The invention relates to the detection of immunological biomarkers, particularly autoantibodies, to predict immunogenicity and therapeutic responses to adalimumab in patients with Rheumatoid Arthritis (RA).
Rheumatoid arthritis (RA), a chronic inflammatory articular disease, is characterized by persistent synovitis, cartilage degradation, and bone erosions [1], and tumor necrosis factor (TNF)-α is a crucial inflammatory mediator in RA-related synovitis and joint damage [2]. The importance of the role of TNF-α in RA pathogenesis is supported by the effectiveness of biologics targeting this cytokine [2-4], although the efficacy diminishes in some patients over time (secondary failure) [5]. Accumulating evidence indicates that the presence of anti-drug antibodies (ADAb) in certain patients may be associated with low or undetectable drug levels and ensuing reduction of therapeutic responsiveness to TNF-α inhibitors [6-10]. Such ADAb responses reflect the differential immunogenicity of the given biologic drug triggered in individual patients, which results in some patients developing a neutralising antibody response against the biologic drug and others not. In the face of such uncertainty about whether individual RA patients will show therapeutic responsiveness to TNF-α inhibitors or not [11], physicians hoping to optimize personalized and precision therapy are thus eager to find biomarkers which can predict the emergence of ADAb and the effectiveness of anti-TNF-α biologics.
Proteomics research has been increasingly applied to the identification of novel biomarkers that might be useful for monitoring therapeutic response in RA patients on specific treatments [12-14]. However, there is currently limited knowledge about circulating biomarkers that are able to predict the development of ADAb in RA patients receiving anti-TNF-α therapy.
Autoantibody biomarkers as described herein are autoantibodies to antigens, autoantibodies being antibodies which are produced by an individual which are directed against one or more of the individual's own proteins (‘self’ antigens).
The aim of the present invention therefore is to provide a novel panel of autoantibody biomarkers that are able to predict immunogenicity of adalimumab and therapeutic responses to adalimumab in individual RA patients, prior to treatment with adalimumab, a widely used TNF-α inhibitor marketed under the brand name HUMIRA® and commonly used for the treatment of autoimmune diseases, such as RA, Crohn's Disease and Psoriasis.
In one aspect of the invention, there is provided a method for predicting a response to adalimumab from a sample extracted from a rheumatoid arthritis patient prior to treating the patient with adalimumab, said response being classified as a good response corresponding to anti-drug antibody negative or a poor response corresponding to anti-drug antibody positive, comprising the steps of:
Advantageously the autoantibody biomarkers can be used to predict immunogenicity of adalimumab and therapeutic responses to adalimumab in individual rheumatoid arthritis (RA) patients at baseline (i.e. prior to treating the patient with adalimumab).
In one embodiment the sample is tested using a panel of antigens that correspond to the autoantibody biomarkers. Typically the antigens are biotinylated proteins. Advantageously the biotinylation ensures that the antigens are folded in their correct form to ensure accuracy of detection by the autoantibody biomarkers.
In one embodiment the antigens may include one or more additional antigens from the group comprising of PPARD, SPANXN2, HNRNPA2B, TRIB2, CEP55, SH3GL1, FN3K, PANK3, HPCAL1, THRA, AIFM1, ODC1, RPS6KA4, EEF1D, KLF10, EPHA2, PRKAR1A and EAPP.
It should be noted that not all human antigens generate an autoantibody response and it is not possible to predict a priori which human antigens will do so in a given patient cohort—of the 1622 antigens tested, only autoantibodies against the 21 antigens described above are suitable as biomarkers in predicting immunogenicity of adalimumab and therapeutic responses to adalimumab in RA patient at baseline.
In one embodiment each biotinylated protein is formed from a Biotin Carboxyl Carrier Protein (BCCP) folding marker which is fused in-frame with the protein.
In one embodiment the biotinylated proteins are bound to a streptavidin-coated substrate.
Advantageously full-length proteins are expressed as fusions to the BCCP folding marker which itself becomes biotinylated in vivo when the fusion partner is correctly folded. By comparison misfolded fusion partners cause the BCCP to remain in the ‘apo’ (i.e. non-biotinylated) form such that it cannot attach to a streptavidin substrate. Thus only correctly folded fusion proteins become attached to the streptavidin substrate via the biotin moiety appended to the BCCP tag.
In one embodiment the substrate comprises a glass slide, biochip, strip, slide, bead, microtitre plate well, surface plasmon resonance support, microfluidic device, thin film polymer base layer, hydrogel-forming polymer base layer, or any other device or technology suitable for detection of antibody-antigen binding.
In one embodiment the substrate is exposed to a sample extracted from a person, such that autoantibody biomarkers from the sample may bind to the antigens. Typically the sample comprises any or any combination of exosomes, blood, serum, plasma, urine, saliva, amniotic fluid, cerebrospinal fluid, breast milk, semen or bile.
Typically the sample is collected at baseline prior to administration of the first dose of adalimumab.
In one embodiment following exposure to the sample, the substrate is exposed to a fluorescently-tagged secondary antibody to allow the amount of any autoantibodies from the sample bound to the antigens on the panel to be determined. Typically the secondary antibody is anti-human IgG, but it will be appreciated that other secondary antibodies could be used, such as anti-IgM, anti-IgG1, anti-IgG2, anti-IgG3, anti-IgG4 or anti-IgA.
In one embodiment the patient's response to treatment with adalimumab (i.e. the immunogenic and/or therapeutic response outcome to adalimumab in RA patient at baseline) corresponds to the relative or absolute amount of autoantibodies from the baseline sample specifically binding to the antigens.
In one embodiment the method is performed in vitro.
In a further aspect of the invention, there is provided a method for manufacturing a kit for predicting a response to adalimumab from a sample extracted from a rheumatoid arthritis patient prior to treating the patient with adalimumab, comprising the steps of:
In one embodiment the antigens further comprise one or more of PPARD, SPANXN2, HNRNPA2B, TRIB2, CEP55, SH3GL1, FN3K, PANK3, HPCAL1, THRA, AIFM1, ODC1, RPS6KA4, EEF1D, KLF10, EPHA2, PRKARIA and EAPP.
In a further aspect of the invention there is provided a method for predicting immunogenicity of adalimumab and therapeutic responses to adalimumab in RA patients at baseline by exposing a composition comprising a panel of antigens as herein described to a sample extracted from a person, and determining the level of autoantibodies from the sample binding to the antigens.
In a yet further aspect of the invention there is provided a method for predicting immunogenicity of adalimumab and therapeutic responses to adalimumab in RA patients at baseline by exposing a composition comprising a panel of antigens as herein described to a sample extracted from a person in vitro, and determining the level of autoantibodies from the sample binding to the antigens.
In further aspect of the invention, there is provided a composition comprising a panel of antigens for predicting an immunogenic and/or therapeutic response to adalimumab in a rheumatoid arthritis patient who has not previously been treated with adalimumab, characterised in that the antigens comprise SSB, TROVE2 and ZHX2.
In one embodiment the antigens further comprise one or more of PPARD, SPANXN2, HNRNPA2B, TRIB2, CEP55, SH3GL1, FN3K, PANK3, HPCAL1, THRA, AIFM1, ODC1, RPS6KA4, EEF1D, KLF10, EPHA2, PRKARIA and EAPP.
In one embodiment the antigens are biotinylated proteins
In one embodiment the amount of one or more autoantibody biomarkers binding in vitro to the antigens in a sample from a patient can be measured to determine the immunogenicity and therapeutic response outcome to adalimumab in an RA patient at baseline.
In yet further aspect of the invention, there is provided a composition comprising a panel of autoantibody biomarkers for predicting an immunogenic and/or therapeutic response to adalimumab in a rheumatoid arthritis patient who has not previously been treated with adalimumab, wherein the level of the autoantibody biomarkers are measured in a sample collected from the patient;
It will be convenient to further describe the present invention with respect to the accompanying drawings that illustrate possible arrangements of the invention. Other arrangements of the invention are possible, and consequently the particularity of the accompanying drawings is not to be understood as superseding the generality of the preceding description of the invention.
The invention utilises the Biotin Carboxyl Carrier Protein (BCCP) folding marker which is cloned in-frame with the gene encoding the protein of interest, as described above and in EP1470229. The structure of the E. coli BCCP domain is illustrated in
BCCP acts not only as a protein folding marker but also as a protein solubility enhancer. BCCP can be fused to either the N- or C-terminal of a protein of interest. Full-length proteins are expressed as fusions to the BCCP folding marker which becomes biotinylated in vivo, but only when the protein is correctly folded. Conversely, misfolded proteins drive the misfolding of BCCP such that it is unable to become biotinylated by host biotin ligases. Hence, misfolded proteins are unable to specifically attach to a streptavidin-coated solid support. Therefore only correctly folded proteins become attached to a solid support via the BCCP tag.
The surface chemistry of the support is designed carefully and may use a three-dimensional thin film hydrogel layer (polyethylene glycol; PEG), which retains protein spot morphologies and ensures consistent spot sizes across the array. The PEG layer inhibits non-specific macromolecule absorption, therefore reducing the high background observed using other platforms. The solid support used to immobilize the selected biomarkers thus provides excellent signal-to-noise ratios and low limits of detection (translating in to improved sensitivity). In addition the PEG hydrogel layer also aids preservation of the folded structure and functionality of arrayed proteins and protein complexes post-immobilisation.
Retention of the correct folded structure of immobilised antigens during antibody binding assays (‘immuno-assays’) is particularly advantageous because human antibodies are known in general to specifically recognise and bind to discontinuous, solvent-accessible epitopes on protein surfaces, yet are also known to bind non-specifically to exposed hydrophobic surfaces on unfolded proteins. Thus serological assays carried out on arrays of unfolded proteins typically give rise to many false positive results due to such non-specific binding events (which have no biological relevance), whilst at the same time also giving rise to many false negative results due to the absence of biologically-relevant discontinuous epitopes. By contrast, serological assays carried out on arrays of folded antigens result in detection of biologically meaningful antibody-antigen interactions that are not obscured by high rates of non-specific binding.
As biotinylated proteins bound to a streptavidin-coated surface show negligible dissociation, this interaction therefore provides a superior means for tethering proteins to a planar surface in a controlled orientation and is thus ideal for applications such as protein arrays, SPR and bead-based assays. The use of a compact, folded, biotinylated, 80 residue domain BCCP affords two significant advantages over for example the AviTag and intein-based tag. First, the BCCP domain is cross-recognised by eukaryotic biotin ligases enabling it to be biotinylated efficiently in yeast, insect, and mammalian cells without the need to co-express the E. coli biotin ligase. Second, the N- and C-termini of BCCP are physically separated from the site of biotinylation by 50 Å (as shown in
The addition of BCCP permits the monitoring of fusion protein folding by measuring the extent of in vivo biotinylation. This can be measured by standard blotting procedures, using SDS-PAGE or in situ colony lysis and transfer of samples to a membrane, followed by detection of biotinylated proteins using a streptavidin conjugate such as streptavidin-horseradish peroxidase. Additionally, the fact that the BCCP domain is biotinylated in vivo is particularly useful when multiplexing protein purification for fabrication of protein arrays since the proteins can be simultaneously purified from cellular lysates and immobilised in a single step via the high affinity and specificity exhibited by a streptavidin surface.
Materials and Methods
Gene synthesis and cloning. The pPRO9 plasmid (see
Recombinant baculovirus was generated via co-transfection of Sf9 cells (a clonal isolate derived from the parental Spodoptera frugiperda cell line IPLB-Sf-21-AE) with a replication-deficient bacmid vector carrying the viral polyhedrin promoter and a transfer vector carrying a specific coding sequence for a specific antigen. Homologous recombination between the transfer vector and the bacmid within Sf9 cells resulted in formation of a replication competent baculoviral vector encoding the specific antigen fused to the BCCP tag. Successful homologous recombination between the transfer vector and the bacmid within Sf9 cells caused the transfected cells to show signs of viral cytopathic effect (CPE) within few days of culture incubation. The most common CPE observed was the significantly enlargement of average cell size, a consequences of viral progeny propagation. These baculoviruses known as P0 were then released into the culture medium, and viral amplification were done to generate a higher titre of P1 viruses.
Protein Expression. Expression of recombinant antigens was carried out in 24 well blocks using 3 ml cultures containing 6×106 Sf9 cells per well. High titre, low passage, viral stocks of recombinant baculovirus (>107 pfu/ml) were used to infect Sf9 insect cells. The infected cells were then cultured for 72 hours to allow them to produce the recombinant protein of interest. The cells were washed with PBS, resuspended in buffer, and were frozen in aliquots at −80° C. ready for lysis as required. Depending on the transfer vector construct and the nature of the antigen itself, the resultant recombinant protein lysate can be recovered either from the cultured cell or the culture medium. Expression of recombinant proteins was confirmed by SDS-PAGE as well as by Western blot using streptavidin-HRP-based detection. In total, 1622 human antigens were cloned and expressed using this methodology.
Array fabrication. Hydrogel coated, streptavidin-derivatised slides were custom manufactured by Schott and used as substrates on to which the biotinylated proteins were then printed. A total of 9 nanoliters of crude protein lysate was printed on a HS slide in quadruplicate using non-contact piezo printing technology. Print buffer that have a pH between 7.0 and 7.5 were used. The slides were dried by centrifugation (200×g for 5 min) before starting the washing and blocking. The printed arrays were blocked with solutions containing BSA or casein (concentration: 0.1 mg/ml) in a phosphate buffer. The pH was adjusted to be between 7.0 and 7.5 and cold solutions were used (4° C.-20° C.). Slides were not allowed to dry between washes, and were protected from light. In total, each resultant ‘Immunome array’ comprised 1622 antigens, each printed in quadruplicate.
1. Study Cohort
The study cohort comprised of a total of 62 plasma and serum samples collected from RA patients at baseline (i.e. prior to treatment administration);
Patients were administered with adalimumab at a dose of 40 mg every other week. The immunogenicity and therapeutic response to adalimumab were evaluated at week 24, the latter by using the EULAR response criteria [15]. EULAR responders were defined as RA patients with good and moderate (“good response”) or poor (“poor response”) EULAR therapeutic responses.
2. Sample Collection and Storage
Peripheral blood samples were collected immediately before the first adalimumab administration (the baseline sample) and also at week 24. After centrifugation at 1000 g for 10 min within 15 min of withdrawal, serum and plasma samples were stored at −70° C.
3. Sample Preparation and Dilution
For each run, samples were placed in a shaking incubator set at 20° C. to allow thawing for 30 minutes. When completely thawed, each sample was vortexed vigorously three times and debris was pelleted by centrifugation for 3 minutes at 13,000 rpm. 11.25 μL of the sample was pipetted into 4.5 mL of Serum Assay Buffer (SAB) containing 0.1% v/v Triton, 0.1% w/v BSA in PBS (20° C.) and vortexed to mix three times. The tube was tilted during aspiration to ensure that the sera was sampled from below the lipid layer at the top but did not touch the bottom of the tube in case of presence of any sediment. Batch records were marked accordingly to ensure that the correct samples were added to the correct tubes. Samples were then randomised prior to assay.
4. Biomarker Assay
Each Immunome array was removed from the storage buffer using forceps, placed in the slide box and rack containing 200 mL cold SAB and shaken on an orbital shaker at 50 rpm, for 5 minutes. After washing, each slide was scanned using a barcode scanner and then placed array side up in an individual slide hybridization chamber containing an individual diluted sera (Step 3 above). All slides were and incubated on a horizontal shaker at 50 rpm for 2 hours at 20° C.
5. Array Washing after Serum Binding
Each Immunome array slide was rinsed twice in individual “Pap jars” with 30 mL SAB, followed by 200 mL of SAB buffer in the slide staining box for 20 minutes on the shaker at 50 rpm at room temperature. All slides were transferred sequentially and in the same orientation.
6. Incubation with Cy3-Anti-Human IgQ
Binding of autoantibodies to the arrayed antigens on the arrays was detected by incubation with Cy3-rabbit anti-human IgG (Dako Cytomation) labelled according to the manufacturer's recommended protocols (GE Healthcare). Arrays were immersed in hybridization solution containing a mixture of Cy3-rabbit anti-human IgG solution diluted 1:1000 in SAB buffer for 2 hours at 50 rpm in 20° C.
7. Washing after Incubation with Cy3-Anti-Human IgG
After incubation, the slide was dipped in 200 mL of SAB buffer, 3 times for 5 minutes at 50 rpm at room temperature. Excess buffer was removed by immersing the slide in 200 mL of pure water for a few minutes. Slides were then dried for 2 min by centrifugation at 240 g at room temperature. Slides were then stored at room temperature until scanning. Fluorescent hybridization signals were measured with excitation at 550 nm and emission at 570 nm using a microarray laser scanner (Agilent) at 10 μm resolution.
Bioinformatic analysis.
1. Image Analysis: Raw Data Extraction
The aim of an image analysis is to evaluate the amount of autoantibody present in the serum sample by measuring the median intensities of all the pixels within each probed spot. A raw .tiff format image file is generated for each slide, i.e. each sample. Automatic extraction and quantification of each spot on the array are performed using the GenePix Pro 7 software (Molecular Devices) which outputs the statistics for each probed spot on the array. This includes the mean and median of the pixel intensities within a spot as well as in its surrounding local background area. A GAL (GenePix Array List) file for the array is generated to enable image analysis. This file contains the information of all probed spots and their positions on the array. Following data extraction, a GenePix Results (.GPR) file is generated for each slide which contains the information for each spot: Protein ID, protein name, foreground intensities, background intensities etc. In the data sheet generated from the experiment, both foreground and background intensities of each spot are represented in relative fluorescence units (RFUs).
2. Data Handling and Pre-Processing
For each slide, antigens and control probes are spotted in quadruplicate on each array. The following steps were performed to verify the quality of the antigen array data before proceeding with data analysis:
Step 1:
Calculate net intensities for each spot by subtracting background signal intensities from the foreground signal intensities of each spot. For each spot, the background signal intensity was calculated using a circular region with three times the diameter of the spot, centered on the spot.
Step 2:
Remove replica spots with net intensity ≤0.
Step 3:
Zero net intensities if only 1 replica spot remaining.
Step 4:
Calculate the coefficient of variant (CV %) for the replica spots on each array.
Flag any replica spots with only 2 or less replica/s remaining and CV %>20% as “High CV”. The mean net intensity of such replica spots (i.e. antigens) is excluded from downstream analysis.
For antigens/controls with a CV %>20% and with 3 or more replica spots remaining, the replica spots which result in this high CV % value were filtered out. This was done by calculating the standard deviation between the median value of the net intensities and individual net intensities for each set of replica spots. The spot with the highest standard deviation was removed. CV % values were re-calculated and the process repeated.
Step 5:
Calculate the mean of the net intensities for the remaining replica spots.
Step 6:
Inspect signal intensities of two positive controls: IgG and Cy3-BSA.
Step 7:
Carry out a composite normalisation [16] using both quantile-based and total intensity-based modules for each dataset. This method assumes that different samples share a common underlying distribution of their control probes while taking into account the potential existence of flagged spots within them. The Immunome array uses Cy3-labelled biotinylated BSA (Cy3-BSA) replicates as the positive control spots across slides. Hence it is considered as a ‘housekeeping’ probe for normalisation of signal intensities for any given study.
The quantile module adopts the algorithm described by Bolstad et al., 2003 [17]. This reorganisation enables the detection and handling of outliers or flagged spots in any of the Cy3BSA control probes. A total intensity-based module was then implemented to obtain a scaling factor for each sample. This method assumes that post-normalisation, the positive controls should have a common total intensity value across all samples. This composite method aims to normalise the protein array data from variations in their measurements whilst preserving the targeted biological activity across samples. The steps are as follows:
Quantile-Based Normalisation of all cy3BSA across all samples
(i=spot number and j=sample number)
Intensity-Based Normalisation
3. Data Analysis
Batch normalisation: The composite normalised data sets from the assays in the two runs were merged using a ComBat normalisation method [18]. For each protein, this method inputs the net intensity values across all the samples from the 2 data sets and adjusts for any possible batch effects between the two data sets using a parametric empirical Bayes frameworks.
Biomarker Panel Selection: A pipeline was developed which utilises a combination of feature selection and machine learning methodologies to determine the optimal combination of antigens eliciting autoantibody responses from the list of 1622 antigens which are able to provide the best stratification between ADAb-positive and ADAb-negative patients [19]. For feature selection, univariate statistical tests, random forest importance and mutual information metrics were the filter methods used to rank biomarkers.
Biomarker panels were generated by additively selecting the top-ranking biomarkers as inputs to machine learning models up to a total of top 160 biomarkers (top 10% of biomarkers). Any further addition of number of biomarkers did not lead to significant improvements of model performance and would lead to further increase of computational time. To estimate the biomarker panel performance, ROC, sensitivity and specificity was evaluated and the biomarker panel with the best sensitivity and specificity was deemed as the optimal panel to stratify ADAb status. For this analysis, machine learning models were built using Random Forests [20], under default settings with leave-one out cross validation (LOOCV). All analyses were performed using packages available in R. Feature selection was performed using ranger [21] package and all machine learning models were performed using the caret [22] package.
Table 1 shows top 4 best performing biomarker panel from the machine learning models using leave-one out cross validation. The lowest number of antigens with the highest sensitivity and specificity was deemed to be the top biomarker panel. This panel comprises SSB, TROVE2, ZHX2, PPARD, SPANXN2, HNRNPA2B, TRIB2, CEP55, SH3GL1, FN3K, PANK3, HPCAL1, THRA, AIFM1, ODC1, RPS6KA4, EEF1D, KLF10, EPHA2, PRKARIA and EAPP.
The biomarkers were ranked based on Random Forests estimated variable importance measure [23] derived from each panel (
Comparison with the preliminary American College of Rheumatology and the World Health Organization/International League Against Rheumatism Criteria. Arthritis Rheum 1996; 39:34-40.
Stat. Data Anal. 143, 106839.
Homo sapiens thyroid hormone receptor alpha
Homo sapiens eukaryotic translation
Number | Date | Country | Kind |
---|---|---|---|
10202011473Y | Nov 2020 | SG | national |
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/SG2021/050690 | 11/11/2021 | WO |