0.59, p-value=0.213.
Furthermore, the data from Section 3.4 suggested that analysis of coverage from the Gamma locus might be predictive of clonal status. Unfortunately, these hypotheses were not substantiated by way of the clinical validation set, from which the AUC for the TRG locus V-gene analysis and TRG locus J-gene analysis were 0.59 and 0.57, respectively.
The clear discordance between these results and those of Section 3.4 likely relates to several factors. First, the sample size in Section 3.4 is one-sixth that of the clinical validation set, making the results of Section 3.4 much more vulnerable to the effects of outliers. Second, the overall coverage in the analytical validation set was lower, owing to base-output restrictions using the mid-output NEXTSEQ kit; as such, coverage correlations made in Section 3.4 might not necessarily be applicable to experiments performed using the high-output NEXTSEQ kit. Thirdly, the clinical validation experiment was not subject to bias of assumption as to the clonality of each input specimen; rather clonality was specifically assayed using an orthogonal method.
Described above is the first hybrid-capture-based T-cell clonality assay designed to assess clonality and provide clonotype data over all four T-cell gene loci. For this purpose, a custom MATLAB-based analysis pipeline was implemented using optimized object-oriented programming integrating the ultra-fast BWA alignment system and the aesthetically-pleasing circos-based genomic data visualization suite. The latter visualization was designed with current methods in mind, in which electropherographic plots serve as the primary means by which clonotypes are visualized.
Advantages of NTRA Over Traditional T-Cell Clonality Testing Assays
Not only can the NTRA identify clonotypes from all four loci, the use of hybrid capture makes the process platform-agnostic. The laboratory work-flow can be integrated into any standard library preparation work-flow with the addition of a single hybridization step, capable of enriching for sequences containing T-cell genes of a several specimens at a time. In addition, as part of laboratory work-flows already using a hybrid-capture approach for other purposes, the probes used as part of the NTRA are amenable to “spike-in” combined hybridization reactions, provided that there is no significant probe-set sequence overlap or complementarity.
In comparison to the current BIOMED-2 based donality assays, the NTRA adds a dearth of extra data, especially as pertaining to clonotype data from the gene-rich alpha-locus. This locus has traditionally been too diffusely distributed within the genome to be amenable to primer-based amplification, a challenge easily overcome using a hybrid-capture approach. Akin to the requirements of the IMGT, the NTRA outputs a clonotype table containing data specific to the best aligned allele. In contrast, however, visualized data is restricted to gene-level only, thereby providing a means of visualization comparable to electropherographic output. In addition, included with the latter, is the in-frame CDR3 sequence (where detected), data currently not available using either standard PCR-based techniques or the mainstream sequencing-based solutions (e.g. Invivoscribe).
In addition to validating the wet-bench and informatics using a number of orthogonal approaches, the NTRA was also shown to be theoretically sensitive to low-level clonotypes. This latter observation is an important boon to the hybrid-capture approach, suggesting that carefully performed hybrid-capture methods can provide signal amplification comparable to flow-cytometric(81) and molecular approaches(32)(82)(83).
Assay Cost & Efficiency Considerations
As highlighted in Section 3.8, the assay may be considered cost effective, depending on the specific scenario of interest. In addition, the use of a hybrid-capture approach allows for spike-ins of additional probes for other genomic regions of interest. This allows the possibility of running multiple assays from a single library preparation step, requiring only bioinformatic separation of the resulting enriched sequences.
Applications
Assessment of lymphocyte clonality is integral to the diagnosis of diseases and cancer affecting the immune system. In addition, sequencing of the T-cell repertoire of a patient has gained clinical value with the recent understanding of T-cell mediated recognition and destruction of neoplasms. Further, the development of adoptive cell therapy and recombinatorial engineering of T-cell receptors requires high-throughput molecular characterization of in vitro T-cell populations before transplant. PCR-based methods such as BIOMED-2 and Immunoseq are currently in use for TCR characterization however their costs and complexity remain barriers for clinical deployment requiring high-throughput multi-patient, multi-sample work-flows at low cost. We have therefore developed a hybrid-capture-based method that recovers rearranged TCR sequences of heavy and light TCR chains from all four classes in one tube per sample at low cost. TCR clonality and CDR3 prevalence can be rapidly assessed in a three-day turn-around time with an automated pipeline generating summary figures that can be rapidly assessed by clinicians.
Adaptive T-cell immunotherapy has become a field of great interest in the treatment of multiple solid-tumor cancer types. Non-childhood cancers, particularly those linked to chronic exposure of known carcinogens, are driven by the accumulation of mutations. Some of these mutations drive pro-tumorigenic changes, while others result in non-tumorigenic changes to proteins expressed by the carrier cell. During normal protein turnover these modified proteins are broken down in to short polypeptides and make their way to the surface of the cell in association with molecular surveillance molecules (MHC I). In this context these modified polypeptides are recognized as foreign neo-antigens by the host immune system, and in the context of other signals, lead to the activation of T-cells that direct the destruction of cells expressing these modified proteins.
It is now understood that many solid-tumours exist in a state where their presence recruits neo-antigen specific T-cell lymphocytes to the margins however further advance and effective destruction of the tumor is prevented by expression of checkpoint inhibition molecules on the tumor cell surfaces. Therefore immunotherapy has become a major area of advance in cancer therapy wherein such checkpoint inhibition molecules are masked through transfusion of antibodies. This allows recognition of tumor and its destruction by neo-antigen specific T cells. In order to further enhance such anti-tumor activity, tumor infiltrating lymphocytes (TIL) can be isolated from tumor biopsies and expanded in vitro, followed by subsequent transfusion in great numbers back in to the patient following immunodepletion to enhance transplant colonization thereby driving a durable antitumor response.
T-cell lymphocytes are fundamental to this process, however due to their exquisite specificity, only neo-antigen specific T-cells are capable of driving anti-tumor activity. As a result there is a need for molecular characterization of circulating T-cells in the patient before and after treatment, infiltrating T-cells in the tumor before and after treatment, and screening of expanded populations in vitro for safety and efficacy. Our method provides a high-throughput, low cost and rapid turn-around method for T-cell receptor characterization in order to facilitate clinical deployment and uptake of adoptive cell transfer immunotherapy.
This method is not only of use in immunotherapy applications, as any disease involving expansion of T-cell clones would benefit from its use. The symptoms of autoimmune diseases are driven largely by T-cell mediated cytotoxicity of “self” tissue and therefore the identification and expansion of specific T-cell clones can be monitored using this method. This method would also be useful to follow immune challenges such as infection or immunization in the development of anti-infectives or vaccines.
There is also described herein a laboratory and bioinformatic workflow for targeted hybrid-capture enrichment of T-cell receptor loci followed by Illumina sequencing to assess the clonality of a range of specimens with variable T-cell clonal complexity as well as a set of 63 T-cell isolates referred for clinical testing at our institution.
Methods and Materials
Probe design—All annotated V, D, J gene segments were retrieved from the IMGT/LIGM-DB website (www.imgt.org9). The 100 bp of annotated 3′ V gene coding regions and up to 100 bp, when available, of annotated 5′ J gene coding regions were selected as baits. Probes with duplicate sequences were not included.
DNA isolation—CD3+ T cells were isolated by flow assisted cell sorting of PBMC populations separated from whole blood. Peripheral blood mononuclear cells (PBMC) were isolated from whole blood by centrifugation followed by DNA isolation with a GENTRA PUREGENE kit (Qiagen) according to manufacturer protocol. In the case of fresh/frozen tissues, a QIAGEN ALLPREP (Qiagen) kit was employed, according to the manufacturer's instructions. In contrast, for FFPE samples a previously optimized in-house approach was used. First, sample FFPE tissue blocks were cored with a sterilized TISSUE-TEK QUICK-RAY punch (Sakura) in a pre-selected area of representative tissue; alternatively, under sterile conditions, 10×10 μm DNA curls/unstained slides were obtained for each submitted block of FFPE tissue. In a fumehood, 400-1000 μL xylene was aliquot into each tube (volume increased for larger FFPE fragments), followed by vigorous vortexing for 10 sec, incubation in a 65° C. water bath for 5 min, and centrifugation at 13200 rpm for 2 min. The supernatant was then discarded and step an additional xylene treatment step was performed. Subsequently, addition of 400-1000 μL ethanol (volume adjusted for larger input tissue volumes) was performed, followed by vigorous vortexing for 10 sec, and centrifugation at 13200 rpm for 2 min. The supernatant was then discarded and the ethanol treatment step repeated. The resulting pellet was then dried using a SPEEDVAC (Thermo Scientific) for 5 min, after which 150 μL of QIAAMP buffer ATL (Qiagen) was added, followed by 48-hour incubation at 65° C. with 50-150 μL of proteinase K (volume increased for higher input volumes). A final ethanol clean-up step was performed, as above, to produce a purified DNA product. Resuspension in TE buffer (Qiagen) was then performed.
Hybrid capture—Isolated genomic DNA was diluted in TE buffer to 130 uL volumes. Shearing to ˜275 bp was then performed on either a Covaris M220 Focused-ultrasonicator or E220 Focused-ultrasonicator, depending on sample throughput, with the following settings: for a sample volume of 130 μL and desired peak length of 200 bp, Peak Incident Power was set to 175 W; duty factor was set to 10%; cycles per burst was set to 200; treatment time was set to 180 s. In addition, temperature and water levels were carefully held to manufacturer's recommendations given the instrument in use.
Illumina DNA libraries were generated from 100-1000 ng of fragmented DNA using the KAPA HYPERPREP Kit (Sigma) library preparation kit following manufacturer's protocol version 5.16 employing NEXTFLEX sequencing library adapters (BIOO Scientific). Library fragment size distribution was determined using the Agilent TAPESTATION D1000 kit and quantified by fluorometry using the Invitrogen QUBIT.
Hybridization with probes specifically targeting V and J loci (Supplemental Table 3) was performed under standard SEQCAP (Roche) conditions with XGEN blocking oligos (IDT) and human Cot-1 blocking DNA (Invitrogen). Hybridization is performed either at 65 C overnight. The target capture panel consists of 598 probes (IDT) targeting the 3′ and 5′ 100 bp of all TR V gene regions, and 95 probes targeting the 5′ 100 bp of all TR J gene regions as annotated by IMGT (four loci, 1.8 Mb, total targeted 36 kb).
Capture Analysis—A custom Bash/Python/R pipeline was employed for analysis of paired read sequencing data generated by Illumina NEXTSEQ 2500 instrument from the hybrid-capture products. First, 150 bp paired reads were merged using PEAR 0.9.6 with a 25 bp overlap parameterA18. This results in a single 275 bp sequence for each sequenced fragment. Next, specific V, J, and D genes within the fragment sequence were identified by aligning regions against a reference sequence database. Specifically, individual BLAST databases were created using all annotated V, D, J gene segments retrieved from the IMGT/LIGM-DB website (www.imgt.orgA9), as these full-length gene sequences were the source of probes used to design the hybrid-capture probe panel. Individual merged reads are iteratively aligned using BLASTn with an e value cut-off of 1 to the V database, J database then D database with word size of 5 for D segment queriesA19. Trimming of identified V or J segments in the query sequence is performed prior to subsequent alignment. From reads containing V and J sequences, we identified V/J junction position and the antigen specificity determining Complementarity Determining Region 3 (CDR3) sequences. In order to identify CDR3 sequences, the V/J junction position is extracted from the previous search data for those fragments containing both a V and J search result. 80 bp of DNA sequence flanking this junction is translated to amino acid sequence in all six open reading frames and sequences lacking stop codons are searched for invariable anchor residues using regular expressions specific for each TR class as determined by sequence alignments of polyclonal hybrid-captured data from rearranged TR polypeptides annotated by IMGT9
Results and Discussion
The CapTCR-seq method employs hybrid capture biotinylated probe sets designed based on all unique Variable (V) gene and Joining (J) gene annotations retrieved from the IMGT database version 1.1, LIGMDB_V129. These probe sets specifically target the 3′ regions of V gene coding regions and the 5′ regions of J gene coding regions that together flank the short Diversity (D) gene fragment in heavy chain encoding loci and which together form the antigen specificity conferring CDR3 (
To identify V(D)J rearrangements from the pool of captured V and J sequences, we used a computational method that performed: (1) Read merging to collapse paired reads in to a single long-read sequence to enhance V(D)J and CDR3 identification, (2) progressive BLASTn-based V, J and D detection utilizing iterative end trimming and (3) CDR3 scoring using regular expression pattern matching (
We employed this method to identify V(D)J rearrangements and CDR3 sequences in PBMCs isolated from a healthy human. With a single step hybridization and capture reaction employing the probe panel targeting TCR V genes, the number of detected unique VJ rearrangements increased with increasing amount of sample genomic DNA used to generate the initial library, with 52 times more rearrangements detected with an input of 1,000 ng compared with 100 ng (1925 vs 37) (
To test the ability of CapTCR-seq to assess TCR clonality of samples with a range of clonal signatures, we analyzed libraries derived from CD3+ flow-sorted Tumor Infiltrating Lymphocytes (TIL) expanded cultures (oligoclonal) and lymphoblast cell lines (clonal) (
To demonstrate the potential clinical utility of our approach, we generated DNA sequencing libraries from an unselected cohort of 63 samples submitted for clinical T-cell receptor rearrangement testing and subjected these to capture, sequencing and analysis (Supplemental Table 1). Samples were found to have varying degrees of clonality, with the predominant CDR3 sequence representing up to 40% of the most clonal sample (average 12.2%; median 6.3%%, range 0.8-100%, data not shown). When a clonal population was defined as having the most abundant to third most abundant rearrangements observed at two or more times the level of the next most abundant rearrangement, we observed three groups of samples: 11 with clonal enrichment of both beta and gamma rearrangements, 12 with clonal enrichment of beta or gamma rearrangements, and 41 that were polyclonal for both beta and gamma. When 61 of these samples were assessed by BIOMED2 assay we observed 73% agreement for beta (44/60) and 77% for gamma (46/60), 60% of samples were in agreement for both beta and gamma clonality measures (36/60). For the beta locus, 13 samples that were scored as clonal by BIOMED2 were scored as polyclonal based on relative prevalence when assessed by hybrid capture profiling. Six had low top clone prevalence (predominant rearrangement relative proportion of 1.3%, 1.8%, 2.6%, 3.1%, 3.4%, 3.8%) with a median unique VJ rearrangement count of 185. Seven had higher top clone prevalence (predominant rearrangement relative proportion of 7.6%, 8.4%, 8.5%, 8.8%, 11.9%, 12.1%, 16.9%) with a considerably lower median unique VJ rearrangement count of 44. These 13 samples had variable diversity but no predominant rearrangement was more than twofold enriched relative to the next most common rearrangement. Conversely, three samples that were scored as polyclonal by BIOMED2 at the beta locus were scored as clonal based on relative prevalence (predominant rearrangement relative proportion of 25.9%, 18.6%, 6.5%) with a median unique VJ rearrangement count of 191. These discrepancies could be resolved with deeper sequencing of these libraries to determine whether insufficient depth was distorting the interpretation or whether these represent incorrect interpretations by the BIOMED2 protocol. Improvements in the BIOMED2 primer sets have led to reduced false positives compared to previous generations, and can be further diminished through the use of higher resolution gel separation and additional analysesA2, however if available, sequencing-based methods provide a more quantitative assessment and relative comparison between all rearrangements. To determine whether there was unexpected enrichment in the A037 or lymphoma data sets we compared their gene usages (data not shown). A037 and the lymphoma collection had similar VJ usage profiles with few individual unique VJ rearrangement proportion enriched in A037 of up to 1% and more enrichments amongst the lymphoma set of up to 3% as expected given the clonal enrichment of select rearrangements in T-cell lymphomas.
In summary, CapTR-Seq allows for rapid, inexpensive and high-throughput profiling of all four loci from multiple samples of diverse types from a given DNA sequencing library with fragment size of 250 bp and sequencing length of 150 bp. This method will permit intensive monitoring of TR repertoires of patients with T-cell malignancies as well as monitoring of tumor-infiltrating lymphocytes in tumors from patients undergoing immune checkpoint blockade, adoptive cell transfer and other immunotherapies.
Adoptive Cell Transfer (ACT) of in-vitro expanded Tumour-Infiltrating Lymphocytes (TIL) has emerged as an effective treatment for numerous types of solid tumours, often resulting in a durable response and in some cases a complete remission by the patientB1. This intervention effectively replaces nearly the entire heterogenous T-cell repertoire of the patient with tumour antigen and patient-specific effector T cells. Effector T-cells are integral for the adaptive immune response due to their roles in cellular cytotoxicity and cytokine production, with specificity conferred by the TCR-MHC interactionB2. The CD8+ effector T-cell repertoire consists of alpha/beta and gamma/delta subtypes, both polyclonal and skewing in the incidence of an antigen-specific response or malignancyB3. In high mutation load neoplasms, the MHC molecule often presents tumour-associated neo-antigens generated as a result of mutation that lead to clonal expansion and infiltration of tumour-infiltrating lymphocytes (TILs)B4. These TILs are largely clonal and distinct from the circulating repertoire in multiple types of neoplasiaB5. While these TILs are capable of driving an effective anti-tumour response in vitro, they are often exhausted within the tumour microenvironment as a result of expression of immunosuppressive cell-surface proteins by the tumour but their activities can be restored with immune checkpoint blockade therapyB6. The combined effect of immunotherapy intervention: immunodepletion, TIL ACT and checkpoint blockade together present an effective treatment for many patients but have a disruptive effect on the endogenous immune repertoire and therefore proper patient care would benefit from longitudinal monitoring of the T-cell repertoire during the course of disease and treatment.
During ACT immunotherapy, both the requisite immunodepletion and T-cell transfer radically disrupt the abundance and diversity of the endogenous T-cell population and therefore molecular profiling methods are required for monitoring of the patient during the course of immunotherapyB7. The TCR repertoire consists of cell-specific heterodimeric receptors uniquely rearranged and expressed from either the alpha/beta or gamma/delta genomic lociB8. The TCR has unique specificity for an antigen presented in the context of the an MHC molecule as defined by the combined interactions of the amino acid residues encoded at the V-(D)-J junction known as the complementarity determining region 3 (CDR3), and by the CDR1 and CDR2 regions in the upstream V gene fragment.
Methods and Materials
Probe design—All annotated V (V-panel), D, J (J panel) gene segments and V 3′-UTR (depletion panel) sequences were retrieved from the IMGT/LIGM-DB website (www.imgt.org). The 100 bp of annotated 3′ V gene coding regions, up to 100 bp, when available, of annotated 5′ J gene coding regions, and 120 bp of V 3′-UTR sequences were selected as baits. Probes with duplicate sequences were not included. The V-panel consists of 299 probes (IDT) targeting the 3′ and 5′ 100 bp of all TR V gene regions, and the J-panel consists of 95 probes targeting the 5′ 100 bp of all TR J gene regions as annotated by IMGT (four loci, 1.8 Mb, total targeted 36 kb). The depletion-panel consists of 131 probes targeting the 5′ 120 bp of 3′-UTR Immunoglobulin V regions, and 107 probes targeting the 5′ 120 bp of 3′-UTR TCR V regions.
DNA Isolation—CD3+ T cells were isolated by flow assisted cell sorting of PBMC populations separated from whole blood. Peripheral blood mononuclear cells (PBMC) were isolated from whole blood by centrifugation followed by DNA isolation with a GENTRA PUREGENE kit (Qiagen) according to manufacturer protocol. In the case of fresh/frozen tissues, a QIAGEN ALLPREP kit (Qiagen) was employed to extract DNA and RNA, according to the manufacturer's instructions. The whole blood plasma fraction was then treated with red blood cell lysis buffer and circulating DNA (cfDNA) was extracted using the QIAGEN NUCLEIC ACID kit (Qiagen) according to manufacturer protocol. cDNA synthesis—mRNA was separated from isolated total RNA using the NEBNEXT Poly(A) mRNA Magnetic Isolation Module (NEB) according to manufacturer's instructions. To generate cDNA, first NEBNEXT RNA First Strand Synthesis Module (NEB) was used followed by NEBNEXT RNA Second Strand Synthesis Module (NEB) according to manufacturer's instructions.
Library preparation—Isolated genomic DNA or synthesized cDNA was diluted in TE buffer to 130 uL volumes. Shearing to ˜275 bp was then performed on either a Covaris M220 Focused-ultrasonicator or E220 Focused-ultrasonicator, depending on sample throughput, with the following settings: for a sample volume of 130 μL and desired peak length of 200 bp, Peak Incident Power was set to 175 W; duty factor was set to 10%; cycles per burst was set to 200; treatment time was set to 180 s. In addition, temperature and water levels were carefully held to manufacturer's recommendations given the instrument in use.
Illumina DNA libraries were generated from 100-1000 ng of fragmented DNA using the KAPA HYPERPREP Kit (Sigma) library preparation kit following manufacturer's protocol version 5.16 employing NEXTFLEX sequencing library adapters (BIOO Scientific). Library fragment size distribution was determined using the Agilent TAPESTATION D1000 kit and quantified by fluorometry using the Invitrogen QUBIT.
Hybrid capture—For cDNA derived libraries, hybridization was performed with a pooled panel of probes targeting V and J loci in equimolar concentrations. For genomic DNA derived libraries, hybridization and capture was performed iteratively with probes specifically targeting V loci, 3′-UTR sequences, or J loci under standard SEQCAP (Roche) conditions with XGEN blocking oligos (IDT) and human Cot-1 blocking DNA (Invitrogen). Hybridization is performed at 50 C overnight. The Capture process consisting of bead incubations and washes are performed at 50 C.
For the iterative hybridization and capture process, the first J hybridization and capture is performed in completion with terminal PCR amplification with 4 steps. Following clean-up by Agencourt AMPURE XP SPRI bead purification (Beckman) this product is used as input for a subsequent depletion step. For depletion, a modified and truncated SEQCAP protocol is employed wherein following incubation of the hybridization mixture with M-270 streptavidin linked magnetic beads (Invitrogen), the 15 uL hybridization reaction is separated on a magnetic rack, the supernatant is recovered and diluted to 100 uL with TE buffer, followed by clean up by standard Agencourt AMPURE XP SPRI bead purification (Beckman). The depletion-probe-target-beads are discarded. The purified supernatant is then used as input for a subsequent V-panel capture and hybridization as described above, but with terminal PCR amplification with 16 or amplifications steps to achieve sufficient library for sequencing.
Capture Analysis—A custom Bash/Python/R pipeline was employed for analysis of paired read sequencing data generated by Illumina NEXTSEQ 2500 instrument from the hybrid-capture products. First, 150 bp paired reads were merged using PEAR 0.9.6 with a 25 bp overlap parameter. This results in a single 275 bp sequence for each sequenced fragment. Next, specific V, J, and D genes within the fragment sequence were identified by aligning regions against a reference sequence database. Specifically, individual BLAST databases were created using all annotated V, D, J gene segments retrieved from the IMGT/LIGM-DB website (www.imgt.org), as these full-length gene sequences were the source of probes used to design the hybrid-capture probe panel. Individual merged reads are iteratively aligned using BLASTn with an e value cut-off of 1 to the V database, J database then D database with word size of 5 for D segment queries. Trimming of identified V or J segments in the query sequence is performed prior to subsequent alignment. From reads containing V and J sequences, we identified V/J junction position and the antigen specificity determining Complementarity Determining Region 3 (CDR3) sequences. In order to identify CDR3 sequences, the V/J junction position is extracted from the previous search data for those fragments containing both a V and J search result. 80 bp of DNA sequence flanking this junction is translated to amino acid sequence in all six open reading frames and sequences lacking stop codons are searched for invariable anchor residues using regular expressions specific for each TR class as determined by sequence alignments of polyclonal hybrid-captured data from rearranged TR polypeptides annotated by IMGT.
Results and Discussion
Methods Improvement
We experimented with alternate capture methods, using an iterative three-step hybridization and capture, first with a J panel then molecular depletion of unrearranged V-gene sequences, then subsequently with a V panel (data not shown). The depletion probes (V-gene and J-gene) are shown in Table D. These altered protocols improved recovery of unique CDR3 sequences when normalized to reads. When compared to a one-step V-panel capture, the one-step combined VJ-panel capture increased signal by 6.84×, the two-step J and V iterative capture increased signal by 12×(no significant difference was observed for J-V or V-J iterative order), and the three-step J-depletion-V iterative capture increased signal by 31.2×(
We experimented with reducing hybridization and wash temperatures to improve recovery (
We determined the best method for depletion (
We tested whether depletion should be preceded by a V or J capture (
Input Source Material Comparisons
To determine whether we could characterize the TCR repertoire from both low and high signal samples, we performed a series of dilution curves for CD3+ genomic DNA (
Since each sequenced sample represents only a snapshot of the TCR repertoire with the extent dependent on the amount of input material and the complexity of the source repertoire, we were interested in whether the method could assay complete VJ or CDR3 saturation of a patient. We looked at unique VJ pair recovery across multiple samples derived from a single patient blood draw (data not shown). Beta locus VJ saturation was achieved with fewer than ten runs. With sufficient input and sequencing depth, VJ saturation could be achieved in a single run. We also looked at CDR3 saturation across these same samples and were able to achieve approximately 50% beta locus saturation (data not shown). This level could be achieved with fewer samples by using cDNA libraries as input with deeper sequencing.
We looked at whether the genomic DNA and cDNA samples were recapitulating the same VJ combinations at the beta locus (
We looked at whether the genomic DNA and cDNA samples were recapitulating the same CDR3 sequences (
Investigation of Samples from Adoptive Cell Transfer Immunotherapy
We next applied the CapTCR-Seq methodology to samples derived from expanded Tumor Infiltrating Lymphocyte (TIL) infusion populations and PBMCs from serial blood draws from patients undergoing adoptive cell transfer immunotherapy. We wanted to track clones from the TIL culture over time to determine whether they successfully colonized the patient and the extent of their population over time (
Comparison Between Uncaptured and Captured Tumor Samples
We wished to demonstrate the value of this method for interrogating existing cDNA RNA-Seq libraries (data not shown). To do this, Illumina cDNA sequencing libraries were generated from FFPE-derived total RNA and subjected to sequencing followed by analysis using the TCR annotation pipeline to identify unique TCR CDR3 sequences (bulk unique CDR3). Residual library then underwent CapTCR-Seq to identify unique TCR CDR3 sequences (capture unique CDR3). The CapTCR-Seq method yielded a greatly increased number of unique CDR3 sequences (mean: 466 fold, median: 353 fold). When normalized to number of total reads sequenced, we observed a 15fold increase in signal per read sequenced (mean: 15.2, median: 14.5, n=41).
Investigation of Tumor Repertoires from Different Cancer Types
We next wanted to characterize tumor repertoires and investigate highly prevalent TIL clones in the blood repertoire before and during anti-PDL1 immunotherapy treatment. We selected five patients, each with a different tumor type: Patient A: Head and neck; Patient B: Breast; Patient C: Ovarian; Patient D: Melanoma; Patient E: Cervical. Each patient had three sample types: Tumor tissue (extracted DNA and RNA), pre-treatment blood (extracted PBMC DNA, PBMC RNA, and plasma cfDNA), on-treatment blood (extracted plasma cfDNA).
We first queried the extent of the TCR signal in the tumor samples in terms of infiltration and clonality. TCR signal is defined as the total number of counts of fragments containing both a V and J gene region (non-unique, reads normalized) while diversity is defined as the total number of unique CDR3 sequences detected (unique, reads normalized). Overall, diversity increased with signal (
Next we assessed the clonality of the tumor sample TIL repertoire. Tumors with clonal infiltration have a larger than expected population of one or more VJ rearrangements, the population of which are significantly greater than the next most prevalent clone. Patient A appears to have a large alpha rearrangement population in its tumor compared to baseline blood, while the most prevalent beta rearrangement is only slightly enriched (
Next we assessed how the most prevalent tumor VJ rearrangements differed in terms of prevalence across the other patient samples (
We performed similar experiments relating to B-cells. Our design targets more than 500 V-regions and 50 J-regions within the IGH, IGK and IGL loci annotated in the IMmunoGeneTics database. This accounts for all known Ig alleles while maximizing depth of coverage in selected regions. A blast-based informatics pipeline calls V(D)J recombinations and an algorithm combining information from large-insert and soft-clipped reads are used to predict candidate rearrangements which are manually verified in Integrated Genome Viewer.
Candidate V(D)J rearrangements and translocations detected through this approach have been validated in three well-characterized cell-lines with publically available whole genome data; an additional 67 MM cell lines have been annotated for V(D)J rearrangements and translocations into IGH, IGL and IGK genes. The limit of detection was established with a cell-line dilution series. We were also able to translate these techniques to cell-free DNA. These methods are applicable to the detection of MRD in mature B-cell malignancies and immunoglobulin repertoire profiling in a many clinical scenarios including cellular immunotherapy and therapeutics with immunomodulatory effects. V(D)J and complex rearrangement annotations in 70 MM cell-lines are highly relevant in further in-vitro studies.
The B-cell V-gene and J-gene capture probes used are shown in Tables B1 and B2 respectively.
Although preferred embodiments of the invention have been described herein, it will be understood by those skilled in the art that variations may be made thereto without departing from the spirit of the invention or the scope of the appended claims. All documents disclosed herein, including those in the following reference list, are incorporated by reference.
‡Assumptions of note: Stock Jurkat Cell Line DNA Concentration: 10.6 ng/μL; Presumes lymphocyte DNA content: 0.0007 ng/cell
This application is a national phase application under 35 U.S.C. § 371 of International Application No. PCT/CA2018/000104 filed 29 May 2018, which claims priority to U.S. Provisional Application No. 62/512,255 filed 30 May 2017. The entire contents of each of the above-referenced disclosures is specifically incorporated by reference herein without disclaimer.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/CA2018/000104 | 5/29/2018 | WO |
| Publishing Document | Publishing Date | Country | Kind |
|---|---|---|---|
| WO2018/218332 | 12/6/2018 | WO | A |
| Number | Name | Date | Kind |
|---|---|---|---|
| 2392551 | Roe | Jan 1946 | A |
| 6189282 | VanderWerf | Feb 2001 | B1 |
| 9453341 | Swierad et al. | Sep 2016 | B1 |
| 20070292216 | Hamel | Dec 2007 | A1 |
| 20080125333 | Labgold et al. | May 2008 | A1 |
| 20100018146 | Aube et al. | Jan 2010 | A1 |
| 20100029498 | Gnirke | Feb 2010 | A1 |
| 20110072753 | MacDonald | Mar 2011 | A1 |
| 20140331343 | Bradley et al. | Nov 2014 | A1 |
| 20150218620 | Behlke et al. | Aug 2015 | A1 |
| Number | Date | Country |
|---|---|---|
| WO 2016090273 | Jun 2016 | WO |
| WO 2017013436 | Jan 2017 | WO |
| WO 2017177308 | Oct 2017 | WO |
| Entry |
|---|
| He et al., “IgH gene rearrangements as plasma biomarkers in Non-Hodgkin's Lymphoma patients,” Oncotarget 2011, 2:178-185. (Year: 2011). |
| Extended European Search Report issued in Corresponding European Application No. 18810749.4, dated Apr. 19, 2021. |
| Yaari et al., “Practical guidelines for B-cell receptor repertoire sequencing analysis” Genome Medicine 2015, 7(1), 14 pages. |
| Bleakley et al., “Recovering probabilities for nucleotide trimming processes for T cell receptor TRA and TRG V-J junctions analyzed with IMGT tools”, BMC Bioinformatics, 9:408, 2008. |
| Jiang et al., “VDJ-Seq: Deep Sequencing Analysis of Rearranged Immunoglobulin Heavy Chain Gene to Reveal Clonal Evolution Patterns of B Cell Lymphoma”, J Vis Exp., 106:053215, 2015. |
| Mamedov et al., “Preparing unbiased T-cell receptor and antibody cDNA libraries for the deep next generation sequencing profiling”, Front Immunol., 4:456, 2013. |
| Matsuda et al., “The complete nucleotide sequence of the human immunoglobulin heavy chain variable region locus”, J Exp Med., 188(11):2151-2162, 1998. |
| Monod et al., “IMGT/JunctionAnalysis: the first tool for the analysis of the immunoglobulin and T cell receptor complex V-J and V-D-J Junctions”, Bioinformatics, 20 Suppl 1 :i379-385, 2004. |
| Partial Supplementary European Search Report Issued in Corresponding EP Patent Application No. 18810749.4, dated Jan. 12, 2021. |
| Woodsworth et al., “Sequence analysis of T-cell repertoires in health and disease”, Genome Med., 5(10):98, 2013. |
| Gao & Wang, “Ligation-Anchored PCR Unveils Immune Repertoire of TCR-Beta from Whole Blood,” BMC Biotechnology, 15-39, 2015. |
| Yaari & Kleinstein, “Practical Guidelines for B-Cell Receptor Repertoire Sequencing Analysis,” Genome Medicine, 7: 121, 2015. |
| He, J. et al., “IgG gene rearrangements as plasma biomarkers in non-Hodgkin's lymphoma patients.” Oncotarget 2011, 2(3), 178-185. |
| International Search Report and Written Opinion issued in corresponding application No. PCT/CA2018/000104, dated Sep. 4, 2018. |
| Bashford-Roger, et al., “Capturing Needles in Haystacks: A Comparison of B-Cell Receptor Sequencing Methods,” BMC Immunology, 15: 29, 2014. |
| Bolotin, et al., “Next Generation Sequencing for TCR Repertoire Profiling Platform-Specific Features and Correction Algorithms,” European Journal Immunology, 42: 3073-3083, 2012. |
| Brochet et al., “IMGTN-QUEST: the Highly customized and integrated system for IG and TR standardized V-J and V-D-J sequence analysis” Nucleic Acids Res., 2008, 36:W503-508. |
| Camacho et al., “BLAST+: architecture and applications” BMC Bioinformatics, 2009, 10:421. |
| Extended European Search Report issued in corresponding Application No. 17781661.8, dated Nov. 18, 2019. |
| Fisher, et al., “Neuroblastoma Killing Properties of V-Delta 2 and V-Delta2 Negative Gamma Delta T Cells Following Expansion by Artificial Antigen Presenting Cells,” Clinical Cancer Research, 20(22): 5720-5732, 2014. |
| Gazzola et al., “The evolution of clonality testing in the diagnosis and monitoring of hematological malignancies” Therapeutic Advances in Hematology., 2014, 5(2):35-47. |
| Halper-Stromberg, et al., “Fine Mapping of V(D)J Recombinase Mediated Rearrangements in Human Lymphoid Malignancies,” BMC Genomics, 14: 565, 2014. |
| Herman et al., “Filter-based hybridization capture of subgenomes enables resequencing and copy-number detection” Nature Methods 2009, 6(7), 7 pages. |
| Jiang, et al., “Deep Sequencing Reveals Clonal Evolution Patterns and Mutation Events Associated with Relapse in B-Cell Lymphomas,” Genome Biology, 15: 432, 2014. |
| Li, HD “Fast and accurate short read alignment with Burrows-Wheeler Transform” Bioinformatics, 2009, 25:1754-1760. |
| Lin, et al., “Highly Sensitive and Unbiased Approach for Elucidating Antibody Repertoires,” PNAS, 113(28):7826-7851, 2016. |
| Linneman, et al., “High-Throughput Identification of Antigen-Specific TCRs by TCR Gene Capture,” Nature Medicine, 19(11): 1534-1541, and Supplementary Information, 2013. |
| Liu et al. “Systematic Comparative Evaluation of Methods for Investigating the TCRβ Repertoire” PLOS One 2016, 18 pages. |
| Logan, et al., “High-Throughput VDJ Sequencing for Quantification of Minimal Residual Disease in Chronic Lymphocytic Leukemia and Immune Reconstitution Assessment,” PNAS, 108(52): 21194-21199, 2011. |
| Mahe, “Design & Validation of a Hybrid-Capture Next-Generation Sequencing-Based T-Cell Clonality Assay,” Master of Science Thesis, University of Toronto, Graduate Department of Laboratory Medicine and Pathobiology, 2016. |
| Mamanova et al. “Target-enrichment strategies for next-generation sequencing” Nat. Methods, 2010, 7(2):111-118. |
| Mamedov, et al., “Quantitative Tracking of T Cell Clones After Haematopoietic Stem Cell Transplantation,” EMBO Mol. Med, 201-207, 2011. |
| Mulder, et al., “CapTCR-Seq: Hybrid Capture for T-Cell Receptor Repetoire Profiling,” Blood Advances, 2(23): 3506-3514, 2018. |
| Office Action Issued in U.S. Appl. No. 16/093,825, dated Aug. 18, 2020. |
| Parkinson, et al., “Violation of the 12/23 Rule of Genomic V(D)J Recombination is Common in Lymphocytes,” Genome Research, 25: 226-234, 2015. |
| Rosati, et al., “Overview of Methodologies for T-Cell Receptor Repertoire Analysis,” BMC Biotechnology, 17: 61, 2017. |
| Ruggiero, et al., “High-resolution analysis of the human T-cell receptor repertoire” Nature Communications 2015, 1-7. |
| Smith et al., “Identification of common molecular subsequences” J. Mot. Biol., 1981, 147(1):195-197. |
| Number | Date | Country | |
|---|---|---|---|
| 20200115749 A1 | Apr 2020 | US |
| Number | Date | Country | |
|---|---|---|---|
| 62512255 | May 2017 | US |
