Patent Application
20150299804
The National Cancer Institute has estimated that in the United States alone, 1 in 3 people will be struck with cancer during their lifetime. Moreover, approximately 50% to 60% of people contracting cancer will eventually succumb to the disease. The widespread occurrence of this disease underscores the need for improved anticancer regimens for the treatment of malignancy.
Due to the wide variety of cancers presently observed, numerous anticancer agents have been developed to destroy cancer within the body. These compounds are administered to cancer patients with the objective of destroying or otherwise inhibiting the growth of malignant cells while leaving normal, healthy cells undisturbed. Anticancer agents have been classified based upon their mechanism of action, and are often referred to as chemotherapeutics or immunotherapeutics (agents whose therapeutic effects are mediated by their immuno-modulating properties). The vertebrate immune system requires multiple signals to achieve optimal immune activation; see, e.g., Janeway, Cold Spring Harbor Symp. Quant. Biol., 54:1-14 (1989); and Paul, W. E., ed., Fundamental Immunology, 4th Edition, Raven Press, NY (1998), particularly Chapters 12 and 13, pp. 411-478. Interactions between T lymphocytes (T cells) and antigen presenting cells (APCs) are essential to the immune response. Levels of many cohesive molecules found on T cells and APC's increase during an immune response (Springer et al., Ann. Rev. Immunol., 5:223-252 (1987); Shaw et al., Curr. Opin. Immunol., 1:92-97 (1988); and Hemler, Immunology Today, 9:109-113 (1988)). Increased levels of these molecules may help explain why activated APCs are more effective at stimulating antigen-specific T cell proliferation than are resting APCs (Kaiuchi et al., J. Immunol., 131:109-114 (1983); Kreiger et al., J. Immunol., 135:2937-2945 (1985); McKenzie, J. Immunol., 141:2907-2911 (1988); and Hawrylowicz et al., J. Immunol., 141:4083-4088 (1988)).
T cell immune response is a complex process that involves cell-cell interactions (Springer et al., Ann. Rev. Immunol., 5:223-252 (1987)), particularly between T and accessory cells such as APCs, and production of soluble immune mediators (cytokines or lymphokines) (Dinarello, New Engl. J. Med., 317:940-945 (1987); and Sallusto, J. Exp. Med., 179:1109-1118 (1997)). This response is regulated by several T-cell surface receptors, including the T-cell receptor complex (Weiss, Ann. Rev. Immunol., 4:593-619 (1986)) and other “accessory” surface molecules (Allison, Curr. Opin. Immunol., 6:414-419 (1994); Springer (1987), supra). Many of these accessory molecules are naturally occurring cell surface differentiation (CD) antigens defined by the reactivity of monoclonal antibodies on the surface of cells (McMichael, ed., Leukocyte Typing Iff, Oxford Univ. Press, Oxford, N.Y. (1987)).
Early studies suggested that B lymphocyte activation requires two signals (Bretscher, Science, 169:1042-1049 (1970)) and now it is believed that all lymphocytes require two signals for their optimal activation, an antigen specific or clonal signal, as well as a second, antigen non-specific signal. (Janeway, supra). Freeman (J. Immunol., 143:2714-2722 (1989)) isolated and sequenced a cDNA clone encoding a B cell activation antigen recognized by MAb B7 (Freeman, J. Immunol., 139:3260 (1987)). COS cells transfected with this cDNA have been shown to stain by both labeled MAb B7 and MAb BB-1 (Clark, Human Immunol., 16:100-113 (1986); Yokochi, J. Immunol., 128:823 (1981); Freeman et al. (1989), supra; and Freeman et al. (1987), supra). In addition, expression of this antigen has been detected on cells of other lineages, such as monocytes (Freeman et al., (1989) supra).
T helper cell (Th) antigenic response requires signals provided by APCs. The first signal is initiated by interaction of the T cell receptor complex (Weiss, J. Clin. Invest., 86:1015 (1990)) with antigen presented in the context of major histocompatibility complex (MHC) molecules on the APC (Allen, Immunol. Today, 8:270 (1987)). This antigen-specific signal is not sufficient to generate a full response, and in the absence of a second signal may actually lead to clonal inactivation or anergy (Schwartz, Science, 248:1349 (1990)). The requirement for a second “costimulatory” signal has been demonstrated in a number of experimental systems (Schwartz, supra; Weaver et al., Immunol. Today, 11:49 (1990)).
CD28 antigen, a homodimeric glycoprotein of the immunoglobulin superfamily (Aruffo et al., Proc. Natl. Acad. Sci., 84:8573-8577 (1987)), is an accessory molecule found on most mature human T cells (Damle et al., J. Immunol., 131:2296-2300 (1983)). Current evidence suggests that this molecule functions in an alternative T cell activation pathway distinct from that initiated by the T-cell receptor complex (June et al., Mol. Cell. Biol., 7:4472-4481 (1987)). Some studies have indicated that CD28 is a counter-receptor for the B cell activation antigen, B7/BB-1 (Linsley et al., Proc. Natl. Acad. Sci. USA, 87:5031-5035 (1990)). The B7 ligands are also members of the immunoglobulin superfamily but have, in contrast to CD28, two Ig domains in their extracellular region, an N-terminal variable (V)-like domain followed by a constant (C)-like domain.
Delivery of a non-specific costimulatory signal to the T cell requires at least two homologous B7 family members found on APCs, B7-1 (also called B7, B7. 1, or CD80) and B7-2 (also called B7.2 or CD86), both of which can deliver costimulatory signals to T cells via CD28. Costimulation through CD28 promotes T cell activation.
CD28 has a single extracellular variable region (V)-like domain (Aruffo et al., supra). A homologous molecule, CTLA-4, has been identified by differential screening of a murine cytolytic-T cell cDNA library (Brunet, Nature, 328:267-270 (1987)). CTLA-4 (CD152) is a T cell surface molecule and also a member of the immunoglobulin (Ig) superfamily, comprising a single extracellular Ig domain. Researchers have reported the cloning and mapping of a gene for the human counterpart of CTLA-4 (Dariavach et al., Eur. J. Immunol., 18:1901-1905 (1988)) to the same chromosomal region (2q33-34) as CD28 (Lafage-Pochitaloff et al., Immunogenetics, 31:198-201 (1990)). Sequence comparison between this human CTLA-4 and CD28 proteins reveals significant homology of sequence, with the greatest degree of homology in the juxtamembrane and cytoplasmic regions (Brunet et al. (1988), supra; Dariavach et al. (1988), supra).
The CTLA-4 is inducibly expressed by T cells. It binds to the B7-family of molecules (primarily CD80 and CD86) on APCs (Chambers et al., Ann. Rev. Immunol., 19:565-594 (2001)). When triggered, it inhibits T-cell proliferation and function. Mice genetically deficient in CTLA-4 develop lymphoproliferative disease and autoimmunity (Tivol et al., Immunity, 3:541-547 (1995)). In pre-clinical models, CTLA-4 blockade also augments anti-tumor immunity (Leach et al., Science, 271:1734-1736 (1996); and van Elsas et al., J. Exp. Med., 190:355-366 (1999)). These findings led to the development of antibodies that block CTLA-4 for use in cancer immunotherapy.
Blockade of CTLA-4 by a monoclonal antibody leads to the expansion of all T cell populations, with activated CD4+ and CD8+ T cells mediating tumor cell destruction (Melero et al., Nat. Rev. Cancer, 7:95-106 (2007); and Wolchok et al., The Oncologist, 13 (Suppl. 4):2-9 (2008)). The antitumor response that results from the administration of anti-CTLA-4 antibodies is believed to be due to an increase in the ratio of effector T cells to regulatory T cells within the tumor microenvironment, rather than simply from changes in T cell populations in the peripheral blood (Quezada et al., J. Clin. Invest., 116:1935-1945 (2006)). One such agent is ipilimumab.
Ipilimumab (previously MDX-010; Medarex Inc., marketed by Bristol-Myers Squibb as YERVOY™) is a fully human anti-human CTLA-4 monoclonal antibody that blocks the binding of CTLA-4 to CD80 and CD86 expressed on antigen presenting cells, thereby, blocking the negative down-regulation of the immune responses elicited by the interaction of these molecules. Initial studies in patients with melanoma showed that ipilimumab could cause objective durable tumor regressions (Phan et al., Proc. Natl. Acad. Sci. USA, 100:8372-8377 (2003)). Also, reductions of serum tumor markers such as CA125 and PSA were seen for some patients with ovarian or prostate cancer, respectively (Hodi et al., Proc. Natl. Acad. Sci. USA, 100:4712-4717 (2003)). Ipilimumab has demonstrated antitumor activity in patients with advanced melanoma (Weber et al., J. Clin. Oncol., 26:5950-5956 (2008); Weber, Cancer Immunol. Immunother., 58:823-830 (2009)). In addition, in a number of phase II and two phase III clinical trials, ipilimumab was shown to increase the overall survival in advanced melanoma patients (Hodi, F. S. et al., “Improved survival with ipilimumab in patients with metastatic melanoma”, New Engl. J. Med., 363:711-723 (2010), and Robert, C. et al., “Ipilimumab plus dacarbazine for previously untreated metastatic melanoma”, New Engl. J. Med., 364:2517-2526 (2011)). Treatment with ipilimumab, however, can result in adverse events in some patients and individual survival outcome may be different.
Provided herein are biomarkers that may be used to predict clinical response of patients to treatment with an immunotherapeutic agent, for example, an anti-CTLA4 antibody such as ipilimumab, prior to receiving the agent, and methods of using the biomarkers for treatment with the immunotherapeutic agent, or for predicting clinical response of a patient treated with the immunotherapeutic agent.
Provided herein are methods for treating a subject having cancer with an immunotherapeutic agent, comprising (a) determining expression level of at least one gene in a blood sample obtained from the subject, wherein the at least one gene is selected from a first group of genes as listed in Table 2 and a second group of genes as listed in Table 3; (b) determining likelihood of clinical response of the subject to the treatment based on the expression level of the at least one gene in the blood sample, wherein the expression level of the at least one gene selected from the first group of genes is positively correlated with the likelihood of clinical response, and wherein the expression level of the at least one gene selected from the second group of genes is negatively correlated with the likelihood of clinical response; and (c) administering to the subject a therapeutically effective amount of the immunotherapeutic agent for treating the cancer.
Also provided herein are methods for predicting likelihood of clinical response of a subject having cancer to treatment with an immunotherapeutic agent, comprising (a) obtaining a blood sample from the subject before the treatment, (b) determining expression level of at least one gene in the blood sample, wherein the at least one gene is selected from a first group of genes as listed in Table 2 and a second group of genes as listed in Table 3; (c) determining likelihood of clinical response to the treatment based on the expression level of the at least one gene in the blood sample, wherein the expression level of the at least one gene selected from the first group of genes is positively correlated with the likelihood of clinical response, and wherein the expression level of the at least one gene selected from the second group of genes is negatively correlated with the likelihood of clinical response.
Also provided herein are methods for determining whether to treat a subject having cancer with a immunotherapeutic agent, comprising (a) obtaining a blood sample from the subject, (b) determining expression level of at least one gene in a blood sample obtained from the subject, wherein the at least one gene is selected from a first group of genes as listed in Table 2 and a second group of genes as listed in Table 3; (c) determining likelihood of clinical response to the treatment based on the expression level of the at least one gene in the blood sample, wherein the expression level of the at least one gene selected from the first group of genes is positively correlated with the likelihood of clinical response, and wherein the expression level of the at least one gene selected from the second group of genes is negatively correlated with the likelihood of clinical response; and (d) determining whether to treat the subject having cancer with the immunotherapeutic agent based on the likelihood of clinical response.
Also provided herein are methods for treating a subject having cancer with an immunotherapeutic agent, comprising (a) determining expression levels of a first gene and a second gene in a blood sample obtained from the subject, wherein the first gene is IL2RB and a second gene is selected from ASGR1 and ASGR2; (b) determining likelihood of longer overall survival of the subject following the treatment based on the expression levels of the first gene and the second gene in the blood sample, wherein the expression levels of the first gene and the second gene are used to calculate a score according to formula:
Score=−C1*Xfirst gene+C2*Xsecond gene,
wherein Xfirst gene and Xsecond gene are normalized mRNA expression levels of the first and the second gene, respectively, and C1 and C2 are each, independently, a number ranging from 0.01 to 3, wherein the score is negatively correlated with the likelihood of longer overall survival; and (c) administering to the subject a therapeutically effective amount of the immunotherapeutic agent for treating the cancer.
Also provided herein are methods for predicting likelihood of longer overall survival of a subject having cancer following treatment with an immunotherapeutic agent, comprising: (a) obtaining a blood sample from the subject before the treatment; (b) determining expression levels of a first gene and a second gene in the blood sample obtained from the subject, wherein the first gene is IL2RB and a second gene is selected from ASGR1 and ASGR2; and (c) determining likelihood of longer overall survival of the subject following the treatment based on the expression levels of the first gene and the second gene in the blood sample, wherein the expression levels of the first gene and the second gene are used to calculate a score according to formula:
Score=−C1*Xfirst gene+C2*Xsecond gene,
wherein Xfirst gene and Xsecond gene are normalized mRNA expression levels of the first and the second gene, respectively, and C1 and C2 are each, independently, a number ranging from 0.01 to 3, wherein the score is negatively correlated with the likelihood of longer overall survival.
Also provided herein are methods for determining whether to treat a subject having cancer with a immunotherapeutic agent, comprising: (a) obtaining a blood sample from the subject; (b) determining expression levels of a first gene and a second gene in the blood sample obtained from the subject, wherein the first gene is IL2RB and a second gene is selected from ASGR1 and ASGR2; and (c) determining likelihood of longer overall survival of the subject following the treatment based on the expression levels of the first gene and the second gene in the blood sample, wherein the expression levels of the first gene and the second gene are used to calculate a score according to formula:
Score=−C1*Xfirst gene+C2*Xsecond gene,
wherein Xfirst gene and Xsecond gene are normalized mRNA expression levels of the first and the second gene, respectively, and C1 and C2 are each, independently, a number ranging from 0.01 to 3, wherein the score is negatively correlated with the likelihood of longer overall survival; and (d) determining whether to treat the subject with the immunotherapeutic agent based on the likelihood of longer overall survival.
Also provided herein are kits for use for the methods disclosed herein. The kits may comprise one or more reagents for determining expression level of at least one gene in a blood sample, wherein the at least one gene is selected from a first group of genes as listed in Table 2 and a second group of genes as listed in Table 3.
Also provided herein are kits for use for the methods disclosed herein. The kits may comprise one or more reagents for determining expression levels of a first gene and a second gene in a blood sample, wherein the first gene is IL2RB and a second gene is selected from ASGR1 and ASGR2.
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.
The methods described herein are based on certain gene expression signatures. The gene expression signatures may be used as biomarkers, e.g., prognostic, predictive biomarkers for clinical efficacy and/or safety.
Provided herein are methods for treating a subject having cancer with an immunotherapeutic agent, comprising (a) determining expression level of at least one gene in a blood sample obtained from the subject, wherein the at least one gene is selected from a first group of genes as listed in Table 2 and a second group of genes as listed in Table 3; (b) determining likelihood of clinical response of the subject to the treatment based on the expression level of the at least one gene in the blood sample, wherein the expression level of the at least one gene selected from the first group of genes is positively correlated with the likelihood of clinical response, and wherein the expression level of the at least one gene selected from the second group of genes is negatively correlated with the likelihood of clinical response; and (c) administering to the subject a therapeutically effective amount of the immunotherapeutic agent for treating the cancer.
Also provided herein are methods of predicting likelihood of clinical response of a subject having cancer to treatment with an immunotherapeutic agent, comprising (a) obtaining a blood sample from the subject before the treatment, (b) determining expression level of at least one gene in the blood sample, wherein the at least one gene is selected from a first group of genes as listed in Table 2 and a second group of genes as listed in Table 3; (c) determining likelihood of clinical response of the subject to the treatment based on the expression level of the at least one gene in the blood sample, wherein the expression level of the at least one gene selected from the first group of genes is positively correlated with the likelihood of clinical response, and wherein the expression level of the at least one gene selected from the second group of genes is negatively correlated with the likelihood of clinical response.
Also provided herein are methods for determining whether to treat a subject having cancer with a immunotherapeutic agent, comprising (a) obtaining a blood sample from the subject, (b) determining expression level of at least one gene in a blood sample obtained from the subject, wherein the at least one gene is selected from a first group of genes as listed in Table 2 and a second group of genes as listed in Table 3; (c) determining likelihood of clinical response of the subject to the treatment based on the expression level of the at least one gene in the blood sample, wherein the expression level of the at least one gene selected from the first group of genes is positively correlated with the likelihood of clinical response, and wherein the expression level of the at least one gene selected from the second group of genes is negatively correlated with the likelihood of clinical response; and (d) determining whether to treat the subject with the immunotherapeutic agent based on the likelihood of clinical response.
Also provided herein are methods for treating a subject having cancer with an immunotherapeutic agent, comprising (a) determining expression levels of a first gene and a second gene in a blood sample obtained from the subject, wherein the first gene is IL2RB and a second gene is selected from ASGR1 and ASGR2; (b) determining likelihood of longer overall survival of the subject following the treatment based on the expression levels of the first gene and the second gene in the blood sample, wherein the expression levels of the first gene and the second gene are used to calculate a score according to formula:
Score=−C1*Xfirst gene+C2*Xsecond gene,
wherein Xfirst gene and Xsecond gene are normalized mRNA expression levels of the first and the second gene, respectively, and C1 and C2 are each, independently, a number ranging from 0.01 to 3, wherein the score is negatively correlated with the likelihood of longer overall survival; and (c) administering to the subject a therapeutically effective amount of the immunotherapeutic agent for treating the cancer.
Also provided herein are methods of predicting likelihood of longer overall survival of a subject having cancer following treatment with an immunotherapeutic agent, comprising: (a) obtaining a blood sample from the subject before the treatment; (b) determining expression levels of a first gene and a second gene in the blood sample obtained from the subject, wherein the first gene is IL2RB and a second gene is selected from ASGR1 and ASGR2; and (c) determining likelihood of longer overall survival of the subject following the treatment based on the expression levels of the first gene and the second gene in the blood sample, wherein the expression levels of the first gene and the second gene are used to calculate a score according to formula:
Score=−C1*Xfirst gene+C2*Xsecond gene,
wherein Xfirst gene and Xsecond gene are normalized mRNA expression levels of the first and the second gene, respectively, and C1 and C2 are each, independently, a number ranging from 0.01 to 3, wherein the score is negatively correlated with the likelihood of longer overall survival.
Also provided herein are methods for determining whether to treat a subject having cancer with a immunotherapeutic agent, comprising: (a) obtaining a blood sample from the subject; (b) determining expression levels of a first gene and a second gene in the blood sample obtained from the subject, wherein the first gene is IL2RB and a second gene is selected from ASGR1 and ASGR2; and (c) determining likelihood of longer overall survival of the subject following the treatment based on the expression levels of the first gene and the second gene in the blood sample, wherein the expression levels of the first gene and the second gene are used to calculate a score according to formula:
Score=−C1*Xfirst gene+C2*Xsecond gene,
wherein Xfirst gene and Xsecond gene are normalized mRNA expression levels of the first and the second gene, respectively, and C1 and C2 are each, independently, a number ranging from 0.01 to 3, wherein the score is negatively correlated with the likelihood of longer overall survival; and (d) determining whether to treat the subject with the immunotherapeutic agent based on the likelihood of longer overall survival.
Also provided herein are kits for use for the methods disclosed herein. The kits may comprise one or more reagents for determining expression level of at least one gene in a blood sample, wherein the at least one gene is selected from a first group of genes as listed in Table 2 and a second group of genes as listed in Table 3.
Also provided herein are kits for use for the methods disclosed herein. The kits may comprise one or more reagents for determining expression levels of a first gene and a second gene in a blood sample, wherein the first gene is IL2RB and a second gene is selected from ASGR1 and ASGR2.
The term “treating” or “treatment” refers to administering an immunotherapeutic agent described herein to a subject that has cancer, or has a symptom of cancer, or has a predisposition toward cancer, with the purpose to cure, heal, alleviate, relieve, alter, remedy, ameliorate, improve, or affect cancer, the symptoms of cancer, or the predisposition toward cancer.
The terms “patient” or “subject” are used interchangeably and refer to mammals such as human patients and non-human primates, as well as experimental animals such as rabbits, rats, and mice, and other animals. Animals include all vertebrates, e.g., mammals and non-mammals, such as sheep, dogs, cows, chickens, amphibians, and reptiles.
The term “immunotherapeutic agent” means an agent that may enhance or alter immune response to a disease or disorder such as cancer. The term “immune response” refers to the concerted action of immune cells, including lymphocytes, antigen presenting cells, phagocytic cells, and granulocytes, and soluble macromolecules produced by the above cells or the liver (including antibodies, cytokines, and complement), that results in selective damage to, destruction of, or elimination from the human body of invading pathogens, cells or tissues infected with pathogens, or cancerous cells. An immunotherapeutic agent may block immuno-regulatory proteins on immune cells, such as cytotoxic T lymphocyte antigen-4 (CTLA-4), Programmed Death 1 (PD-1), PD-1 ligand 1 (PD-L1), OX40, KIR (Killer-cell Immunoglobulin-Like Receptor), or CD137. The immunotherapeutic agent may be, for example, an anti-CTLA-4 antibody, an anti-PD-1 antibody, an anti-PD-L1 antibody, an anti-KIR antibody, an OX40 agonist, a CD137 agonist, IL21 or other cytokines. In some embodiments, the immunotherapeutic agent may be an anti-CTLA-4 antibody, such as ipilimumab or tremelimumab.
The term “effective amount” refers to an amount of an immunotherapeutic agent described herein effective to “treat” a disease or disorder in a subject. In the case of cancer, the effective amount may cause any of the changes observable or measurable in a subject as described in the definition of “treating” and “treatment” above. For example, the effective amount can reduce the number of cancer or tumor cells; reduce the tumor size; inhibit or stop tumor cell infiltration into peripheral organs including, for example, the spread of tumor into soft tissue and bone; inhibit and stop tumor metastasis; inhibit and stop tumor growth; relieve to some extent one or more of the symptoms associated with the cancer, reduce morbidity and/or mortality; improve quality of life; increase or prolong overall survival; or a combination of such effects. In some embodiments, an effective amount may be an amount sufficient to decrease the symptoms of the cancer, or an amount sufficient to prolong overall survival. Efficacy in vivo can, for example, be measured by assessing the duration of survival (e.g. overall survival), time to disease progression (TTP), the response rates (RR), duration of response, and/or quality of life. Effective amounts may vary, as recognized by those skilled in the art, depending on route of administration, excipient usage, and co-usage with other agents.
The term “clinical response” refers to positive clinical outcome of a patient to the treatment defined above, and may be expressed in terms of various measures of clinical outcome. Positive clinical outcome may be considered as an improvement in any measure of patient status, including those measures ordinarily used in the art, such as tumor regression, a decrease in tumor (or lesion) size or growth, a decrease in tumor (or lesion) burden, an increase in the duration of Recurrence-Free interval (RFI), an increase in the time of Progression Free Survival (PFS), an increase in the time of Overall Survival (OS) (from treatment to death), an increase in the time of Disease-Free Survival (DFS), an increase in the duration of Distant Recurrence-Free Interval (DRFI), and/or an increase in the duration of response, and the like. Clinical response may be a complete or partial response, or stable or controlled disease progression. Clinical response may be measured, for example, at 2-4 weeks, 4-8 weeks, 8-12 weeks, 12-16 weeks, 4-6 months, 6-9 months, 9 months to 1 year, 1-2 years, 2-5 years, 5-10 years or longer, from initiation of treatment. For example, clinical response may be measured at week 8, 12, 16, 20, 24, or 36, survival at one year, 18 months, 2 years, 3 years, 4 years, 5 years, or 10 years, from initiation of treatment.
In some embodiments of the methods described herein, the likelihood of clinical response may be expressed in terms of the likelihood of an increase in the time of survival, such as longer overall survival, as compared to some patients, for example, a control or test patient group; patients who have a higher or lower expression level of a gene than the subject; patients who have a higher or lower score based on a formula and expression level of one or more genes; other patients treated with the immunotherapeutic agent; patients not treated with the immunotherapeutic agent; or patients treated with a different anti-cancer agent or procedure (e.g. surgical procedure). In some embodiments of the methods described herein, clinical response is expressed in terms of longer overall survival as compared to patients receiving the immunotherapeutic agent, e.g., ipilimumab or tremelimumab, who have a higher or lower expression level of a gene than the subject; or patients receiving the immunotherapeutic agent, e.g., ipilimumab or tremelimumab, who have a higher or lower score based on a formula and expression level of one or more genes. In some embodiments the term “longer overall survival” may mean overall survival longer than 6, 8, 9, 10, 12, or 18 months, or longer than 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 15, or 20 years. In some embodiments, “longer overall survival” may mean overall survival longer than 10, 20, 30, 40, 50, or 60 months.
In some embodiments, “likelihood of clinical response” may mean higher probability of survival at certain time points, for example, at 6, 8, 9, 10, 12, 18, 20, 30, 40, 50, or 60 months, or 1 year, 2 years, 3 years, 4 years, 5 years, 6 years, 10 years, or more than 10 years, from initiation of treatment, as compared to some patients, for example, a control or test patient group; patients who have a higher or lower expression level of a gene than the subject; patients who have a higher or lower score based on a formula and expression level of one or more genes; other patients treated with the immunotherapeutic agent; patients not treated with the immunotherapeutic agent; or patients treated with a different anti-cancer agent or procedure.
In some embodiments, the likelihood of clinical response may be expressed in terms of likelihood of an increase in the time of progression free survival (PSF). In some embodiments, “likelihood of clinical response” may mean the likelihood of an increase in the time of PSF as compared to some patients, for example, a control or test patient group; patients who have a higher or lower expression level of a gene than the subject; patients who have a higher or lower score based on a formula and expression level of one or more genes; a group of other patients treated with the immunotherapeutic agent; patients not treated with the immunotherapeutic agent; or patients treated with a different anti-cancer agent or procedure. In some embodiments, “likelihood of clinical response” may mean higher probability of PSF at certain time points, for example, at 1 year, 18 months, 2 years, 3 years, 5 years, 10 years, or more than 10 years, from initiation of treatment, as compared to some patients, for example, a control or test patient group; patients who have a higher or lower expression level of a gene than the subject; patients who have a higher or lower score based on a formula and expression level of one or more genes; other patients treated with the immunotherapeutic agent; patients not treated with the immunotherapeutic agent; or patients treated with a different anti-cancer agent.
The term “advanced cancer” means cancer that is no longer localized to the primary tumor site, or a cancer that is Stage III or IV according to the American Joint Committee on Cancer (AJCC). In some embodiments, the subject may have advanced cancer, such as advanced melanoma. Advanced melanoma may be, for example, metastatic melanoma, or stage III or IV melanoma, such as unresectable stage III or IV melanoma.
In some embodiments of the methods described herein, a blood sample may be obtained from the subject having cancer, and the expression level of at least one gene in the blood sample may be determined. The at least one gene may be selected from the genes listed in the first group of genes as listed in Table 2, wherein the expression level of the at least one gene is positively correlated with the likelihood of clinical response. For example, the at least one gene may be selected from IL2RB, KLRK1, G3BP, PPP1R16B, CLIC3, PRF1, SPON2, HOP, GNLY, TMEM161A, PRKCH, RUNX3, EOMES, SLC25A5, GZMB, IMP3, and ZAP70. It may be determined that the subject may have a high likelihood of clinical response, for example, longer overall survival, if the expression level of the at least one gene is higher than a predetermined value.
In some embodiments, the at least one gene may be selected from the genes listed in the second group of genes as listed in Table 3, wherein the expression level of the at least one gene is negatively correlated with the likelihood of clinical response. For example, the at least one gene may be selected from ASGR1, ASGR2, CENTA2, PGLS, MAPBPIP, STX10, C16ORF68, and RAB31. It may be determined that the subject may have a high likelihood of clinical response, for example, longer overall survival, if the expression level of the at least one gene is lower than a predetermined value.
In some embodiments, the expression level of at least two genes in the blood sample may be determined, and the likelihood of clinical response may be predicted based on the expression level of the at least two genes in the blood sample. The at least two genes may be selected from the genes listed in Tables 2 and 3. In some embodiments, the first gene of the at least two genes may be selected from the first group of genes as listed in Table 2, and a second gene of the at least two genes may be selected from the second group of genes as listed in Table 3. For example, the first gene may be selected from IL2RB, KLRK1, G3BP, PPP1R16B, CLIC3, PRF1, SPON2, HOP, GNLY, TMEM161A, PRKCH, RUNX3, EOMES, SLC25A5, GZMB, IMP3, and ZAP70. In some embodiments, the first gene may be IL2RB.
In some embodiments, the second gene of the at least two genes may be selected from ASGR1, ASGR2, CENTA2, PGLS, MAPBPIP, STX10, C16ORF68, and RAB31. For example, the second gene may be selected from ASGR1 and ASGR2.
In some embodiments, the at least two genes may be selected from the pairs of genes (two-gene signatures) listed in Tables 7 and 10 (see the Example section). In some embodiments, the first gene may be IL2RB and the second gene may be ASGR2. In some embodiments, the first gene may be IL2RB and the second gene may be ASGR1.
In some embodiments, the expression level of at least three genes in the blood sample may be determined, and the likelihood of clinical response may be predicted based on the expression level of the at least three genes in the blood sample. The at least three genes may be selected from the genes listed in Tables 2 and 3. A first gene of the at least three genes may be selected from the first group of genes as listed in Table 2. A second gene of the at least three genes may be selected from the second group of genes as listed in Table 3. In some embodiments, the at least three genes may be selected from three-gene groups (three-gene signatures) listed in Table 8 (see the Example section).
In some embodiments of the methods described herein, determining the likelihood of clinical response may comprise subjecting the expression level of the at least two genes to a formula to calculate a score, wherein the formula may be pre-determined by statistical analysis of (a) clinical response of a plurality of patients having the cancer to treatment with the immunotherapeutic agent and (b) the expression level of the at least two genes in pre-treatment blood samples from the plurality of patients. For example, coefficients may be calculated for each gene based on the clinical response and the gene expression level in the pre-treatment blood samples. The statistical analysis may be performed with any statistical method that is suitable for analyzing gene expression data, for example, Cox proportional-hazards (PH) regression.
In some embodiments, the formula for calculating the score is
Score=−C1*Xfirst gene+C2*Xsecond gene,
wherein Xfirst gene and Xsecond gene may be expression level of the first and the second gene, respectively, and C1 and C2 may be, independently, pre-determined values. For example, C1 and C2 may be, independently, pre-determined coefficients of the first and the second gene, respectively, based on gene expression data obtained from pre-treatment blood samples from a patient group. For example, C1 and C2 may be each, independently, a number ranging from 0.01 to 3, wherein the score may be negatively correlated with the likelihood of survival.
In some embodiments, C1 may range from 0.1 to 2.5, from 0.2 to 1.8, or from 0.3 to 1.4. In some embodiments, C1 may be about 1.3.
In some embodiments, C2 may range from 0.1 to 1.2, from 0.1 to 1.0, or from 0.2 to 0.8. In some embodiments, C2 may be about 0.7 to 0.8.
In some embodiments, Xfirst gene and Xsecond gene may be mRNA expression level of the first and the second gene, respectively. For example, Xfirst gene and Xsecond gene may be mRNA expression level of IL2RB and ASGR2, respectively, or Xfirst gene and Xsecond gene may be mRNA expression level of IL2RB and ASGR1, respectively. The mRNA expression level may be normalized. In some embodiments, where the mRNA expression level is measured by microarray, the mRNA expression level may be normalized using a standard robust multichip average (RMA) approach.
In some embodiments, Xfirst gene and Xsecond gene may be mRNA expression level of IL2RB and ASGR2, respectively, C1 may be about 1.3, and C2 may be about 0.7 to 0.8.
The score described above may be compared to a predetermined threshold. A score that is lower than the threshold may be indicative of high likelihood of clinical response, for example, longer overall survival, or higher probability of survival at a time point, while a score that is higher than the threshold may be indicative of low likelihood of clinical response, for example, shorter overall survival, or lower probability of survival at a time point, as compared to a selected or control group of patients, such as, patients treated with the immunotherapeutic agent, patients not treated with the immunotherapeutic agent, or patients treated with a different anti-cancer agent or procedure.
The expression level of the at least one gene may be measured by at least one method selected from microarray, quantitative polymerase chain reaction (qPCR), and flow cytometry. “Microarray” refers to an ordered arrangement of hybridizable array elements, preferably polynucleotide probes, on a substrate.
The immunotherapeutic agent may be an antibody. In some embodiments, the immunotherapeutic agent may be an anti-CTLA4 antibody, such as a human or humanized or chimeric anti-CTLA4 antibody. In some embodiments, the immunotherapeutic agent may be ipilimumab or tremelimumab. In some embodiments, the immunotherapeutic agent may be ipilimumab
In some embodiments, the subject may have cancer selected from melanoma; prostate cancer, prostatic neoplasms, adenocarcinoma of the prostate; lung cancer, e.g., small cell lung cancer and non-small cell lung cancer; ovarian cancer; gastric cancer; adenocarcinoma of the gastric and gastro-esophageal junction; gastrointestinal stromal tumor; glioblastoma; cervical cancer; adenocarcinoma; breast cancer, invasive adenocarcinoma of the breast; pancreatic cancer; duct cell adenocarcinoma of the pancreas; sarcoma, such as chondrosarcoma, clear cell sarcoma of the kidney, endometrial stromal sarcoma, Ewing's sarcoma, osteosarcoma, peripheral primitive neuroectodermal tumor, ovarian sarcoma, soft tissue sarcoma, uterine sarcoma, adult soft tissue sarcoma, and synovial sarcoma; transitional cell carcinoma; urothelial carcinoma; Wilm's tumor and neuroblastoma; lymphoma; leukemia; ocular melanoma, intraocular melanoma, cutaneous melanoma; and kidney cancer. In some embodiments, the subject may have cancer selected from melanoma; prostate cancer, prostatic neoplasms, adenocarcinoma of the prostate; lung cancer, e.g., small cell lung cancer, non-small cell lung cancer; ovarian cancer; gastric cancer; and glioblastoma. In some embodiments, the subject may have advanced melanoma or metastatic melanoma. In some embodiments, the subject may have stage III or IV melanoma, such as unresectable stage III or IV melanoma. In some embodiments, the subject may have prostate cancer. In some embodiments, the subject may have lung cancer, e.g., small cell lung cancer or non-small cell lung cancer.
In some embodiments of the methods described herein, determining the likelihood of clinical response may be based on the gene expression level and at least one additional factor. In some embodiments, the at least one additional factor may be selected from baseline serum LDH level and disease stage (e.g., M category). In some embodiments, the at least one additional factor may be baseline serum LDH level.
In some embodiments, at the time the likelihood of clinical response of the subject is determined, the subject may be not being treated, or may have not been treated, with the immunotherapeutic agent. In some embodiments, the subject may have been treated with the immunotherapeutic agent at the time the likelihood of clinical response of the subject is determined. For example, the expression level of the at least one gene may change over time in the subject. Thus, the likelihood of clinical response may be determined to decide whether to administer (or re-administer) the immunotherapeutic agent to the subject.
Also provided are kits comprising one or more reagents for determining expression level of at least one gene in a blood sample, wherein the at least one gene is selected from a first group of genes as listed in Table 2 and a second group of genes as listed in Table 3. In some embodiments, the one or more reagents may be used to determine mRNA expression level of the at least one gene. For example, the kit may comprise at least one nucleic acid or polynucleotide capable of specifically hybridizing to the at least one gene. For example, the kit may comprise at least one probe set capable of specifically hybridizing to the at least one gene. In some embodiments, the kit may comprise at least one probe set for microarray. In some embodiments, the kit may comprise at least one reagent for performing quantitative polymerase chain reaction (qPCR). In some embodiments, the kit may comprise at least one reagent for flow cytometry.
In some embodiments, the kit may comprise one or more reagents for determining expression level of at least one gene selected from IL2RB, KLRK1, G3BP, PPP1R16B, CLIC3, PRF1, SPON2, HOP, GNLY, TMEM161A, PRKCH, RUNX3, EOMES, SLC25A5, GZMB, IMP3, and ZAP70. In some embodiments, the kit may comprise one or more reagents for determining expression level of at least one gene selected from ASGR1, ASGR2, CENTA2, PGLS, MAPBPIP, STX10, C16ORF68, and RAB31.
In some embodiments, the kit may comprise one or more reagents for determining expression level of at least two genes in the blood sample. The at least two genes may be selected from the genes listed in Tables 2 and 3. In some embodiments, the first gene of the at least two genes may be selected from the first group of genes as listed in Table 2. In some embodiments, a second gene of the at least two genes may be selected from the second group of genes as listed in Table 3. For example, the first gene may be selected from IL2RB, KLRK1, G3BP, PPP1R16B, CLIC3, PRF1, SPON2, HOP, GNLY, TMEM161A, PRKCH, RUNX3, EOMES, SLC25A5, GZMB, IMP3, and ZAP70. For example, the first gene may be IL2RB. In some embodiments, the second gene may be selected from ASGR1, ASGR2, CENTA2, PGLS, MAPBPIP, STX10, C16ORF68, and RAB31. For example, the second gene may be selected from ASGR1 and ASGR2. In some embodiments, the first gene may be IL2RB and the second gene may be ASGR2. In some embodiments, the first gene may be IL2RB and the second gene may be ASGR1. In some embodiments, the at least two genes may be selected from the pairs of genes listed in Tables 7 and 10 (Example section).
In some embodiments, the kit may comprise one or more reagents for determining expression level of at least three genes in the blood sample. The first gene of the at least three genes may be selected from the first group of genes as listed in Table 2. The second gene of the at least three genes may be selected from the second group of genes as listed in Table 3. In some embodiments, the at least three genes may be selected from three-gene groups listed in Table 8 (Example section).
The following Example contains additional information, exemplification and guidance which can be adapted to the practice of this invention in its various embodiments and the equivalents thereof. The example is intended to help illustrate the invention, and is not intended to, nor should it be construed to, limit its scope.
Ipilimumab, a fully human monoclonal antibody against the cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4), promotes antitumor immunity and improves overall survival (OS) in metastatic melanoma patients.1,2
Several markers have been found to associate with OS or tumor response in patients receiving ipilimumab, including tumor expression of immune-related genes,3 changes in absolute lymphocyte count (ALC),4 EOMES-positive CD8+ T cells,5 ICOShi CD4+ T cells,6 NY-ESO-1 seropositivity,7 polyfunctional NY-ESO-1 specific T cell responses,8 and baseline myeloid-derived suppressor cell (MDSC) levels.9
Despite these insights, no marker has yet emerged that meets five key criteria: (1) can be measured prior to treatment in a readily-accessible sample (e.g. blood), (2) is significantly associated with OS in patients receiving ipilimumab, (3) has a clear mechanistic explanation rooted in the underlying biology, (4) has been repeated in a test cohort independent from the training cohort on which it was developed, and (5) has an effect of a magnitude sufficient to provide clinically meaningful predictions of OS.
In this study biomarkers that meet those five criteria were identified by analyzing gene expression levels in blood drawn from 88 patients prior to receiving ipilimumab and then testing candidate predictive models in a separate cohort of 69 patients.
1. Study Design
The multicenter, phase II clinical trial CA184-004 enrolled 82 previously-treated and untreated patients with unresectable stage III or IV melanoma, randomized 1:1 into 2 arms to receive up to 4 intravenous infusions of either 3 or 10 mg/kg ipilimumab every 3 weeks (Q3W) in the induction phase. In the phase II CA184-007 trial, treatment-naïve or previously treated patients with unresectable stage III/IV melanoma (N=115) received open-label ipilimumab (10 mg/kg every 3 wks for four doses) and were randomized to receive concomitant blinded prophylactic oral budesonide (9 mg/d with gradual taper through week 16) or placebo. Data for baseline (pre-treatment) serum lactate dehydrogenase (LDH) were available for 154 out of 157 patients in the two studies (67 in CA184004 and 87 in CA184007). Clinical variables including OS and disease stage (M category) were recorded. Patient disease stage (M category) information for each cohort appears in Table 1. Complete study design, patient characteristics and endpoint reports of these trials have been described elsewhere10,11. Both studies were conducted in accordance with the ethical principles originating from the current Declaration of Helsinki and consistent with International Conference on Harmonization Good Clinical Practice and the ethical principles underlying European Union Directive 2001/20/EC and the United States Code of Federal Regulations, Title 21, Part 50 (21 C.F.R. 50). The protocols and patient informed consent forms received appropriate approval by all Institutional Review Boards or Independent Ethics Committees prior to study initiation. All participating patients (or their legally acceptable representatives) gave written informed consent for these biomarker focused studies.
2. Affymetrix Gene Expression Analysis
Whole blood was collected prior to treatment. Total RNA was extracted using the Prism 6100 (Applied Biosystems, Foster City, Calif.), purified by RNAClean Kit (Agencourt Bioscience Corporation; Beverly, Mass.), and evaluated on a 2100 Bioanalyzer (Agilent Technologies, Santa Clara, Calif.). Complementary DNA preparation and hybridization on HT-HG-U133A 96-array plates followed manufacturer's protocols (Affymetrix, Santa Clara, Calif.).
3. Computational Analysis
The training cohort consisted of 88 patients from CA184007, and the test cohort comprised 69 patients from CA184004. All raw microarray data for the training and test cohorts were normalized together using a standard robust multichip average (RMA) approach,12 which combines background adjustment, quantile normalization, and summarization, implemented in the Bioconductor package (v2.10, http://www.bioconductor.org)13 of the statistical computing language R (v2.15.1, http://www.r-project.org). For genes with multiple probes, the probe with the greatest mean expression level was selected.14
Feature Selection
A pathwise algorithm for Cox proportional-hazards (PH) regression, regularized by a lasso or elastic-net penalty, was applied to all probe sets for unique genes in the pre-treatment gene expression data from the training cohort to identify genes predictive of OS. This method has been previously described at length15 and is implemented as the glmnet package in the statistical computing language R. For much of the work the glmnet default alpha=1 (lasso penalty) was used, but it was also verified that alpha=0.95 yielded comparable results.
As a second method, a univariate Cox regression was applied to the pre-treatment gene expression data from the training cohort to rank the genes that were most significantly associated with OS.
Two-Gene Signature: Coefficient Estimation and Threshold Selection
Cox PH regression was used to estimate the coefficients for selected genes in order to best fit the OS data in the training cohort. Using the resulting coefficients and the gene expression values of the candidate genes, a two-gene score for each patient was calculated. For purposes of illustration, these scores were dichotomized by application of a classification threshold. This threshold was selected by minimizing, over all possible thresholds, the log-rank test p-value for comparing the OS curve in training-cohort patients with scores below the threshold to that in training-cohort patients with scores above the threshold.
Two-Gene Signature: Testing
For each patient in the test cohort, the coefficients previously estimated using the training cohort were used to calculate a score. Then the previously selected threshold was applied to classify patients into 2 groups, the Kaplan-Meier method16 was used to estimate the survival functions, and a log-rank test was used to compare OS in the 2 groups.
The scores for the training and test cohorts were then pooled, and the previously selected classification threshold was applied. Survival curves for the resulting 2 groups again were estimated by the Kaplan-Meier method and compared using a log-rank test.
Three-Factor Signature
Multivariable Cox PH regression was used to explore the relationship between selected genes and two of the most established prognostic factors in advanced melanoma: baseline serum lactate dehydrogenase (LDH) levels and disease stage (M category).17
An optimal three-factor signature (combining the previously-identified two-gene signature with LDH) was identified by performing a multivariable Cox regression on the training cohort to determine the best-fitting coefficients. Next, the comprehensive threshold exploration method described above was used to determine a good threshold.
Cell-Type Enrichment Analysis
A statistical method was developed to determine whether genes specific to particular cell types were over-represented in the set of genes positively associated with OS, and whether genes specific to particular cell types were over-represented in the set of genes negatively associated with OS. The publicly available Broad Institute Differentiation Map Portal (DMAP)18 data set was used. This data set contains a comprehensive collection of genome-wide gene expression profiles for all major human hematopoietic cell types in several replicates. To evaluate a given gene's cell-type specificity, for each gene profiled in the DMAP data an enrichment score was computed based on a published algorithm.19 Each enrichment score is a measure of how specific the expression of a particular gene is for a particular cell type. Next, for each cell type, cell-type specific gene sets were compiled using an enrichment score cut off of 10 as the criterion for inclusion of the gene into the gene set. Finally, separately for the set of genes positively associated with OS and the set of genes negatively associated with OS, a hypergeometric test was used to evaluate whether each gene set was enriched in genes specific for each of the cell types. The resulting hypergeometric p-values are reported in Tables 15-16, along with the hypergeometric p-values adjusted to control for false discovery rate (FDR) using the Benjamini-Hochberg method.
qPCR Data Analysis
Quantitative polymerase chain reaction (qPCR) was conducted using the TAQMAN® Gene Expression Assay (Life Technologies/Applied Biosystems) with Assay IDs Hs00172872_ml (EOMES) (target sequence RefSeq ID: NM—005442.2) and Hs99999905_ml (GAPDH) (target sequence RefSeq ID: NM—002046.4), respectively, according to methods previously described.3 The qPCR data were normalized using GAPDH as the housekeeping gene. An optimal threshold was identified using methods described above, and then a Kaplan-Meier plot was generated using R. The association with OS was determined by univariate Cox regression. In addition, Spearman's rank correlation was determined between the normalized EOMES expression by qPCR and the expression of selected genes by microarray.
Two analytical methods were used to identify genes predictive of OS: elastic-net regularized Cox PH regression, and univariate (unregularized) Cox PH regression.
When the elastic-net regularized regression method was applied to the gene expression profiles for the selected probe sets for 13,341 unique genes from 88 patients in the training cohort (treated in the CA184007 trial), with the regularization parameter, lambda between 0.3713 and 0.2443, it identified a combination of two genes predictive of OS: IL2RB (interleukin-2 receptor beta, also known as CD122; probe 205291_at) and ASGR1 (asialoglycoprotein receptor 1; probe 206743_s_at). Relaxing lambda to a number between 0.2443 and 0.2226 to identify the next gene yielded ASGR2 (asialoglycoprotein receptor 2; probe 206130_s_at). Further, the gene expression profiles of ASGR1 and ASGR2 were found to be highly correlated in the training cohort (Spearman's rank correlation, R=0.562, P=1.22×10−14) (Table 4). The two genes also have a close biological relationship, encoding two proteins that together form the asialoglycoprotein receptor.20
Applying the univariate (unregularized) Cox PH regression approach to the pre-treatment blood gene expression data from the 88 patients in the training cohort yielded 73 genes associated with OS with p<0.005 (Table 5), including a subset of 16 genes with p<0.001 (Table 6). IL2RB had the smallest p-value (p=4.62×10−7) in the training cohort, and higher expression of this gene was positively associated with longer survival (hazard ratio=0.28, 95% CI=0.17 to 0.46). Among the genes for which higher expression was associated with shorter survival (hazard ratio>1), ASGR1 and ASGR2 had the smallest p-values in the training cohort (P=1.18×10−6 and 1.42×104, respectively).
Next, the 73 genes identified above were analyzed in all 2,628 possible two-gene and all 62,196 possible three-gene combinations. For each such combination, an unregularized Cox PH model to predict OS as an additive function of the two or three expression values was fit to the training-cohort data. A likelihood-ratio test was used to compare each model to a null (constant) model. Among the top 10 two-gene signatures in the training cohort (Table 7) by p-value (where p-value is used solely for ranking), two stood out as being the highest ranked: IL2RB+ASGR1 (p=1.56×10−10) and IL2RB+ASGR2 (p=2.79×10−10).
The three-gene signature with the smallest p-value in the training cohort was comprised of the combination of IL2RB, ASGR2, and CAT (catalase, probe 201432_at), p=2.41×1041. However, the p-value of this signature in the test cohort (as determined by applying the training model coefficients and threshold to the test cohort and calculating the log-rank p-value) was p=6.40×10−3, not below the p<0.001 threshold. To further explore the potential value of adding a third gene, possible three-gene signatures with a p<0.001 in the test cohort were examined. Among these, the three-gene signature with the smallest p-value in the training cohort (p=1.94×10−10) was IL2RB+ASGR2+ZBP1 (Z-DNA binding protein 1, probe 208087_s_at), with a significant p value also in the test cohort (p=9.53×10−4). For the training cohort, adding a third gene decreased the p-value for association with OS by at most one order of magnitude over the best two-gene signature (IL2RB+ASGR2). Furthermore, time-dependent Receiver Operating Characteristic (ROC) curves at 12 months21 show that the majority of the predictive power comes from IL2RB+ASGR2 (
In summary, two different methods converged on two signatures associated with OS in metastatic melanoma patients receiving ipilimumab: IL2RB+ASGR1 and IL2RB+ASGR2. Both signatures yielded comparable log-rank p-values and Kaplan-Meier plots in the training, test, and pooled cohorts (IL2RB+ASGR2,
The two coefficients for combining IL2RB and ASGR2 in a two-gene signature to predict OS were estimated using unregularized Cox PH regression in the training cohort. The estimated coefficients were −1.312 for IL2RB and 0.748 for ASGR2 (Table 9). The two-gene score for each patient could thus be calculated from the following equation: −1.312*XIL2RB+0.748*XASGR2, where Xj gives the log 2-scale RMA-normalized expression level for gene j. The signs of the coefficients indicate that higher expression of IL2RB was associated with longer survival (lesser hazard) whereas higher expression of ASGR2 was associated with shorter survival (greater hazard).
In order to generate Kaplan-Meier plots evaluating the association of the two-gene score with OS, it was necessary to select a threshold separating scores for high risk patients (shorter survival) from those with low risk (longer survival). Thus, each possible threshold was applied to classify the training cohort into two risk groups, and a log-rank test was used to compare OS in the two groups (
In order to test our findings from the training cohort, the same coefficients and threshold were applied to the gene expression data from patients in the test cohort (CA184004 trial). The two-gene signature maintained a highly significant association with OS in the test cohort (log-rank p=1.74×10−4) with a clear separation of the survival curve estimates (
Finally, for illustration purposes, training- and test-cohort scores were pooled for the same two-gene signature, using the coefficients and threshold estimated from the training-cohort data alone, and again estimated OS curves for the two resulting risk groups (
While the two-gene signatures comprised of IL2RB+ASGR2 and IL2RB+ASGR1 were optimal with regard to our model-selection criteria in the training cohort, and were significant and had good predictive accuracy in the test cohort, for completeness this study sought to identify additional pairs of genes that were strongly associated with OS in both the training and test cohorts. For the 2,628 possible two-gene signatures derived from the 73 best genes in the training cohort, Cox PH regression was used to estimate the coefficients and p-values in the training cohort, then the coefficients from the training cohort was applied to the test cohort and the resulting p-values determined. All signatures that had p<0.001 in both the training cohort and the test cohort were retained (Table 10). Then the same procedure was used in reverse: all genes with a univariate Cox regression p<0.005 in the test cohort were selected, then all two-gene combinations formed from those genes were evaluated and the ones with p<0.001 in both the test and training cohorts were retained. More than 88% of the resulting signatures included IL2RB or ASGR2 (Table 11).
To determine whether the two-gene signature, IL2RB+ASGR2, was an independent predictor of OS given established prognostic factors in metastatic melanoma, we performed a multivariable Cox PH regression analysis including the expression levels of each of the genes or that of the two-gene signature as well as baseline serum LDH levels or disease stage (M category). The results suggest that the two-gene signature was an independent predictor of OS in this context in the training, test, and pooled cohorts (Table 12). Each p-value is for a likelihood-ratio test comparing the full model to a model that excludes the corresponding variable. Similarly, expression of each of the individual genes that comprise the two-gene signature (Table 13) also was an independent predictor of OS given baseline serum LDH levels or disease stage (M Category) in the training, test, and pooled cohorts. The two-gene signature was also an independent predictor of OS when absolute lymphocyte count (ALC) at baseline or prior to the third ipilimumab dose was added to the multivariable Cox PH model (Table 14).
As it was established that LDH and the two-gene signature, IL2RB+ASGR2, were independent predictors of OS, it was next determined whether the two-gene signature could be improved by combining it with LDH to create a three-factor signature. Coefficients were estimated using Cox PH regression on the training cohort (0.00158 for LDH and 0.816 for the two-gene signature). The three-factor score for each patient could thus be calculated from the following equation: 0.00158*YLDH+0.816*(−1.312*XIL2RB+0.748*XASGR2), where Yj gives the concentration of factor j. Next the log-rank p-value was calculated for all possible thresholds. The threshold with the smallest p-value was −4.437 (
It was next determined whether using two thresholds instead of one could provide better separation among survival curves. Using the three-factor signature described above with coefficients from the training cohort, two-threshold exploration was performed on the pooled cohort. Using thresholds at both −5.29 and −3.62 (
Time dependent ROC curves at 12 months were then plotted for both the two-gene signature (IL2RB+ASGR2) and the three-factor signature (IL2RB+ASGR2+LDH) in the training cohort (
This study also sought to determine whether the various gene sets emerging in the above analyses were characteristic of particular blood cell types. Among the genes most highly correlated with IL2RB across the pooled training and test cohorts, the top two were PRF1 (perforin 1, probe 214617_at) (Spearman R=0.735, p=2.77×10−28) and RUNX3 (runt-related transcription factor 3, probe 204197_s_at) (Spearman R=0.729, p=1.24×10−27) (Table 5), genes that are highly interrelated, established to be associated with T-cells,22,23 and point clearly to underlying biological mechanisms (see Discussion). Also present among the 100 genes most correlated with IL2RB are a number of other genes established to be associated with T-cells including CD247,24 LCK,25 FYN,25 ZAP70,26 CBLB,27 and TXK.28 RUNX3, PRF1, and ZAP70 are also present on the list of genes associated with OS by univariate Cox regression with p<0.005. RUNX3 has been reported to induce transcription of PRF1 and EOMES (eomesodermin),22 which has been implicated in the regulation of IL2RB expression.29 These analyses pointed to a role for EOMES as a central regulator of the expression of various genes in our model (
Among the genes most highly correlated with ASGR2 are ASGR1, CD14 (cluster of differentiation 14, probe 201743_at) (Spearman R=0.588, p=3.75×10−16), and CD33 (cluster of differentiation 33, probe 206120_at) (Spearman R=0.457, p=1.34×10−9) (Table 5). CD14 expression is a characteristic of myeloid-derived suppressor cells (MDSCs) in melanoma patients,9 and CD33 expression is a characteristic of myeloid cells more generally.30 Our cell type enrichment analysis found that among the 73 genes associated with OS by univariate Cox PH regression (p<0.005), the set of genes negatively associated with OS was most enriched in genes specific for CD14+ monocytes (P=2.17×10−7) (P values by hyper-geometric test as described in Methods), and also highly enriched in genes specific for CD33+ monocytes (P=2.62×10−4) as well as two types of granulocytes (Table 15). This is illustrated graphically (
The set of genes positively associated with OS was most enriched in genes specific for two types of NK cells (CD56+CD16+CD3−, P=2.50×10−18 and CD56−CD16−CD3−, P=7.95×10−12) and two types of T cells (CD8+CD62L−CD45RA+, P=3.41×10−17 and CD8+CD62L−CD45RA−, P=8.05×10−14) (Table 16) (P values by hyper-geometric test as described in Methods). This is illustrated graphically (
Taken together, these analyses suggest that greater expression of genes more highly expressed in natural killer (NK) and T-cells (such as IL2RB) was associated with longer survival, while greater expression of genes expressed in CD14+ cells and other myeloid lineage cells (such as ASGR1 and ASGR2) was associated with shorter survival (
Ongoing research aims to discover biomarkers that could select patients with an enhanced benefit/risk profile. Whereas ipilimumab has shown significant survival benefit in a subset of metastatic melanoma patients, in some patients the treatment can result in adverse events. Thus, identification of biomarkers that can predict a patient's response and are easily measured in peripheral blood is important. In the present study, a novel approach was used to identify blood gene-signatures that may predict OS in metastatic melanoma patients receiving ipilimumab.
When using microarray data to develop predictive gene-signatures there is a high likelihood of developing a signature that may be strongly associated with OS in a training cohort, but not significantly associated with OS in a test cohort, due to over-fitting in the training cohort. Signatures consisting of large numbers of genes are more likely to suffer from over-fitting and are less practical in the clinical context.
Using gene expression microarray data from a training cohort of 88 patients, two independent methods were applied to evaluate association of gene expression with OS. Results from both methods pointed to a lead two-gene signature of IL2RB+ASGR2 that was highly associated with OS in the training cohort. Using these two genes, a signature was calculated that included two coefficients and a threshold in the training cohort, and it was determined that the same signature was also significantly associated with OS in an independent test cohort of 69 patients (p<0.001). The signature also had strong predictive performance in the independent test cohort (AUC=0.818 for a time-dependent ROC curve at 12 months).
The size of the signature is noteworthy. While signatures comprised of many genes carry risk of over-fitting, a two-gene signature significantly mitigates this risk. Adding additional genes improved the signature incrementally, but in this study, the majority of the predictive power came from the combination of two top genes, IL2RB and ASGR2.
Mechanistic investigation of the two genes with expression most highly correlated with that of IL2RB (RUNX3 and PRF1) yielded insights into its underlying biology. RUNX3 has been reported to induce transcription of PRF1 and EOMES (eomesodermin),22 which has been implicated in the regulation of IL2RB expression.29 Based on the high correlation between IL2RB, RUNX3, and PRF1 expression and the mechanistic linkage between EOMES, RUNX3 and IL2RB, it may be hypothesized that EOMES is a core transcription factor that underlies the observed coexpression of IL2RB, RUNX3 and PRF1 in the data. Further analyses of EOMES by qPCR supported this notion, as we found strong correlation of the expression levels of EOMES and other genes in our model. Greater baseline expression levels of this gene were also associated with longer survival in the data set. Moreover, a direct relationship between EOMES and CTLA-4 has been established,31 as well as interactions between EOMES and IFNγ,22 the factor underlying many of the tumor chemokine changes linked with ipilimumab response (FIG. 3A).3
Mechanistic investigation of ASGR2 linked it to myeloid cells and particularly MDSCs, as its expression was highly correlated with the MDSC surface markers CD14 and CD33.9,30 MDSCs have the capacity to suppress both the cytotoxic activities of natural killer (NK) and natural killer T (NKT) cells, and the adaptive immune response mediated by CD4+ and CD8+ T cells. MDSCs act through multiple pathways including upregulation of nitric oxide synthase 2 (NOS2) and production of arginase 1 (ARG1). ARG1 and NOS2 metabolize L-arginine and either together, or separately, block translation of the T cell CD3 zeta chain, inhibit T cell proliferation, and promote T cell apoptosis.32 Additionally, MDSCs are believed to secrete immunosuppressive cytokines such as TGFβ and induce regulatory T cell development.30 High frequency of MDSCs have been reported in the peripheral blood of patients affected by breast, lung, renal and head and neck carcinomas33 and in melanoma.34
While in this study gene expression was mainly measured via microarray, it may also be assayed via quantitative polymerase chain reaction (qPCR). Moreover, IL2RB and ASGR2 are both cell surface markers and therefore may be detected via flow cytometry. The magnitude of the two-gene signature may change over time in a given patient (either inherently or in response to additional therapies such as a CD137-agonist), and may be monitored to determine the best times to administer or re-administer ipilimumab.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/US13/69975 | 11/14/2013 | WO | 00 |
| Number | Date | Country | |
|---|---|---|---|
| 61726953 | Nov 2012 | US |
