WO2015092444A2 - Biomarkers - Google Patents

Abstract

Biomarkers are provided herein that are useful in predicting patient responses to treatment of disease, and to related assays using those biomarkers. Biomarkers provided herein are capable of identifying patients who will benefit from treatment with IGF-1R inhibitors.

Description

BIOMARKERS

The present invention relates to biomarkers useful in predicting patient responses to treatment of disease, and to related assays using those biomarkers. In more detail, the invention relates to one or more biomarkers capable of identifying patients who will benefit from treatment with IGF-1 R inhibitors.

Type 1 insulin-like growth factor receptor (IGF-1 R) signals via recruitment of adaptors, including insulin receptor substrate- 1 (IRS-1 ) and She, to drive proliferation, invasion and cell survival (Chitnis MM, Yuen JS, Protheroe AS, Pollak M, Macaulay VM; Clin Cancer Res 2008; 14: 6364-70; incorporated herein by reference). In both a clinical and experimental setting, it has been found that low IGF bioactivity protects from tumor development and metastasis, suggesting that IGFs provide a potent pro-tumorigenic signal (Pollak M, Nat Rev Cancer 2012; 12: 159-69; and Guevara-Aguirre J, et al. Science Translational Medicine 2011 ; 3: 70ra13; both of which are incorporated herein by reference). In early clinical trials, IGF-1 R inhibitory drugs induced objective regressions, some dramatic and durable, as monotherapy in Ewing sarcomas and other uncommon tumors, and in combination with chemotherapy or targeted agents in common cancers (Olmos D, et al. Lancet Oncol 2010; 11 : 129-35; Tolcher AW, et al. J Clin Oncol 2009; 27: 5800-7; Javle MM, et al. J Clin Oncol 2011 ; 29: suppl: abstr 4026; Molife LR, et al. Br J Cancer 2010; 103: 332-9; Macaulay VM, et al. J Clin Oncol 2011 ; 29: suppl; abstr 3098; Macaulay VM, et al. Ann Oncol 2013; 24: 784-91 ; all of which are incorporated herein by reference).

Thus, inhibition of IGF-1 R represents a potentially promising route for treatment of many tumours. For example, almost all solid tumours express IGF-1 R (Yuen and Macaulay Expert Opin Ther Targets 2008; 12: 589-603), and many tumour types overexpress IGF- 1 R (ie contain more IGF-1 R than the equivalent normal tissue), including: prostate cancer (Hellawell GO et al Cancer Res 2000; 62: 2942-50; incorporated herein by reference)

bladder cancer (Rochester MA et al BJU Int 2007; 100: 1396-401 ; incorporated herein by reference)

renal cancer (Yuen JS et al Oncogene 2007; 26: 6499-508; incorporated herein by reference) small cell and non-small cell lung cancer (Minuto et al Cancer Res 46: 985, 1988; Macaulay et al Cancer Res 50: 251 1 -7, 1990, Shigematsu et al Cancer Res 50: 2481 , 1990; reviewed in Macaulay V Br J Cancer 65: 311 -320, 1992; incorporated herein by reference).

colon cancer (Hakam et a/ Hum Pathol 1999; 30: 1128-33; incorporated herein by reference)

pancreatic cancer (Hakam A et al Dig Dis Sci 2003; 48: 1972-8; incorporated herein by reference)

melanoma (Kanter Leensohn L et al Growth Factors 2000; 17; 193-202; incorporated herein by reference)

Therefore, many different types of cancer could potentially be suitable for treatment with IGF-1 R inhibitory drugs. In practice, however, only a few tumour types have been shown to respond to IGF-1 R inhibition as monotherapy i.e. to show objective tumour regressions of >30% from baseline measurements - the currently used threshold for Objective response' by RECIST criteria - Response Evaluation Criteria in Solid Tumours. These are: Ewing sarcoma family tumours; desmoplastic small round cell tumours (DSRCT); and neuroendocrine tumours including carcinoid (Tolcher et al J Clin Oncol 34: 5800-5807, 2009; Olmos et al Lancet Oncology 1 1 :129-135, 2010; Kurzrock et al Clin Cancer Res 16: 2458-65, 2010; Pappo et al J Clin Oncol 29: 4541 -47, 201 1 , Atzori ef a/ Clin Cancer Res 17: 6304-12, 201 1 ; Tap et al J Clin Oncol 30: 1849-56, 2012, Soria et al Eur J Cancer 49: 1799-807, 2013; all of which are incorporated herein by reference). At present it is not known why these tumour types should be particularly sensitive to IGF- 1 R inhibition.

There have been recent reports of single agent activity in common cancers, so far in abstract only: ovarian cancer (Ray Coquard et al J Clin Oncol 31 : suppl; abstr 5515, 2013; incorporated herein by reference); and non-small cell lung cancer (Ekman et al J Clin Oncol 31 : suppl; abstr 7539, 2013; incorporated herein by reference)

However, IGF-1 R inhibitory drugs have rarely induced objective responses as monotherapy when given to patients with solid tumours, but these drugs have shown evidence of clinical activity in that they can induce: • durable disease stabilizations in patients with multiple tumour types including breast cancer, liver, colorectal, prostate, leiomyosarcoma, cervical and endometrial cancer, prostate, adrenocortical cancer, pancreatic cancer, thymoma (Haluska Clin Cancer Res 13: 5834, 2007; Rodon et al Mol Cancer Ther 7: 2575- 88, 2008; Chitnis ef a/ Clin Cancer Res 14: 6364- 70, 2008; Yee D. J Natl Cancer

Inst 104: 975-981 , 2012; all of which are incorporated herein by reference);

• objective responses in subsets of patients when IGF-1 R inhibitory drugs are added to chemotherapy or targeted agents, in patients with:

- bone and soft tissue sarcomas (Schwartz et al Lancet Oncol 14: 371 -82, 2013; incorporated herein by reference)

- melanoma, leiomyosarcoma (Macaulay et al Ann Oncol 24: 784-91 , 2013; incorporated herein by reference)

- non-small cell lung cancer, colorectal cancer, chordoma (Macaulay et al J Clin Oncol 29: suppl; abstr 3098, 201 1 ; incorporated herein by reference); - gastrointestinal cancers including pancreatic and colorectal cancers

(Reidy et al J Clin Oncol 28: 4240-6, 2010, Golan and Javle Oncology 25: 518-26, 201 1 , McCaffery et al Clin Cancer Res 19: 4282-9, 2013; all of which are incorporated herein by reference);

- ovarian cancer (Harb et al J Clin Oncol 31 : suppl; abstr e13502, 2013; incorporated herein by reference)

- Breast cancer, non-small cell lung cancer, nasopharyngeal cancer, colon cancer, melanoma, gastrointestinal stromal tumour (GIST; Mahadevan et al Clin Oncol 31 : suppl; abstr e13574, 2013; incorporated herein by reference)

However, so far it has not been possible to demonstrate the clinical efficacy of inhibition of IGF-1 in unselected patients in Phase III clinical trials (Yee D. J Natl Cancer Inst 2012; 104: 975-81 ; King H. et al. Cancer Treat Rev 2014; 40(9): 1096-1 105; both incorporated herein by reference). There appear to be two major problems that currently are preventing this approach being applied successfully in clinical treatment. Firstly, it has not been possible to identify patients likely to respond positively to IGF-1 R inhibition. Secondly, IGF-1 R is widely expressed in cancers, and engages in complex cross-talk with numerous cell signalling pathways. This makes it very difficult to prioritize selection of drugs to combine with IGF-1 R inhibitory drugs in clinical trials. There is accordingly a need to improve the current treatment of cancer based on IGF-1 R inhibition. There is a need to be able to identify patients likely to respond to treatment with IGF-1 R inhibition. There is a need to identify IGF-1 R cell signalling pathways. There is in particular a need to identify additional drugs that can be used effectively in combination with IGF-1 R inhibitors to treat cancer.

The present invention meets one or more of the above needs by providing a method of determining whether a patient is likely to benefit from treatment with IGF-1 R inhibition, the method comprising determining the level, in a tumour sample taken from the patient, of one or more biomarkers selected from the list consisting of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C, BRCA2, CDK1 and/or RAD51 , wherein a high level of the biomarker in the sample indicates that a patient is unlikely to benefit from treatment with IGF-1 R inhibition and a low level or absence of biomarker in the sample indicates that the patient is likely to benefit from treatment with IGF-1 R inhibition. Preferred biomarkers may be selected from the list consisting of: CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C and/or RAD51. Preferred biomarkers may be selected from the list consisting of: DVL3, HUNK, LMTK3, CDKN2C, CNKSR1 , DUSP5 and/or RAD51. It is well known that high level of biomarker in the sample is typically associated with high activity of that biomarker in the sample. Low level of biomarker in the sample is typically associated with low activity of that biomarker in the sample. Thus, the invention also provides a method of determining whether a patient is likely to benefit from treatment with IGF-1 R inhibition, the method comprising determining the activity, in a tumour sample taken from the patient, of one or more biomarkers selected from the list consisting of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C, BRCA2, CDK1 and/or RAD51 , wherein reduced activity of biomarker in the sample indicates that the patient is likely to benefit from treatment with IGF-1 R inhibition.

The present invention also provides a method of determining whether a patient is likely to benefit from treatment with IGF-1 R inhibition, comprising determining whether the patient has impaired homologous recombination. The method comprises determining the level or activity, in a tumour sample taken from the patient, of one or more biomarkers associated with homologous recombination, wherein biomarker levels or activity associated with impaired homologous recombination indicate that the patient is likely to benefit from treatment with IGF-1 R inhibition. Methods of determining whether the patient has impaired homologous recombination are known in the art, for example by determining the level of RAD51 foci formation in response to cisplatin (Birkelback et al. J. Thorac. Oncol. Mar 2013; 8(3) 279-286, incorporated herein by reference), and comparing said level to a control, as described herein. Impaired homologous recombination indicates that the patient is likely to benefit from treatment with IGF-1 R inhibition. Impaired homologous recombination may be characterised by low level of one or more biomarkers of homologous recombination in the sample. In some embodiments, said low level of one or more biomarkers of homologous recombination is identified by a Intensity X Percent Score in the sample that is lower than the Intensity X Percent Score of a control (see below). In some embodiments, said low level of one or more biomarkers of homologous recombination is identified by an Intensity X Percent Score of less than 5. Preferred biomarkers of homologous recombination are selected from RAD51 , BRCA2 and/or CDK1.

Level of biomarker in a sample can be scored by many means, e.g. by the intensity of staining and the % of the tumour that is stained. Said staining is preferably by immunohistochemistry. For example, measurements of staining intensity can be scored as follows: 0 = no staining; 1 = weak staining; 2 = moderate staining; 3 = intense staining. Measurements of percentage staining can be scored as follows: 0 = no staining; 1 = 1-10% positive; 2 = 1 1 -50% positive; 3 = 51-80% positive; 4 = >80% positive. The Intensity x Percent Score (IPS) can be used to compare DVL3 staining between tumours (with controls in each staining run for quality control purposes). Using this semi-quantitative scoring system, the relative levels appear to vary between tumour types. IPS values can be obtained for each of the biomarkers disclosed herein.

As set out in the Examples at pages 21 and 22, a "high level" of biomarker in the sample corresponds to an IPS of greater than 5, and a "low level" of biomarker in the sample corresponds to an IPS of less than 5.

In some embodiments, the IPS is determined relative to a control, such as a quality control. In some embodiments, the control is tissue sample in which the subject biomarker is absent. In some embodiments, the control is cells in which the subject biomarker is absent. Impaired homologous recombination may be characterised by reduced activity of one or more biomarkers of homologous recombination in the sample, compared to a control. For example, impaired homologous recombination may be identified by determining the kinase activity of CDK1 in a sample, comparing said activity with a control, and identifying reduced kinase activity in the sample. Suitable controls are readily contemplated by the skilled person. In some embodiments, said reduced activity corresponds to an activity of 90% or less, c, e.g. 90% or less, 80% or less, 70% or less, 60% or less, 50% or less, 40% or less, 30% or less, 20% or less, 10% or less, 5% or less or 1 % or less compared to said control. In some embodiments, said reduced activity corresponds to no activity compared to said control.

Preferred biomarkers of homologous recombination are selected from RAD51 , BRCA2 and/or CDK1. Thus, in some embodiments, impaired homologous recombination may be characterised by reduced activity of RAD51 , BRCA2 and/or CDK1 , compared to a control. In some embodiments, said control is tissue sample in which the subject biomarker is absent. In some embodiments, the control is cells in which the subject biomarker is absent. The present inventors have surprisingly found that the above-recited proteins influence the ability of IGF-1 R to activate cell signalling, and this property tracks with the ability to regulate sensitivity to IGF-1 R inhibition. Thus, patients being considered for treatment with IGF-1 R (kinase) inhibitors or antibody drugs may have tumour tissue tested for the expression or activity of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C, BRCA2, CDK1 and/or RAD51. Preferred biomarkers may be selected from the list consisting of: CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C and/or RAD51. Preferred biomarkers may be selected from the list consisting of: DVL3, HUNK, LMTK3, CDKN2C, CNKSR1 , DUSP5 and/or RAD51. The results can be used to select patients whose tumours have low biomarker expression or activity, and which are therefore more likely to respond to IGF-1 R inhibition.

Preferably, the tumour sample is not treated ex vivo with a modulator of IGF-1 R activity prior to determining the level or activity of expression of biomarker(s) in said sample. Thus, in some embodiments, the tumour sample is not treated with a modulator of IGF- 1 R activity prior to determining the level of expression or activity of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C, BRCA2, CDK1 and/or RAD51 in said sample. In some embodiments, the tumour sample is not treated with a modulator of IGF-1 R activity prior to determining the level of expression or activity of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C and/or RAD51. In some embodiments, the tumour sample is not treated with a modulator of IGF-1 R activity prior to determining the level of expression or activity of CDKN2C, CNKSR1 , HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C, BRCA2, CDK1 and/or RAD51 in said sample. In some embodiments, the tumour sample is not treated with a modulator of IGF-1 R activity prior to determining the level of expression or activity of CDKN2C, CNKSR1 , HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C and/or RAD51. Typically, the patient has not received prior treatment with a modulator of IGF-1 R prior to sampling of said tumour. In embodiments wherein said patient has received prior treatment with a modulator of IGF- 1 R, said prior treatment is preferably at least 24 hours prior to said sampling e.g. at least 2 days; at least 3 days; at least 1 week; at least 2 weeks; at least 1 month; at least 3 months, prior to sampling of said tumour.

In a related aspect, the present invention provides a method of treating a subject having tumour, the method comprising administering to the subject:

(i) A drug which blocks expression or function of one or more of the identified resistance mediators CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C, BRCA2, CDK1 and/or RAD51 ; and

(ii) An IGF-1 R inhibitor.

The invention also provides a method of treating a patient having tumour, the method comprising administering to the patient:

(i) A drug which blocks activity of one or more of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C, BRCA2, CDK1 and/or RAD51 ; and

(ii) An IGF-1 R inhibitor. Preferred mediators may be selected from the list consisting of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C and/or RAD51. Preferred mediators may be selected from the list consisting of DVL3, HUNK, LMTK3, CDKN2C, CNKSR1 , DUSP5 and/or RAD51

This approach may be appropriate for patients whose tumours are resistant to IGF-1 R inhibition by virtue of the fact that their cancer over-expresses one of the identified resistance mediators.

In some embodiments of the invention, the treatment comprises concurrent administration of an inhibitor of homologous recombination. In some embodiments, the inhibitor of homologous recombination is suitable for depletion of the level of one or more of RAD51 , BRCA2 and CDK1. In some embodiments, the inhibitor of homologous recombination is suitable for blocking the activity of one or more of RAD51 , BRCA2 and CDK1. In some embodiments blocking the activity of said one or more of RAD51 , BRCA2 and CDK1 comprises reducing the activity of said one or more of RAD51 , BRCA2 and CDK1 , to an activity of 90% or less, compared to an untreated control, e.g. 90% or less, 80% or less, 70% or less, 60% or less, 50% or less, 40% or less, 30% or less, 20% or less, 10% or less, 5% or less or 1 % or less compared to said untreated control. In some embodiments, said reduced activity corresponds to no activity compared to said control In some embodiments of the invention, the patient has received prior treatment with an inhibitor of homologous recombination. Inhibitors of homologous recombination are known in the art and include, for example, B02 (Huang et al J Med Chem 2012 55:301 1- 3020, incorporated herein by reference). In some embodiments, the method of treating a subject does not comprise administering to the subject a drug which blocks expression or activity of DUSP5.

The tumour or tumour sample may be any one of the cancers listed above that expresses IGF-1 R; and may also express a relatively high level of the resistance mediator(s). One preferable embodiment of the method relates to head and neck squamous cell cancers (HNSCC) and blocking the expression of DVL3. One preferable embodiment of the method relates to head and neck squamous cell cancers (HNSCC) and blocking the activity of DVL3. The IGF-1 R inhibitor may be a small molecule IGF-1 R kinase inhibitor. The IGF-1 R inhibitor may be an IGF-1 R blocking antibody. The IGF-1 R inhibitor may be an IGF ligand antibody.

The inhibitors of the resistance mediators may be any small molecule chemical inhibitors designed to block the proteins listed below. Preferred inhibitors of DVL3 include DVL- PDZ inhibitor II (DVLi), which was developed by Grandy D, et al. J Biol Chem 2009; 284: 16256-63 (incorporated herein by reference). Preferred inhibitors of RAD51 include BO2, which was developed by Huang et al (ACS Chem Biol 6: 628-35, 2011 ; incorporated herein by reference).

The present inventors have shown that the presence of these proteins is associated with resistance to IGF-1 R inhibition. Accordingly, blocking their expression will allow use of IGF-1 R inhibitors in treatment. Similarly, blocking their activity will also allow use of IGF- 1 R inhibitors in treatment

Thus, the subject to be treated with the above-recited combination of drugs may be a subject who has been previously identified as unlikely to benefit from treatment with IGF- 1 R alone, whether by means of the above-described method or by other means (e.g. a previous, unsuccessful, clinical trial).

The present inventors have identified WNT component dishevelled homolog 3 (DVL3) - among others - as a mediator of resistance to IGF-1 R inhibition. Without being bound by any theory, it is thought that DVL3 has a function in regulating signal transduction from IGF-1 R to RAS. Inhibition of proximal WNT signalling has been found to recapitulate DVL3 depletion in activating RAS-MEK-ERK and enhancing sensitivity to IGF-1 R inhibition in vitro and in vivo. DVL3 protein levels are significantly lower in Ewing sarcomas than oropharyngeal cancers; these have previously been characterized respectively as responsive and refractory to IGF-1 R inhibition. The present inventors have also found inverse correlation between tumour DVL3 and progression-free survival in patients treated on IGF-1 R antibody trials. The present invention will now be described, by way of non-limiting example, with reference to the accompanying Figures: Figure 1 : DVL3 mediates resistance to IGF-1 R inhibition. 1A) Hits from triplicate second-round siRNA screens in DU145 cells, showing inhibition of cell viability expressed as mean ± SEM log2 cell surviving fraction, induced by single siRNAs (grey bars), siRNA pools (black), with non-silencing Allstars (AS) control. Dashed line: -0.2 threshold for significant growth inhibition in AZ12253801 -treated cells vs controls. 1 B) Isoform-specific qPCR showing expression of DVLs 1 -3 relative to GAPDH. 1 C) DU145 cells were transfected with 50nM Allstars (AS) siRNA or DVL3 siRNAs, 48hr later treated with AZ1253801 , and after 5 days assayed for viability. Graph: pooled data from 3 independent assays, curve-fitted to interpolate AZ12253801 Gl₅₀ values, shown in legend, with fold sensitization as ratio of Gl₅₀ values in Allstars and DVL siRNA transfectants. Inset: DVL3 western blot. 1 D) DU145 cells were transfected with DVL siRNAs, after 48hr the expression of DVL isoforms was assessed by qPCR, normalized to GAPDH and expressed relative to expression in Allstars (AS) transfectants. Table below shows AZ12253801 Gl₅₀ values in cells depleted of each DVL, and fold sensitization calculated as 1 B). 1 E) Cells were infected with lentiviral vectors encoding GFP or FLAG-DVL3, after 48hr transfected with Allstars (AS) siRNA or 3' UTR siRNA DVL3_6 and after 5 days assayed for viability. Legend shows AZ12253801 Gl₅₀ values from triplicate experiments, and fold sensitization as ratio to Gl₅₀ GFP-Allstars. Inset: parallel cultures were used for western blotting. 1 F) DU145 cells were transfected with control (Allstars), DVL3 and/or IGF-1 R siRNA (100nM total siRNA) and: left, assayed for cell viability after 5 days; right, used in western blot for DVL3 and IGF-1 R. Compared with control transfectants, depletion of DVL3 or IGF-1 R suppressed viability to 75 ± 9% and 72 ± 5% respectively, and combined depletion to 24 ± 1 1 %, suggesting supra- additive growth inhibition (**p<0.01 , ***p<0.001 by one way ANOVA.). Figure 2: DVL3 depletion enhances IGF signaling via MEK-ERK. 2A) DU145 cells were Allstars (AS) or DVL3 siRNA transfected and after 48hr treated with 10nM IGF-1 for 10min. 2B) Allstars or DVL3 siRNA-transfected cells were serum starved for 12hr and stained for total ERK 1/2, with DAPI nuclear staining. 2C) qPCR for ELK1 target genes in control (Allstars-transfected) and DVL3-depleted cells, mean ± SEM of triplicate independent analyses (*p<0.05, **p<0.01). 2D) Cells were siRNA-transfected as 2A), treated with 120 nM AZ12253801 for one hour and in the final 10min with 10nM IGF-1.

Figure 3: DVL inhibition recapitulates effects of DVL3 depletion and sensitizes to AZ12253801 in vitro and in vivo. 3A) Serum-starved DU145 cells were treated with DVLi for 16 hours. 3B) DU145 whole cell extract was treated with lambda phosphatase (λ) in the absence or presence of phosphatase inhibitors (PI). The upper band of DVL3 immunoreactivity was abolished by λ phosphatase and restored by PI, consistent with DVL phosphorylation, as reported (Gao C, Chen YG. Cell Signal 2010; 22: 717-27; incorporated herein by reference). 3C), 3D) DU145 cells were treated with 3C) AZ12253801 alone or with 100μΜ DVLi, and 3D) DVLi alone or with 120nM AZ12253801. Legends show Gl₅₀ values derived from pooled data in 3 independent experiments. 3E) Prostate cancer cells were serum-starved and treated with 100μΜ DVLi for 16hr. Upper, western blot for phospho-ERK; lower, parallel culture were treated with AZ12253801 alone or with 100μΜ DVLi. Table shows pooled data from at least 3 independent CTG assays, showing Gl₅₀ data for AZ1225380 alone (IGF-1 Ri) or with DVLi (Comb), fold sensitization calculated as Gl₅₀ ratio (Gl₅₀ IGF-1 Ri / Gl₅₀ Comb). Genotypes from www.sanger.ac.uk. 3F) Male mice bearing DU145 xenografts were treated for 14 days with intraperitoneal solvent (DMSO), 25mg/kg AZ12253801 twice daily, 50mg/kg DVLi once daily or the combination (DVU+AZ12253801 ). Tumour volumes were measured twice a week and expressed as % baseline. Tumour growth in groups treated with AZ12253801 or DVLi was not significantly different from controls. The combination treatment group showed significant tumor growth retardation compared with control (p<0.001), AZ12253801 alone (p<0.001 ) and DVLi alone (p<0.01 ) groups by repeated measures ANOVA.

Figure 4: DVL3 regulates signaling from IGF-1 R to RAS. 4A) Serum-starved DU145 cells were treated with 100μΜ DVLi for 16hr and in the final 0-60min with 10nM IGF-1. 4B) Serum-starved DU145 cells were treated with 100μΜ DVLi for 1 -16hr and in the final 30min with 10nM IGF-1. Graphs below 4A) and 4B) show phospho-ERK (mean ± range), corrected for total ERK, from two independent experiments. 4C) MCF7 cells were treated with 100μΜ DVLi for 16hr and in the final 30min with 10nM IGF-1. Graph below: parallel cultures were treated with AZ12253801 alone or with 100μΜ DVLi and cell viability was assayed after 5 days. Legend shows Gl₅₀ values derived from pooled data in 3 independent experiments, and fold sensitization (Gl₅₀ ratio) to AZ12253801 in DVL-inhibited cells. 4D) DU145 cells were treated with 100μΜ DVLi for 16hr, lysates were incubated with GST or GST-RAS binding domain (RBD) of RAF, and pull-downs analyzed for activated (RBD-bound) RAS. Blot to right confirms ERK activation in whole cell extracts. Graph below: RAS activity in two independent assays, expressed relative to control (DVLi-untreated) cells; *p<0.05. E) Upper: IGF-1 R activates RAS via a protein complex that includes She, Grb2 and SOS. Lower: Allstars (AS) transfectants or cells depleted of IGF-1 R, She, Grb2 or SOS were treated with 100μΜ DVLi for 16hr. DVLi- induced ERK activation was abolished by depletion of She, SOS or Grb2 but not IGF-1 R. 4F) DU145 whole cell extracts were immunoprecipitated with control (IgG) or DVL3 antibodies and analyzed by western blotting, in parallel with DVL3 IP supernatant (s/n) to confirm DVL3 immunodepletion. 4G) DU145 whole cell extract was incubated with GST or GST-Grb2, and precipitated proteins were analyzed by western blot. 4H) DU145 cells were transfected with Allstars (AS) or DAB2 siRNA and after 48hr: left, treated with AZ12253801 and cell viability assayed after 5 days, showing Gl₅₀ values from 3 independent assays; right, serum-starved overnight and treated with 10nM IGF-1 for 10min. 4I) Allstars-transfected or DAB2-depleted cells were treated with DVLi and/or AZ12253801 and cell viability assayed after 5 days. Graph: mean ± SEM viability, pooled data from 3 independent assays. Legend shows Gl₅₀ values and fold sensitization to AZ12253801. Blot to right: parallel cultures analyzed by western blot.

Figure 5: DVL3 protein expression is inversely correlated with response to IGF-1 R antibody. DVL IHC performed on 5A) Control cells and tissue. DU145 cells were transfected with Allstars or DVL3 siRNA, and after 48hr were formalin-fixed, paraffin- embedded and used as controls for the specificity of immunostaining, in parallel with sections of a TURP in which 100% of chippings were involved by Gleason grade 4+5 prostate cancer from a previously untreated patient. These cell and tissue controls were included in every staining run for quality control purposes. 5B) Tissue microarrays of Ewing sarcoma and HNSCC underwent DVL3 immunohistochemical staining in the same staining run. Representative images are shown of: upper: Ewing sarcoma; to the left is a sample of normal kidney included as a control on the same TMA slide; lower: HNSCC. Most of the Ewing sarcomas had light or no DVL3 signal; one of the tumors had scattered haemosiderin deposition. 5C) DVL3 Intensity x Percentage (IPS) scores for 9 cases of Ewing sarcoma and 24 cases of HNSCC, showing significantly higher mean score in HNSCC. 5D) Tumours from patients on IGF-1 R antibody trials. 5E) Graph of progression-free survival (PFS) vs DVL3 IPS in 22 patients treated on IGF-1 R antibody trials.

Figure 6: Model for DVL3 as a regulator of RAS activation and mediator of resistance to IGF-1 R inhibition. The WNT and IGF signaling pathways, showing DVL3 in complex with adaptor proteins downstream of IGF-1 R. This newly-identified location allows DVL3 to regulate signal transduction to SOS-RAS, and mediate resistance to IGF-1 R inhibition. Figure 7: IGF-1 R influences repair of endogenous DNA damage. 7A) Serum-starved DU145 cells were treated with AZ12253801 for 1 hr and in the final 15min with 50nM IGF-1. 7B) DU145 cells were treated with solvent (control) or 100nM AZ12253801 , fixed 0-3 days later and stained for γΗ2ΑΧ and DAPI. Irradiated cells (3Gy, 6hr) served as a positive control for γΗ2ΑΧ foci. 7C) Graph: mean ± SEM foci per cell from 2 independent experiments. Foci increased with time in IGF-1 R inhibited cells (*p<0.05, ***p<0.001 ). 7D) DU145 were transfected with AHStars siRNA (siControl) or silGFI R, and fixed and stained 1 -3 days later as b), with irradiated cells as a positive control. 7E) Cells transfected as D) were analysed as C). Graph: mean ± SEM foci per cell. Three days post-siRNA transfection, IGF-1 R depleted cells contained more γΗ2ΑΧ foci than controls (***p<0.001). 7F) DU145 cells were siRNA-transfected and lysed on days 1-3 for western blotting.

Figure 8: PTEN wild-type tumor cells are sensitized to IGF-1 R inhibition by depletion or loss of RAD51 or BRCA2. 8A) Survival of DU145 cells transfected with siControl, IGF- 1 R or RAD51 siRNAs. Graph: mean ± SEM colonies showing reduced survival on depletion of IGF-1 R (*p<0.05) or RAD51 (***p<0.001 ). To right: parallel cultures blotted for IGF-1 R and RAD51. 8B) DU145 cells were siRNA-transfected as a) and treated with solvent or AZ12253801. Graph: survival (% untreated) from two independent experiments (6 data points; ***p<0.001 by two-way ANOVA). Data were curve-fitted to interpolate AZ12253801 SF₅₀: 102nM for controls, and 108, 50 and 45 nM for IGF-1 R, RAD51_1 , RAD51_7 siRNA transfectants respectively. Graph to right: relative survival at 100nM AZ12253801 , showing sensitization by RAD51 -depletion (***p<0.001). 8C) DU145 cells were reverse-transfected with siControl, siRAD51_1 , or siRAD51_7 and viability assayed 5 days later. Graph: mean ± SEM luminescence (***p<0.001). 8D) DU145 cells were siRNA-transfected as c), 48hr later treated with AZ12253801 , and after 5 days viability assayed. Graph: mean ± SEM viability as % siControl, pooled data from 3 independent assays (***p<0.001 ), generating AZ12253801 Gl₅₀ values in controls 122nM, RAD51_1 transfectants 65nM, RAD51_7 transfectants 55nM. Graph to right: relative viability at 100nM AZ12253801 (***p<0.001 ). 8E) Cell lysates analyzed for components of IGF axis, androgen receptor (AR), RAD51. Table below: PTEN status, AZ12253801 Gl₅₀ values (3 experiments in each cell line as d), fold sensitization as Gl₅₀ Ratio in siControl/siRAD51 transfectants. 8F) DU145 cells were transfected with siControl or BRCA2 siRNAs (siBRCA2_6, siBRCA2_7, siBRCA2), and AZ12253801 - treated and viability assayed as d). Graph: mean ± SEM of 2 independent experiments. Only siBRCA2 sensitized to AZ12253801 (** p<0.01 ). Inset: parallel cultures analysed by western blotting. Graph to right: relative viability at 100nM AZ12253801 (**p<0.01 ). 8G) BRCA2^+/" and BRCA2^"'^" DLD-1 cells analysed by: left, western blotting; right, immunofluorescent staining for γΗ2ΑΧ (red), RAD51 , DAPI, 6hr after 5Gy irradiation. 8H) DLD-1 cells were treated with AZ12253801 and viability assayed as d). Graph: mean ± SEM viability, pooled data from 3 experiments. BRCA2^"'^" cells were more sensitive to AZ12253801 than BRCA2^+/" cells (** p<0.01 ). Graph to right: relative viability at 100 nM AZ12253801 (***p<0.001 ).

Figure 9: Effects of RAD51 depletion on sensitivity of prostate cancer cells to AZ12253801. 9A) Prostate cancer cells were transfected with siRAD51 _1 or _7 and after 48hr lysed and analysed by western blot to check RAD51 depletion. 9B) PC3, 9C) LNCaP and 9D) 22Rv1 cells were siRNA-transfected, AZ12253801 -treated and assayed for viability as Figure 2D (*p<0.05, **p<0.01 , ***p<0.001 by 2-way ANOVA). Data from triplicate independent experiments for each cell line were pooled and curve-fitted to interpolate Gl₅₀ values, summarized in Figure 2E. Graphs to right: relative viability at 50nM AZ12253801 (*p<0.05, **p<0.01 , ***p<0.001 by one-way ANOVA). 9E) siControl or siRAD51_7 -transfected DU145 cells were serum-starved overnight and treated with 50nM IGF-1 for 15min. Figure 10: Prostate cancer cells are sensitized to IGF-1 R inhibition by inhibitors of RAD51 or CDK1 . 10A) DU145 cells were treated with solvent (Control), B02 (10 // M), RL-1 (10 /i M), or RO-3306 (1 μ M), after 1 hr irradiated (5Gy), and after 6hr stained for y H2AX, RAD51 , DAPI . 10B) DU145 cells were treated with AZ12253801 with solvent or 10 /i M B02 or RL-1 , and viability assessed 5 days later. Graph: mean ± SEM of 3 independent experiments. Sensitivity to AZ12253801 was enhanced by B02 (*** p<0.001 by 2-way ANOVA) but not RL-1. Graph to right: relative viability at 100nM AZ12253801 (*** p<0.001 ). 10C) DU145 cells were treated for 1 -5 days with 100nM AZ12253801 (AZ3801 ) and/or 1 μΜ RO-3306 and analyzed by flow cytometry. 10D) DU145 cells were treated with AZ12253801 alone or with 1 μ M RO-3306 and viability assayed after 5 days. Graph: mean ± SEM viability from 3 independent experiments. Sensitivity to AZ12253801 was enhanced by RO-3306 (p<0.001 by two-way ANOVA). Legend shows Gl₅₀, GI₈o and fold sensitization (ratio of Gl₅₀ or GI₈o values). Graph to right: relative viability at 100nM AZ12253801 (***p<0.001 ). Figure 11 : Characterizing effects of homologous recombination inhibitors on prostate cancer cells. Cells were treated with 11 A) B02, 11 B) RL-1 , 11 C) RO-3306 and cell viability was quantified after 5 days. 11 D) Pooled data from 2 experiments (6 data points) were curve-fitted to determine Gl₅₀ values for each compound in each cell line. 11 E) Cells were treated with solvent (control) or compounds at the Gl₅₀, and 1 hr later were irradiated (5Gy) to determine effects on induction of RAD51 foci as Figure 3a.

Figure 12: Effects of inhibiting RAD51 or CDK1 on response of prostate cancer cells to AZ12253801. 12A) PC3, 12B) LNCaP, 12C) 22Rv1 cells were treated with AZ12253801 together with solvent or B02 at or near the pre-determined Gl₅₀ value (LNCaP 25μΜ, PC 4μΜ, 22Rv1 5μΜ; see Figure S2d), and viability was assessed 5 days later. Graphs represent pooled data (mean ± SEM) from 3 independent experiments in each cell line. Legends show Gl₅₀ values and statistical significance of differences between controls and B02-treated cells. B02 sensitized to AZ12253801 in PC3 and 22Rv1 (*** p<0.001 by 2-way ANOVA) but not LNCaP cells. 12D) PC3, 12E) LNCaP, 12F) 22Rv1 cells were treated with AZ12253801 together with solvent or RO-3306 at the Gl₅₀ for each cell line, and viability was assessed 5 days later. Graphs represent pooled data (mean ± SEM) from two independent experiments. RO-3306 had no effect on sensitivity to AZ12253801 in PC3 or LNCaP cells. In 22Rv1 cells there was a shift in the AZ12253801 dose-response curve (p<0.001 by 2-way ANOVA); fold-differences in Gl₅₀ and GI₈o values are shown in legend.

Figure 13: Effect of inhibiting HUNK on MCF-7 cells treated with IGF-1 R inhibitors 13A) OSI-906, 13B) BMS 754807 (BMS807) and 13C) N VP-AD W742, and 13D) also to monoclonal blocking antibody figitumumab. IGF-1 R over-expression has previously been found in a number of cancers, but is a poor positive predictor of clinical sensitivity to IGF-1 R inhibition (Yee D. J Natl Cancer Inst 2012; 104: 975-81 ; incorporated herein by reference). Consistent with this, IGF-1 R over-expression in low IGF-1 R PC3 prostate cancer cells did not influence the response to IGF-1 R tyrosine kinase inhibitor (TKI) AZ12253801 (AstraZeneca, described in Aleksic T, et al. Cancer Res 2010; 70: 6412-9; incorporated herein by reference). In order to identify proteins that influence sensitivity or resistance to AZ12253801 , siRNA screens were performed using DU145 prostate cancer and MCF7 breast cancer cells (materials and methods are described in detail below).

Initial experiments confirmed that AZ12253801 inhibited IGF-1 R phosphorylation and cell viability. Primary siRNA screens were then performed to deplete -1000 targets; given the involvement of IGF-1 R kinase in the DNA damage response (Turney BW, et al. Radiother Oncol 2012; 103: 402-9 ; and Chitnis MM, et al. Oncogene 2013; in press; both of which are incorporated herein by reference), siRNA libraries were selected which target kinase-related and DNA repair-associated proteins, together with positive (siPLKI ) and negative non-silencing (Allstars) control siRNA (as described in Lord CJ, et al. DNA Repair (Amst) 2008; 7: 2010-9 and lorns E, et al. Cancer Cell 2008; 13: 91 - 104; both of which are incorporated herein by reference). After 48hr to allow target protein depletion, cells were treated with solvent or AZ12253801 at the pre-determined Gl₅₀ and cell viability was assayed 5 days later. Duplicate DU145 screens were highly reproducible (R2 values >0.8) and sensitive, with Z'-factors of 0.23-0.6, indicating good discrimination between positive and negative controls. To identify potential mediators of sensitivity or resistance to AZ12253801 , drug sensitization Z-scores (Boutros M, Ahringer J. Nat Rev Genet 2008; 9: 554-66; incorporated herein by reference) were calculated for each siRNA. Ordered siRNAs were ranked according to Z-score, and 54 genes were selected for further validation (Table 1 ).

Gene Z- Gene Z-score

score

DUSP5 -5.395 S0X4 -4.264

CDKN2C -2.811 MGMT -3.246

KIT -2.742 INO80C -3.068

AKAP1 -2.658 RAD23B -2.525

MAP3K10 -2.453 XRCC6 -2.501

MPP2 -2.422 XRCC2 -2.465

FLT4 -2.396 MPG -2.423 LMTK3 -2.389 SND1 -2.421

APEG1 -2.32 MCM3 -2.419

MAP4K3 -2.173 LDHA -2.343

TIE -2.172 CCT5 -2.229

LIM -2.047 CDKN3 -2.146

PDK4 -2.041 DVL3 -2.143

PIK3C2A -1.984 TREX2 -2.120

MYLK -1.943 Rad51 -1.741

CNKSR1 -1.941 RAD54B -1.733

DUSP4 -1.889 BRCA1 -1.635

SGK2 -1.851 KIAA0101 -1.604

CCL4 -1.744 CKN1 -1.554

STK39 -1.703 DCLRE1A -1.534

HUNK -1.605 ADPRT -1.500

CDK6 -1.421 LIG4 -1.063

LIMK2 -0.942 UBE2V2 -0.546

PLK3 -0.918

CDK4 -0.828

LIMK1 -0.803

CDK3 -0.388

CDC2 2.385

PANK4 2.535

PRKAG3 2.894

Table 1 : Candidate resistance mediators tested in second round screens. Table shows averaged Z-scores from duplicate primary screens in DU145 cells of: left, kinome library; right, DNA repair library. Potential hits were selected as those with averaged Z-score <-1.8, or proteins predicted to interact with higher score hits. The final three hits listed from the kinome screen had averaged Z-scores >2.0, suggesting that hit depletion mediated resistance to AZ12253801 , but none of these could be validated in second round screens.

Triplicate second round screens identified 12 putative resistance mediators, including regulators of the cell cycle and DNA damage response, and proteins with poorly- characterized functions (Figure 1A). Each was validated in a low-throughput format by confirming that depletion by screen siRNAs enhanced sensitivity to AZ12253801 (Figure 1 B, Table 2).

Table 2: Validation of screen hits. DU145 cells were reverse transfected with 50nM Allstars or gene specific siRNAs, after 48hr treated with solvent (0.01 % DMSO) or AZ12253801 , and viability was assayed after 5 days. Each target was tested in three independent experiments, data were pooled and curve-fitted to interpolate Gl₅₀ values. Fold sensitization to AZ12253801 is shown as a ratio of Gl₅₀ in cells transfected with Allstars and gene-specific siRNA.

Six hits in DU145 screens (CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3) were also identified as candidate resistance mediators in MCF7 cells. RAD51 was a clear screen hit in DU145 cells and is thus also considered to be useful in the methods of the present invention.

One of the most robust hits common to both cell lines was Dishevelled homolog 3 (DVL3), a poorly characterized member of the WNT pathway. DVL3 is one of 3 mammalian homologs of Drosophila Dsh, a cytoplasmic protein that is phosphorylated on binding of Wnt ligands to Frizzled (Fz) receptors, blocking activity of the β-catenin destruction complex (Gao C, Chen YG. Cell Signal 2010; 22: 717-27; incorporated herein by reference). Assessment of endogenous DVL expression indicated that DVL3 was the dominant isoform in both DU145 and MCF7 cells (Figure 1 B), and depletion of DVL3 but not DVL 1 or 2 sensitized to AZ12253801 (Figure 1C, D). Expression of siRNA-resistant FLAG-DVL3 was able to rescue from AZ12253801 sensitization induced by DVL3 siRNA targeting the 3'-untranslated region of endogenous DVL3 mRNA (Figure 1 E), suggesting that sensitization by DVL3 depletion was unlikely to be an off-target effect of DVL3 siRNA. Furthermore, DVL3-depleted cells were sensitized to IGF-1 R depletion (Figure 1 F), supporting the contention that functional interaction between DVL3 and AZ12253801 is related to the ability of AZ12253801 to block IGF-1 R.

Without being bound by any theory, the present inventors suggest that the ability of DVL3 to regulate AZ12253801 sensitivity may be because DVL3 has a (WNT- independent role) in regulating RAS activation. The adaptor protein IRS-1 is known to undergo IGF-induced interaction with β-catenin, promoting β-catenin stabilization, nuclear translocation and transcriptional activity (Playford MP, et al. Proc Natl Acad Sci U S A 2000; 97: 12103-8; and Chen J, et al. J Biol Chem 2005; 280: 29912-20; both of which are incorporated herein by reference). However, the present inventors have found no evidence that IGF-1 affected levels of active β-catenin or phosphorylated DVL3 in DU145 prostate cancer cells (Figure 2A). Consistent with known roles for DVL3 in canonical WNT signalling and mTOR activation (Gao C, Chen YG. Cell Signal 2010; 22: 717-27; and Inoki K, et al. Cell 2006; 126: 955-68; both of which are incorporated herein by reference), DVL3 depletion suppressed active β-catenin and phosphorylation of mTOR effector S6. DVL3-depleted cells showed no change in IGF-1 R expression (Figure 2A), but unexpectedly exhibited enhanced basal and IGF-stimulated ERK phosphorylation (Figure 2A), accompanied by ERK nuclear translocation and up- regulation of ERK-ELK target genes (Figure 2B, 2C). The IGF-induced component of ERK activation in DVL3-depleted cells was effectively suppressed by AZ12253801 (Figure 2D). These data validate DVL3 as a mediator of resistance to IGF-1 R inhibition, and also suggest that DVL3 regulates signal transduction from IGF-1 R to ERKs. The present inventors have identified two further proteins with newly-identified roles as kinase regulators: lemur tyrosine kinase 3 (LMTK3) is a serine/threonine kinase reported to regulate AKT (Giamas G, et al. Nat Med 201 1 ; 17: 715-9; incorporated herein by reference), and hormonally upregulated neu-associated kinase (HUNK), a SNF1/AMPK- related serine-threonine kinase found to influence EGFR activation (Komurov K, et al. J Biol Chem 2010; 285: 21134-42; incorporated herein by reference). The present inventors have additionally found that HUNK depletion enhanced IGF signaling via AKT, suggesting a common theme in that sensitivity to IGF-1 R inhibition is regulated by factors that influence IGF signaling downstream of the receptor.

Compounds were tested that block WNT signalling at different levels, with the aim of finding a drug that recapitulates DVL3 depletion. Consistent with a role in Axin stabilization (Huang SM, et al. Nature 2009; 461 : 614-20; incorporated herein by reference), the toolbox tankyrase inhibitor XAV939 upregulated Axinl and also inhibited mTOR, but did not activate ERKs or sensitise to AZ12253801. The effects of WNT inhibition were subsequently tested at a more proximal step, using DVL-PDZ inhibitor II (DVLi) that blocks binding of Fz to DVL (Grandy D, et al. J Biol Chem 2009; 284: 16256- 63; incorporated herein by reference). This agent inhibited DVL3 phosphorylation and phenocopied DVL3 depletion, reducing active β-catenin and phospho-S6, activating ERKs and sensitizing to AZ12253801 (Figure 3A-C). There was evidence for reciprocal sensitisation: despite suppressing active β-catenin and mTOR, DVLi alone caused negligible growth inhibition, perhaps related at least in part to its ability to activate ERKs, but with AZ12253801 caused significant loss of viability (Figure 3D). The DVLi also sensitized to three additional IGF-1 R inhibitory agents including two that have been evaluated clinically: the IGF-1 R blocking antibody figitumumab, and TKI BMS-754807 (Olmos D, et al. Lancet Oncol 2010; 1 1 : 129-35; and Haluska P, et al. J Clin Oncol 201 1 ; 29: (suppl; abstr TPS1 1 1); both of which are incorporated herein by reference). These data suggest that DVL inhibition influences responses to two different classes of IGF-1 R inhibitor.

While the DVLi used herein has relatively low potency, the WNT pathway is an intense focus for drug development (Zimmerman ZF, et al. Cold Spring Harbor Perspectives in Biology 2012; 4; incorporated herein by reference), offering the prospect of more potent inhibitors in future. The DVLi used is sufficient to perform proof of principle experiments. In this regard, its ability to influence sensitivity to IGF-1 R inhibition was tested in a panel of 5 prostate and 6 breast cancer cell lines in which expression of IGF axis components was characterised. Of 1 1 cell lines, three (LNCaP, LNCaP-LN3 and BT474) were resistant to AZ12253801 with no sensitization upon DVLi treatment (Figure 3E). BT474 are known to be refractory to IGF-1 R inhibitors (Chakraborty AK, et al. Breast Cancer Res Treat 2010; 120: 327-35; incorporated herein by reference) and had high basal phospho-ERK, which is correlated in small cell lung cancer cell lines with resistance to IGF-1 R inhibition (Zinn RL, et al. Mol Cancer Ther 2013; 12: 1 131 -9; incorporated herein by reference), but this association was not manifest in LNCaP or LNCaP-LN3. The remaining cell lines showed variable AZ12253801 sensitivity, which was enhanced by DVLi in those cell lines in which DVLi also activated ERKs (Figure 3E). These comprised 2 of 5 prostate cancer cell lines and 5 of 6 breast cancer cell lines, including 4 of 4 triple negative cell breast cancer (TNBC) lines. Previous work highlighted the intrinsic sensitivity of TNBC to IGF-1 R inhibition (Litzenburger BC, et al. Clin Cancer Res 201 1 ; 17: 2314-27; incorporated herein by reference); the present inventors propose that further sensitization can be achieved by proximal WNT inhibition. This is consistent with reported WNT pathway activation in basal-like breast cancers, many of which have a TNBC phenotype (Khramtsov Al, Am J Pathol 2010; 176: 2911 -20; incorporated herein by reference). Supporting the existence of functional cross-talk between the IGF axis and proximal WNT components, up-regulation of IGF binding protein 5 was recently shown to mediate growth inhibitory effects of a soluble Wnt inhibitor in murine MMTV- Wnt1 driven tumors (Liu BY, et al. Cancer Res 2012; 72: 1568-78; incorporated herein by reference). Therefore, to test the potential clinical relevance of these findings, mice bearing DU145 prostate cancer xenografts were treated with AZ12253801 and DVLi alone or in combination. AZ12253801 or DVLi alone had no significant effects on the growth of DU145 xenografts, but tumor growth in the combination treatment group was significantly retarded compared with control-treated (p<0.001), AZ12253801 -treated (p<0.001 ) and DVLi-treated animals (p<0.01 ; Figure 3F).

These data indicate that inhibition of proximal WNT signalling enhances sensitivity to IGF-1 R inhibition, and suggest that this property tracks with regulation of IGF signalling to MEK-ERK (Figure 2A-C, 3A-C, 3E). To further characterize this effect, time-course experiments were performed in control or DVL-inhibited DU145 cells. IGF-1 induced rapid activation of IGF-1 R and AKT that persisted in control cells for at least 60min, while ERK activation peaked at 10min and resolved to basal levels by 60min (Figure 4A). In contrast there was clear persistence of IGF-induced ERK phosphorylation at 30-60min in cells where DVL3 was inhibited (Figure 4A) or depleted. Persistent ERK activation in DVL-inhibited cells was apparent within 4hr of DVLi treatment, and was strikingly enhanced upon IGF-treatment (Figure 4B). An increase in IGF-induced ERK activation was also observed in MCF7 cells (Figure 4C). In contrast, DVL inhibition did not influence the response to IGF-1 in PC3 prostate cancer cells that were not sensitized to AZ12253801 by DVLi. These results support the existence of a link between the ability of DVL3 to influence sensitivity to IGF-1 R inhibition, and to attenuate the response of the IGF axis to a mitogenic stimulus via MEK-ERK. Consequently, DVL3 depletion or inhibition creates an environment that is permissive for signalling, recently characterized as 'signalability' (Lito P, et al. Cancer Cell 2012; 22: 668-82; incorporated herein by reference). To explore the mechanism of this effect, RAS activation assays were performed, which indicated that RAS was activated in DVL-inhibited cells (Figure 4D). This contrasts with previously reported WNT:ERK cross-talk, which occurs at more distal signalling nodes, and generates positive feedback between the two pathways (Kim D, et al. Oncogene 2007; 26: 4571 -9; Schlange T, et al. Breast Cancer Research : BCR 2007; 9: R63; Shin SY, et al. Cancer Res 2010; 70: 6715-24; all of which are incorporated herein by reference). Given that DVL3 depletion did not activate IGF-1 R itself (Figure 2A), and DVL-inhibited cells showed no change in EGF-induced ERK activation or sensitivity to EGFR inhibitor gefitinib, it is thought that RAS activation in DVL depleted or inhibited cells was unlikely to be initiated at the level of these RTKs.

It is increasingly recognized that mitogenic signals generated by RTKs are integrated by complexes of adaptor and scaffolding proteins, including IRS-1 , a well-recognized focus for feedback signaling via ERKs and mTOR-S6 kinase (Yuen JS, et al. Mol Cancer Ther 2009; 8: 1448-59 and Buck E, et al. Cancer Res 2008; 68: 8322-32; both of which are incorporated herein by reference). However, IRS-1 knockdown did not influence AZ12253801 sensitivity or ERK activation induced by DVL inhibition. In contrast, ERK activation in DVL-inhibited cells was clearly suppressed by depletion of the adaptor protein She, the exchange factor son-of-sevenless (SOS), or growth factor receptor- bound-2 (Grb2), or although not IGF-1 R itself (Figure 4E). These data suggest that DVL3 suppresses signal transduction downstream of IGF-1 R, at the level of the Shc- Grb2-SOS complex. DVLs were not identified as Grb2 interactors in HEK293 cells (Bisson N, et al. Nat Biotechnol 201 1 ; 29: 653-8; incorporated herein by reference), but DVL3 does contain the atypical proline-rich region shown in DVL2 to bind Grb2 and to promote canonical Wnt signaling (Crampton SP, et al. PLoS One 2009; 4: e7841 ; incorporated herein by reference). Indeed, complexes containing Grb2 and DVL3 were detectable in DU145 cells by immunoprecipitation and Grb2 pulldown (Figure 4F, G). These complexes also contained the putative tumor suppressor Disabled 2 (DAB2), reported to limit RAS activation by competing with SOS for Grb2 binding (Xu XX, et al. Oncogene 1998; 16: 1561-9; Zhou J, Hsieh JT. J Biol Chem 2001 ; 276: 27793-8; and Hocevar BA, et al. Embo J 2003; 22: 3084-94; all of which are incorporated herein by reference). Consistent with this role, and with the association we found earlier between ERK activation and sensitization to IGF-1 R inhibition (Figure 2A-C, 3A-C, 3E), DAB2 depletion mimicked DVL3 depletion in sensitizing to AZ12253801 and enhancing IGF- induced ERK activation (Figure 4H). It was found that DVL3 and DAB2 showed similar binding patterns to individual Grb2 domains, interacting principally with the amino- terminal SH3 domain and central SH2 domain, while as reported, SOS bound to Grb2 amino and carboxy-terminal SH3 domains (Egan SE, et al. Nature 1993; 363: 45-51 ; incorporated herein by reference). It was investigated whether DVL3 might interact with Grb2 via DAB2, but it was instead found that DAB2-depleted cells still contained DVL3:Grb2 complexes, and could be further sensitized to IGF-1 R inhibition by DVLi, with 18-fold reduction in AZ12253801 Gl₅₀ values in DAB2-deleted, DVL inhibited cells, compared with 2.8 and 4.6 fold sensitization induced separately by DAB2-depletion or DVL-inhibition (Figure 4I). These data identify DVL3 as a major regulator of IGF signalling to RAS, and suggest that DVL3 has a different role from DAB2 in the Shc- Grb2-SOS complex.

To investigate the relevance of these findings to cancer patients, DVL3 expression was evaluated in clinical tumors. Using controls including DVL3-depleted DU145 cells, a specific immunohistochemistry protocol was developed, in which DVL3 in breast and prostate cancers was scored by intensity and extent of staining (Figure 5A). Approximately 50% of breast and prostate cancers had moderate or heavy cytoplasmic DVL3 protein expression, with no correlation with stage, grade or patient survival (Table 3).

Table 3: DVL3 expression and clinical parameters in breast and prostate cancer. Tables show analyses for a) Prostate cancer, 101 cases (38 deaths); b) Breast cancer, 97 cases (22 deaths) scored for DVL3 by Intensity x Percentage Score (IPS). Upper tables show analysis of DVL3 IPS by clinical parameters using Chi-squared tests and *Fisher's exact test, showing association with EGFR in breast cancer but no other significant associations. Lower tables: multivariate analysis for survival, showing the expected correlation with known prognostic variables, but not with DVL3 IPS. PSA, prostate specific antigen; ER, estrogen receptor; RFS, relapse-free survival; OS, overall survival.

As a first approach to testing for correlation between DVL3 protein expression and sensitivity to IGF-1 R inhibition, DVL3 was assessed in cohorts of patients with Ewing sarcoma, reported to be responsive to IGF-1 R inhibitor monotherapy (Olmos D, et al. Lancet Oncol 2010; 1 1 : 129-35; and Tolcher AW, et al. J Clin Oncol 2009; 27: 5800-7; both of which are incorporated herein by reference), and head and neck squamous cell cancer (HNSCC) that appears to be resistant (Schmitz S, et al. Ann Oncol 2012; 23: 2153-61 ; incorporated herein by reference). The results are shown in Figure 5B and 5C; all but one of the Ewing sarcomas contained low or no detectable DVL3, and the mean DVL3 staining score was significantly higher in the HNSCC cohort (p=0.0033). DVL3 expression was then quantified in archival tumor tissue from patients recruited to early phase clinical trials of IGF-1 R antibodies figitumumab or AVE164 (Macaulay VM, et al. Ann Oncol 2013; 24: 784-91 ; and Schmitz S, et al. Ann Oncol 2012; 23: 2153-61 ; both of which are incorporated herein by reference). Figure 5D shows examples of DVL3 staining in 8 of the trial cases, and analysis of the 22 case cohort with respect to progression-free survival (PFS) is shown in Figure 5E. While there was overlap in DVL3 expression between patients experiencing early progression vs prolonged control, it was notable that of 8 patients achieving prolonged disease control (>84 days, including one partial remission), the tumors of 6 of these patients had low DVL3 expression (IPS <5; Table 4).

HNSCC 6 44

Ovarian 12 44

Ovarian 6 42

Melanoma 4 42

HNSCC 12 38

Colorectal 4 36

HNSCC 12 31

Table 4: DVL3 IPS on tumor tissue from 22 patients recruited to trials of IGF-1 R antibodies. HNSCC, head and neck squamous cell cancer. DVL3 IPS was derived as the product of intensity and percentage of DVL3 immunohistochemical staining as described in Methods. PFS, progression-free survival.

In this series of 22 cases, DVL3 IPS showed a negative correlation with PFS (Spearman r = -0.47, p = 0.028), and mean PFS was longer in patients whose tumours showed no/low DVL3 (n=10, PFS 127 ± 27 days) compared with patients whose tumours had moderate or strong DVL3 immunostaining (n=12, PFS 68 ± 12 days, p = 0.044; Figure 5E). Included in this series were melanomas, ovarian gastrointestinal and head and neck cancers (Table 4), suggesting that DVL3 expression may have predictive value for response to IGF-1 R inhibition in a range of tumour types.

The present inventors have accordingly demonstrated that the proportion of potentially- responsive patients in IGF-1 R trials could be significantly increased by selecting cases with low-DVL3 tumours.

As mentioned above, RAD51 was a clear screen hit in DU145 cells. RAD51 catalyses the strand-invasion step of homologous recombination (HR) repair of DNA damage. IGF-1 R targeting has been reported to sensitize tumor cells to ionizing radiation and cytotoxic drugs, and delay double-stranded break (DSB) repair by non-homologous end joining (NHEJ) and homologous recombination (Rochester MA, Riedemann J, Hellawell GO, Brewster SF, Macaulay VM. Silencing of the IGF1 R gene enhances sensitivity to DNA-damaging agents in both PTEN wild-type and mutant human prostate cancer. Cancer Gene Ther 2005; 12: 90-100; Ferte C, Loriot Y, Clemenson C, et al. IGF-1 R targeting increases the antitumor effects of DNA damaging agents in SCLC model: an opportunity to increase the efficacy of standard therapy. Mol Cancer Ther 2013; Turney BW, Kerr M, Chitnis MM, et al. Depletion of the type 1 IGF receptor delays repair of radiation-induced DNA double strand breaks. Radiother Oncol 2012; 103: 402-9; Chitnis MM, Lodhia KA, Aleksic T, Gao S, Protheroe AS, Macaulay VM. IGF-1 R inhibition enhances radiosensitivity and delays double-strand break repair by both non- homologous end-joining and homologous recombination. Oncogene 2013; Chitnis MM, Lodhia KA, Aleksic T, Gao S, Protheroe AS, Macaulay VM. IGF-1 R inhibition enhances radiosensitivity and delays double-strand break repair by both non-homologous end- joining and homologous recombination. Oncogene 2013; all of which are incorporated herein by reference). DSBs also arise from endogenous damage, typically following collapse of stalled replication forks, and depend on homologous recombination for repair due to their one-ended structure (Helleday T. Pathways for mitotic homologous recombination in mammalian cells. Mutation research 2003; 532: 103-15, incorporated herein by reference). In view of these findings and our identification of RAD51 as a candidate mediator of resistance to IGF-1 R inhibition, the inventors propose that IGF-1 R inhibition leads to accumulation of DSBs that form at endogenous DNA lesions.

In DU145 cells, AZ12253801 inhibited phosphorylation of IGF-1 R and its downstream effectors (Figure 7A), confirming previous results (Chitnis MM, Lodhia KA, Aleksic T, Gao S, Protheroe AS, Macaulay VM. IGF-1 R inhibition enhances radiosensitivity and delays double-strand break repair by both non-homologous end-joining and homologous recombination. Oncogene 2013, incorporated herein by reference). Using γΗ2ΑΧ as a DSB marker, the inventors assessed whether IGF-1 R inhibition influences accumulation of endogenous damage. Compared with controls, AZ12253801 -treated cells showed progressive accumulation of γΗ2ΑΧ foci over a 3-day time-course (Figure 7B,C). To assess the specificity of this effect, the inventors repeated the experiment using siRNA to deplete IGF-1 R. There was no difference 1 -2 days after siRNA transfection, but by 3 days IGF-1 R depleted cells contained more γΗ2ΑΧ foci than controls (Figure 7D,E). Without wishing to be bound by theory, the relative delay compared with effects of AZ12253801 may be because IGF-1 R depletion was achieved only after 2-3 days (Figure 7F), consistent with the relatively long half-life (~16-20hr) of IGF-1 R protein (Yuen JS, Cockman ME, Sullivan M, et al. The VHL tumor suppressor inhibits expression of the IGF1 R and its loss induces IGF1 R upregulation in human clear cell renal carcinoma. Oncogene 2007; 26: 6499-508, incorporated herein by reference).

While γΗ2ΑΧ foci can form at single-stranded DNA regions, their detection correlates closely with numbers of DSBs (Lobrich M, Shibata A, Beucher A, et al. gammaH2AX foci analysis for monitoring DNA double-strand break repair: strengths, limitations and optimization. Cell cycle 2010; 9: 662-9). Therefore, these results suggest that IGF-1 R inhibition or depletion enhances accumulation of DSBs induced by endogenous DNA damage, consistent with previous reports that IGF-1 R influences homologous recombination (Turney BW, Kerr M, Chitnis MM, et al. Depletion of the type 1 IGF receptor delays repair of radiation-induced DNA double strand breaks. Radiother Oncol 2012; 103: 402-9; Chitnis MM, Lodhia KA, Aleksic T, Gao S, Protheroe AS, Macaulay VM. IGF-1 R inhibition enhances radiosensitivity and delays double-strand break repair by both non-homologous end-joining and homologous recombination. Oncogene 2013, both of which are incorporated herein by reference). This effect may contribute to reduced cell survival in IGF-1 R depleted cells (Figure 8A). In light of the identification of RAD51 in a screen for proteins whose depletion sensitizes to IGF-1 R inhibition, the inventors speculated that IGF-1 R inhibited cells are capable of residual HR, albeit with reduced efficiency, and are susceptible to manoeuvres that further comprise repair. Therefore, the inventors tested the effects of depleting RAD51 using two RAD51 siRNAs from the above-mentioned screen. Both mediated effective gene silencing, and RAD51 - depleted DU145 cells showed reduction in cell survival (Figure 8A). These results are consistent with the known toxicity of RAD51 loss in mammalian cells (Suzuki T, Fujii Y, Sakumi K, et al. Targeted disruption of the Rad51 gene leads to lethality in embryonic mice. Proc Natl Acad Sci U S A 1996; 93: 6236-40, incrporated herein by reference). As is standard (lorns E, Turner NC, Elliott R, et al. Identification of CDK10 as an important determinant of resistance to endocrine therapy for breast cancer. Cancer cell 2008; 13: 91-104, incorporated herein by reference), our screen included a cut-off for hits that were toxic in the absence of IGF-1 R inhibition. RAD51 had not been excluded on this basis, suggesting that RAD51 silencing had not crossed a threshold that impairs fitness, and that IGF-1 R inhibited cells are more sensitive to RAD51 depletion than controls. Indeed, both RAD51 siRNAs sensitized DU145 cells to AZ12253801 (Figure 8B). Depletion of IGF-1 R itself did not affect AZ12253801 sensitivity (Figure 8B), consistent with findings that IGF-1 R expression has little predictive value for response to IGF-1 R inhibition (King H, Aleksic T, Haluska P, Macaulay VM. Can we unlock the potential of IGF-1 R inhibition in cancer therapy? Cancer Treat Rev 2014, incorporated herein by reference). RAD51 depleted cells also showed reduced cell viability, and induced ~2-fold reduction in AZ12253801 Gl₅₀ (Figure 8C,D), comparable to the siRNA screen results. The inventors performed similar assays in three further prostate cancer cell lines, confirming RAD51 depletion by Western blotting (Figure 9A). The 22Rv1 cells were sensitized to IGF-1 R inhibition by both RAD51 siRNAs, with ~2-fold reduction in AZ12253801 Gl₅₀, as in DU145. There was no change in AZ12253801 sensitivity in RAD51 -depleted LNCaP cells, and a minor effect in PC3, unlikely to be of biological significance (Figures 8E, 9B-D). PC3 and LNCaP lack functional PTEN, while DU145 and 22Rv1 express wild-type (WT) PTEN (Figure 8E), as do >40% of prostate cancers (Phin S, Moore MW, Cotter PD. Genomic Rearrangements of in Prostate Cancer. Frontiers in oncology 2013; 3: 240, incorporated herein by reference). Given that PTEN encodes a phosphatase that reverses the action of PI3K (Phin S, Moore MW, Cotter PD. Genomic Rearrangements of in Prostate Cancer. Frontiers in oncology 2013; 3: 240, incorporated herein by reference), the inventors tested whether RAD51 depletion affects IGF signalling, but detected no evidence of altered activation of IGF-1 R or its effectors in RAD51 -depleted DU145 cells (Figure 11 E). It is possible that a link with PTEN status could be provided by the reported ability of AKT and PTEN to influence DSB repair, although the nature of the effect can vary with cellular context (Mendes-Pereira AM, Martin SA, Brough R, et al. Synthetic lethal targeting of PTEN mutant cells with PARP inhibitors. EMBO molecular medicine 2009; 1 : 315-22 and Xu N, Lao Y, Zhang Y, Gillespie DA. Akt: a double-edged sword in cell proliferation and genome stability. Journal of oncology 2012; 2012: 951724, both of which are incorporated herein by reference).

A key step in homologous recombination is the loading of RAD51 nucleofilaments onto DNA, a process that requires BRCA2 (Esashi F, Christ N, Gannon J, et al. CDK- dependent phosphorylation of BRCA2 as a regulatory mechanism for recombinational repair. Nature 2005; 434: 598-604, incorporated herein by reference). BRCA2 was present in the DNA repair set used in our siRNA screen, but was not identified as a hit. The inventors used two models to investigate whether BRCA2 influences sensitivity to IGF-1 R inhibition. Aiming to deplete DU145 cells of BRCA2, the inventors tested screen siRNAs BRCA2_6 and _7, and found that both were ineffective, while a third (Yata K, Lloyd J, Maslen S, et al. Plk1 and CK2 act in concert to regulate Rad51 during DNA double strand break repair. Mol Cell 2012; 45: 371 -83, incorporated herein by reference) did mediate effective target depletion (siBRCA2; Figure 8f). In viability assays, the response of DU145 cells to AZ12253801 was enhanced by siBRCA2 but not by the ineffective screen siRNAs (Figure 10F). This suggests that depletion of BRCA2, like RAD51 , is capable of sensitizing to IGF-1 R inhibition, and also that BRCA2 was a false negative in the screen. As a second approach, the inventors used isogenic DLD1 colorectal cancer cells that express WT PTEN and are BRCA2 WT (BRCA2^+/") or null (BRCA2^"'^") (Hucl T, Rago C, Gallmeier E, Brody JR, Gorospe M, Kern SE. A syngeneic variance library for functional annotation of human variation: application to BRCA2. Cancer Res 2008; 68: 5023-30, incorporated herein by reference). The cell lines had similar expression of IGF-1 R and downstream effectors, undetectable AKT phosphorylation consistent with WT PTEN status, and detectable BRCA2 only in BRCA2^+/" cells (Figure 8G, left). The effect of BRCA2 loss was tested by measuring the ability to generate irradiation-induced repair foci. Both cell lines formed γΗ2ΑΧ foci, indicative of damage induction, and BRCA2^+/" cells contained RAD51 foci that co-localized with γΗ2ΑΧ, while RAD51 foci were clearly reduced in BRCA2^"'^" cells, indicating defective homologous recombination (Figure 8G, right). Consistent with effects of BRCA2 knockdown in DU145 cells, BRCA2^"'^" DLD1 cells were more sensitive than isogenic BRCA2^+/" cells to AZ12253801 (Figure 8H), indicating that patients with BRCA2 mutant cancers may be sensitive to IGF-1 R inhibition.

The inventors next explored chemical approaches to inhibit HR. First, the inventors used small molecule drugs B02 and RL-1 , identified as capable of blocking RAD51 binding to DNA, and/or DNA strand exchange activity (Budke B, Logan HL, Kalin JH, et al. RI-1 : a chemical inhibitor of RAD51 that disrupts homologous recombination in human cells. Nucleic Acids Res 2012; 40: 7347-57 and Huang F, Mazina OM, Zentner I J, Cocklin S, Mazin AV. Inhibition of homologous recombination in human cells by targeting RAD51 recombinase. J Med Chem 2012; 55: 3011 -20, both of which are incorporated herein by reference). Viability assays were used to determine Gl₅₀ values for use in subsequent experiments (Figure 11A-D). In DU145 cells, radiation-induced RAD51 foci were inhibited by B02 but not by RL-1 (Figure 10A). RL-1 was thus identified as an ineffective compound. B02 sensitized to IGF-1 R inhibition, with 2.2 fold reduction in AZ12253801 Gl₅₀ (Figure 10B), similar to effects of RAD51 depletion (Figure 8D), while RL-1 did not affect the response to AZ12253801 (Figure 10B). These data suggest that the sensitization effect tracks with the ability to suppress RAD51 foci. B02 also suppressed RAD51 focus formation in PC3, LNCaP and 22Rv1 cells (Figure 1 1e), and induced AZ12253801 sensitization in 22Rv1 and PC3 cells, and no change in LNCaP (Figure 12A-C). Cyclin-dependent kinase (CDK) activity is required for serine phosphorylation of BRCA1 and 2 prior to RAD51 nucleofilament formation (Esashi F, Christ N, Gannon J, et al. CDK-dependent phosphorylation of BRCA2 as a regulatory mechanism for recombinational repair. Nature 2005; 434: 598-604, incorporated herein by reference). CDK1 inhibitor RO-3306 was reported to impair BRCA1 localization to DSBs, (Johnson N, Li YC, Walton ZE, et al. Compromised CDK1 activity sensitizes BRCA-proficient cancers to PARP inhibition. Nat Med 201 1 ; 17: 875-82, incorporated herein by reference), and consistent with this, the inventors observed that RO-3306 suppressed RAD51 focus formation in DU145 cells when applied at the Gl₅₀ (1 μΜ; Figures 10a, 11 C- E). Analysis of cell cycle distribution showed that cells co-treated with 100nM AZ12253801 and 1 μΜ RO-3306 manifested a marked increase in pre-G1 DNA, indicating apoptosis induction (Figure 10C). In viability assays, RO-3306 induced a significant shift to the left of the AZ12253801 dose-response curve, with 2.4-fold reduction in Gl₅₀, comparable to RAD51 depletion or inhibition. At sub-50nM AZ12253801 concentrations the inventors observed more striking sensitization, with 13- fold reduction in Gl₈₀ (Figure 10D). Paralleling effects of RAD51 knockdown, RO-3306 did not affect AZ12253801 sensitivity of PC3 or LNCaP cells (Figure 12D.E), but did sensitize 22Rv1 cells (Figure 12F); as in DU145, CDK1 inhibition was more effective than RAD51 depletion at sensitizing to low AZ12253801 concentrations, suggesting that CDK1 may influence response to IGF-1 R inhibition via processes in addition to homologous recombination modulation.

These results indicate that the response to IGF-1 R inhibition is enhanced in PTEN WT cells by manipulations that suppress BRCA2 and RAD51 function. It was previously reported that IGF-1 R inhibited cells show delayed repair of radiation-induced damage, with inhibition of repair by both NHEJ and homologous recombination in reporter assays (Chitnis MM, Lodhia KA, Aleksic T, Gao S, Protheroe AS, Macaulay VM. IGF-1 R inhibition enhances radiosensitivity and delays double-strand break repair by both nonhomologous end-joining and homologous recombination. Oncogene 2013, incorporated herein by reference). IGF-1 R inhibition was phenotypically silent in cells with an NHEJ defect induced by inhibition or loss of DNA-PK (Chitnis MM, Lodhia KA, Aleksic T, Gao S, Protheroe AS, Macaulay VM. IGF-1 R inhibition enhances radiosensitivity and delays double-strand break repair by both non-homologous end-joining and homologous recombination. Oncogene 2013, incorporated herein by reference), implying that with respect to its role in DSB repair, IGF-1 R acts primarily through NHEJ. This is consistent with the results of the current study, which support the concept that IGF-1 R inhibited cells depend on residual homologous recombination for repair of endogenous damage, and are sensitized to its inhibition. These findings suggest that HR-deficient tumors may be intrinsically sensitive to IGF-1 R inhibitory drugs.

The present inventors have thus demonstrated that the proportion of potentially- responsive patients in IGF-1 R trials can be significantly increased by selecting cases with impaired homologous recombination, for example cases having reduced levels or activity of RAD51 , BRCA2 and/or CDK1.

The present inventors have likewise demonstrated that sensitivity to IGF-1 R inhibition is enhanced by suppression of homologous recombination. Efficacy of treatment can thus be improved by prior and/or concurrent treatment with inhibitors of homologous recombination, for example by depletion or inhibition of RAD51 , BRCA2 and/or CDK1. To further validate the IGF-1 R TKI AZ12253801 hits in MCF7 and DU145 cells, additional experiments were carried out in MCF7 cells to test CDKN2C, CNKSR1 , DUSP5, HUNK and LMTK3 against a panel of small molecule TKIs and antibody antagonists of IGF-1 R. Initial assays focused on HUNK, which when depleted was capable of sensitizing MCF-7 cells to all tested IGF-1 R inhibitors including TKIs AZ12253801 , OSI-906, BMS 754807 (BMS807) and NVP-ADW742, and also to monoclonal blocking antibody figitumumab (Figure 13, Table 5).

Table 5: HUNK depletion sensitises breast cancer cells to IGF-1 R inhibitory drugs. MCF7 cells underwent transfection with non-silencing Allstars or HUNK siRNA. After 2 days, cells were treated with IGF-1 R inhibitory drugs and viability assayed after a further 5 days, as described herein. Pooled data from 3 independent assays were graphed and curved-fitted using GraphPad Prism v6, to interpolate Gl₅₀ values. Fold-sensitisation was calculated as ratio of Gl₅₀ values in Allstars siRNA transfected controls vs HUNK depleted cells. As in previous assays, and data from others Cosaceanu et al. Oncogene. 2007; 26:2423-2434), IGF-1 R antibody figitumumab was relatively ineffective at inhibiting cell viability compared with the TKIs. It was notable that HUNK depletion caused major reduction in viability of figitumumab-treated cells, with greater sensitisation (measured by fold-change in Gl₅₀) to figitumumab than to any of the TKIs (Figure 13D). Subsequent assays tested effects of depleting LMTK3, DUSP5, CDKN2C and CNKSR1 (Table 6).

Table 6: Effects of depletion of screen hits on response to IGF-1 R inhibitory drugs. MCF7 cells were transfected with non-silencing Allstars siRNA or siRNAs to deplete LMTK3, DUSP5, CDKN2C or CNKSR1. After 2 days, cells were treated with IGF-1 R inhibitory drugs and viability assayed as in legend to Table 5. Pooled data from 3 independent assays were pooled, and graphed and curved-fitted using GraphPad Prism v6, to interpolate Gl₅₀ values. N/D, not done. Figures in parentheses represent fold-sensitisation, calculated as ratio of Allstars Gl₅₀/Gene depletion Gl₅₀.

All hits were successfully re-validated against AZ12253801 , with fold-sensitisation (calculated as ratio of Gl₅₀ values) ranging from 2 - 13. In addition, LMTK3 was validated as a mediator of resistance to BMS807, DUSP5 to all agents except OSI-906, and CNKSR1 to OSI-906 and NVP-ADW742. CDKN2C depletion was capable of sensitizing to all tested TKIs and also to blocking antibody figitumumab.

These data demonstrate that major enhancement of sensitivity can be induced by depletion of individual resistance mediators. Furthermore, the biomarkers identified herein are capable of predicting sensitivity to IGF-1 R inhibitory drugs of different classes, including TKIs and receptor antibodies. Materials and methods

Cell lines and reagents

DU145 prostate cancer and MCF-7, MDA-MB-231 and MDA-MB-468 breast cancer cells were from Cancer Research UK Cell Services (Clare Hall Laboratories, Hertfordshire UK), PC3, LNCaP, LNCaP-LN3 and 22Rv1 prostate cancer cells were from Sir Walter Bodmer, and BT20, BT474 and BT549 from Dr Anthony Kong, University of Oxford. AZ12253801 and gefitinib were provided by AstraZeneca, DVL-PDZ inhibitor II (DVLi) was purchased from Calbiochem, BMS-754807 and XAV939 from Selleck Chemicals, and MAB391 from R&D. DVL3 cDNA in pcDNA3.1 vector (www.addgene.org/16758) was amplified using forward and reverse primers 5'- GG ATCCATG G ACTACAAG G ACG ACG ACG A-3' and 5'-

CTCGAGTCACATCACATCCACAAAGAACT-3' incorporating BamH1 and Xho1 sites respectively (underlined). The PCR product was digested with BamH1 and Xho1 (New England Biolabs), cloned into BamHI-Xhol digested pHRSIN-CSGW HIV vector, and after verification of the insert by DNA sequencing, used for virus production in HEK293 cells (as described in Demaison C, et al. Human Gene Therapy 2002; 13: 803-13; incorporated herein by reference). Supernatants containing viral particles were applied directly to DU145 cells. Western blotting, immunoprecipitation and pull-down assays

Whole cell extracts in IGF-1 R lysis buffer (1 % Triton-X-100) were analyzed by western blotting and immunoprecipitation (as described in Turney BW, et al. Radiother Oncol 2012; 103: 402-9; incorporated herein by reference). For immunoprecipitation and co- precipitation, cells were lysed in IGF-1 R lysis buffer or Co-IP buffer (150mM NaCI, 50mM Tris-HCI PH 7.5, 0.5% NP-40, with EDTA-free protease inhibitor cocktail, 1.5mM Pefabloc SC Plus, Roche and phosphatase inhibitor cocktails II and III, Sigma). For lambda phosphatase assay, cells were lysed in 150mM NaCI, 50mM Tris.CI pH7.5, 0.5% NP-40 for 30min on ice and cleared by centrifugation at 13,000rpm for 30min at 4°C. Protein extracts (100 g) were incubated with solvent or 800 units of lambda protein phosphatase (Sigma) with or without phosphatase inhibitor cocktail II and III (Sigma) at 30°C for 20min prior to analysis by western blot. GST, GST-c-RAF RAS binding domain (aa1 -149; GST-RBD) and GST fusion proteins representing full length GRB2 (GST- GRB2), GRB2 amino- or carboxy terminal SH3 domain (GST-N-SH3, GST-C-SH3) or GRB2 SH2 domain (GST-SH2) fusion proteins were purified and used (as described in Harkiolaki M, et al. Structure 2009; 17: 809-22; incorporated herein by reference). Cells were lysed in RAS assay lysis buffer (25mM Hepes pH7.5, 150mM NaCI, 10mM MgCI2, 1 mM EDTA, 10% glycerol, 1 % NP-40, 0.25% deoxycholate with protease inhibitors as above. Lysates were cleared by centrifugation at 13,000 rpm for 30min at 4°C. Each RAS activation assay used 500 g cell lysate, incubated with 50 g of GST or GST-RBD beads at 4°C for one hour with rotation. After washing 3-6 times with RAS assay lysis buffer, samples were analyzed by western blot to detect binding of activated (GTP- bound) RAS to RAF. For pulldown assays, cells were lysed in Co-IP lysis buffer, and precleared lysates were incubated with 50 g GST or GST fusion protein overnight with rotation at 4°C. After incubation with glutathione sepharose beads for 2hr at 4 °C and 3 washes with Co-IP lysis buffer, samples were analyzed by western blot using antibodies listed in Table 7.

IRS-1 Cell Signaling (2382]

Pan-Ras Calbiochem (OP40)

DAB 2 Abeam (ab76253)

Active β-catenin Millipore (05-665]

SHC BD Transduction Lab (610878]

Grb2 BD Transduction Lab (610111]

S0S1 BD Transduction Lab (6100960]

Androgen receptor Cell Signaling (3202]

Cyclin Dl Cell Signaling (2922]

β-tubulin Sigma

RAD51 Abeam (ab213]

BRCA2 (AB-1) Millipore (OP95]

Table 7: Antibodies used for western blotting, immunoprecipitation and immunofluorescence.

Radiographs were scanned and imported into Adobe Photoshop and faint exposures were used for quantification using ImageJ software.

Cell viability and cell survival assays

Cells were seeded or reverse transfected with siRNAs into u-clear 96-well plates (Grenier Bio-one) and after two days to allow target depletion, triplicate wells were treated with inhibitor or solvent, and assayed 5 days later using CellTiter-Glo (CTG) Luminescent cell viability assay (Promega). Clonogenic assays were performed as Chitnis MM, Lodhia KA, Aleksic T, Gao S, Protheroe AS, Macaulay VM (2013) IGF-1 R inhibition enhances radiosensitivity and delays double-strand break repair by both nonhomologous end-joining and homologous recombination. Oncogene, epub 4 Nov, by forward-transfecting with siRNAs, the following day reseeding into 10 cm dishes at 3000 cells/dish, and 24hr later treating with solvent or inhibitor. After 10-12 days, visible colonies were stained and counted. Values were normalized to solvent-treated cells and data from at least 3 independent experiments were pooled and curve-fitted using GraphPad Prism v5 to interpolate Gl₅₀ values (concentration required to inhibit 50% of growth) and SF₅₀ values (50% of survival fraction). Cell cycle distribution assays

Cells were fixed in ice cold 80% ethanol, stored at -20°C, and analyzed by propidium iodide staining for DNA content as described [2]. Samples were analysed on a CyAn ADP Analyzer (Beckman-Coulter, UK) with FlowJo 7.6.5 software (www.flowjo.com). Immunofluorescence

Cells were siRNA-transfected with siRNA as above and after 2 days were fixed and stained (as described in Aleksic T, et al. Cancer Res 2010; 70: 6412-9; incorporated herein by reference) with primary ERK antibody and secondary antibody conjugated to Alexa Fluor 488 (Molecular Probes AF21206), and mounted with ProLong Gold antifade reagent with DAPI (Molecular Probes). Slides were imaged on a Zeiss Axioskop 2 Plus fluorescence microscope and images were acquired using AxioVision software.

Other cells were stained using the methodology above, using antibodies to y H2AX and RAD51. Detection used highly cross-adsorbed Alexa-Fluor 488- or 594-conjugated anti- mouse and/or anti-rabbit secondary antibodies (Invitrogen Molecular Probes, Eugene, USA). After mounting in Fluoromount G with 2 μ g/ml 4'6'-diamidino-2-phenylindole (DAPI), foci were imaged on an Axioskop 2 or Observer.ZI microscope. Foci were counted in >50 cells per condition from >2 separate experiments. siRNA screens

The screens were performed (as described in Lord CJ, et al DNA Repair (Amst) 2008; 7: 2010-9; incorporated herein by reference), using protein kinase siRNA library (siARRAY, targeting 779 known and putative human protein kinase genes; Dharmacon) containing SMARTPools of four distinct siRNAs targeting each transcript, and human DNA Repair siRNA Set V1.0 siRNA library (Qiagen), together with siPLKI and Allstars siRNA as positive and negative controls respectively. The siRNAs were diluted to 2μΜ (final concentration in transfection 50nM) and aliquoted into 96-well plates. Cells were reverse-transfected into duplicate plates using Dharmafect 1 reagent for DU145 and Dharmafect 3 for MCF-7 cells in serum-free medium (SFM) without antibiotics. After 6.5hr, transfection medium was replaced with complete medium including 10% fetal calf serum (FCS). Two days later the cells were exposed to vehicle (0.01 % DMSO) or AZ1223580 at the Gl₅₀ concentration for 5 days, after which cell viability was assessed by CTG assay. Data from duplicate primary screens were analyzed to derive Z' factors (dynamic range) and Z-scores (effect of siRNA on cell viability, corrected for within plate and between plate variation), as described in Lord CJ, et al DNA Repair (Amst) 2008; 7: 2010-9; Boutros M, Ahringer J. Nat Rev Genet 2008; 9: 554-66; and Martin SA, et al. Cancer Res 201 1 ; 71 : 1836-48 (all of which are incorporated herein by reference). Second round screens of 53 potential hits (Table 8) used 4 siRNAs to each target; these siRNAs and additional siRNAs are listed in Table 8.

Hs_RAD51_7 QIAgen (SI026636682]

Hs_BRCA2_6 QIAgen (SI02653434)

Hs_BRCA2_7 QIAgen (SI02653595)

BRCA2 siRNA Sigma; siBRCA2

Table 8: siRNA libraries and individual siRNAs used in this study.

Triplicate second round screens were performed in both cell lines, using the four individual siRNAs and the 4 siRNAs pooled. The data were analyzed (as described in Lord CJ, et al. DNA Repair (Amst) 2008; 7: 2010-9; incorporated herein by reference) to generate log2 surviving fractions as the final score for the effect of IGF-1 R inhibition on viability; scores of < -0.2 were regarded as significant. Hit validation

To assess knockdown, cells were reverse-transfected with siRNAs using Oligofectamine (InVitrogen, USA) or Dharmafect reagent 1 or 3 (Dharmacon) according to the manufacturer's instructions. After 48-72 hours, cells were collected and lysed in IGF-1 R lysis buffer for western blot as above, or used for RNA extraction using the RNeasy micro kit (QIAgen). After treatment to remove DNA using the DNA-freeTM Kit (AB Applied Biosystems), RNA concentrations were measured by spectrophotometer (Nanodrop technologies, Wilmington, DE, USA). cDNA was synthesized by reverse transcribing total RNA using the High Capacity cDNA Archive® Kit (AB Applied Biosystems) according to the manufacturer's instructions. qRT-PCR reactions were performed in triplicate in the Rotor Gene 3000® qPCR system (Corbett Research, Sydney, Australia) using the SensiMixTM SYBR Kit (www.quantace.com) according to the manufacturer's instructions.

Xenografts

In vivo work was carried out at Breakthrough Breast Cancer Research Centre London UK under a Home Office approved Project License. Prostate cancer xenografts were established by injecting 107 DU145 cells with an equal volume of matrigel (BD Biosciences) into the flanks of 6-7 week old male Balb/c immunodeficient (Nu/Nu) mice. Tumor volumes were measured twice a week, and mice were randomly allocated to four treatment groups when tumor volumes reached 100-200mm3. Treatments were administered by intraperitoneal injection in 0.05ml DMSO per dose twice daily for 5 days a week and once daily for 2 days a week for a total of 14 days. The treatment groups comprised: solvent (DMSO) alone, 25mg/kg AZ12253801 twice daily, 50mg/kg DVLi once daily or combination treatment (DVLJ+AZ12253801 ).

Immunohistochemistry (IHC)

Permission to use human tissues in research was covered by Research Ethics Committee approved study 07/H0606/120. DVL3 IHC was performed on a Bond Max auto-stainer (Leica) using antigen retrieval in Bond ER2 epitope retrieval solution at 100oC for 20 min followed by cooling to room temperature for 12 min in ER2 retrieval solution. DVL3 antibody (4D3, Santa Cruz) was applied at a 1 :200 dilution in primary antibody diluent (Bond) for 30min at room temperature and detection used the Polymer Refine detection kit (Bond). Scoring of DVL3 was performed by CV and RA by intensity (0, no staining; 1 , weak staining; 2, moderate; 3 intense staining) and percentage positivity (0, no staining; 1 , 1-10% positive; 2, 1 1-50% positive; 3, 51-80% positive and 4, >80% positive). The product of these scores (Intensity x Percentage Score, IPS) was correlated with progression-free survival from trial data (Macaulay VM, et al. Ann Oncol 2013; 24: 784-91 ; and Schmitz S, et al. Ann Oncol 2012; 23: 2153-61 ; both of which are incorporated herein by reference).

Statistical analysis

Prism v.5.0 (GraphPad) and Excel (Microsoft) were used to plot and analyze data. Graphs were plotted to show mean values and standard error of the mean (SEM). The student's t-test and analysis of variance (ANOVA) were used to compare mean values between two and multiple (>2) data sets respectively, and two-way ANOVA to study interactions between two variables. Xenograft growth rates were compared by repeated measures ANOVA. A minimum of 95% level of significance (p<0.05) was used to define statistical significance.

Claims

1. A method of determining whether a patient is likely to benefit from treatment with IGF-1 R inhibition, the method comprising determining the level, in a tumour sample taken from the patient, of one or more biomarkers selected from the list consisting of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C, BRCA2, CDK1 and/or RAD51 , wherein a high level of the biomarker in the sample indicates that a patient is unlikely to benefit from treatment with IGF-1 R inhibition and a low level of biomarker in the sample indicates that the patient is likely to benefit from treatment with IGF-1 R inhibition.

2. A method of determining whether a patient is likely to benefit from treatment with IGF-1 R inhibition, the method comprising determining the activity, in a tumour sample taken from the patient, of one or more biomarkers selected from the list consisting of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C, BRCA2, CDK1 and/or RAD51 , wherein reduced activity of biomarker in the sample indicates that the patient is likely to benefit from treatment with IGF-1 R inhibition.

3. The method of claim 1 or claim 2, wherein the one or more biomarkers are selected from the list consisting of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C and/or RAD51.

4. The method of any one of the preceding claims, wherein the one or more biomarkers are selected from the list consisting of DVL3, HUNK, LMTK3, CDKN2C, CNKSR1 , DUSP5 and/or RAD51.

5. A method of determining whether a patient is likely to benefit from treatment with IGF-1 R inhibition, the method comprising determining the level, in a tumour sample taken from the patient, of one or more biomarkers associated with homologous recombination, wherein biomarker levels associated with impaired homologous recombination indicate that the patient is likely to benefit from treatment with IGF-1 R inhibition.

The method of claim 5, wherein said impaired homologous recombination is characterised by a low level of one or more of RAD51 , BRCA2 and CDK1 in the sample.

A method of determining whether a patient is likely to benefit from treatment with IGF-1 R inhibition, the method comprising determining the activity, in a tumour sample taken from the patient, of one or more biomarkers associated with homologous recombination, wherein biomarker activity associated with impaired homologous recombination indicates that the patient is likely to benefit from treatment with IGF-1 R inhibition.

The method of claim 7, wherein said impaired homologous recombination is characterised by reduced activity of one or more of RAD51 , BRCA2 and CDK1 in the sample.

The method of any one of the preceding claims, further comprising the step of treating a patient having low expression of the biomarker with an IGF-R1 inhibitor.

The method of any one of the preceding claims, further comprising the step of treating a patient having reduced activity of the biomarker with an IGF-R1 inhibitor.

The method of claim 9 or claim 10, wherein the IGF-1 R inhibitor is a kinase inhibitor.

The method of claim 9 or claim 10, wherein the IGF-1 R inhibitor is an antibody, directed against IGF-1 R itself or the ligands IGF-1 and IGF-2.

A method of treating a patient having tumour, the method comprising administering to the patient:

(i) A drug which blocks expression of one or more of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C, BRCA2, CDK1 and/or RAD51 ; and

(ii) An IGF-1 R inhibitor.

14. The method of claim 13, wherein the drug blocks expression of one or more of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C and/or RAD51.

15. The method of claim 13 or claim 14, wherein the drug blocks expression of one or more of DVL3, HUNK, LMTK3, CDKN2C, CNKSR1 , DUSP5 and/or RAD51.

16. A method of treating a patient having tumour, the method comprising administering to the patient:

(iii) A drug which blocks activity of one or more of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C, BRCA2, CDK1 and/or RAD51 ; and

(iv) An IGF-1 R inhibitor.

17. The method of claim 16, wherein the drug blocks activity of one or more of CDKN2C, CNKSR1 , DUSP5, HUNK, LMTK3, DVL3, AKAP1 , APEG1 , MPP2, CDKN3, INO80C and/or RAD51.

18. The method of claim 16 or claim 17, wherein the drug blocks activity of one or more of DVL3, HUNK, LMTK3, CDKN2C, CNKSR1 , DUSP5 and/or RAD51.

19. The method of any one of claims 13 to 18, wherein the IGF-1 R inhibitor is an IGF-1 R kinase inhibitor.

20. The method of claim 19, wherein the IGF-1 R inhibitor is an antibody, directed against IGF-1 R itself or the ligands IGF-1 and IGF-2.

21. The method of any one of claims 13 to 20, wherein the patient has been previously identified as unlikely to benefit from treatment with an IGF-1 R inhibitor.

22. The method of claim 21 , wherein the patient has been identified using the method of any one of claims 1 to 12.

23. The method of any one of claims 13 to 22, wherein the patient has received prior treatment with an inhibitor of homologous recombination.

24. The method of any one of claims 13 to 23, wherein said treatment comprises concurrent administration of an inhibitor of homologous recombination.

25. The method of claim 23 or claim 24, wherein said inhibitor of homologous recombination is suitable for depletion of the level of one or more of RAD51 , BRCA2 and CDK1.

26. The method of any one of claims 23 to 25, wherein said inhibitor of homologous recombination is suitable for blocking the activity of one or more of RAD51 , BRCA2 and CDK1.

27. The method of any one of the preceding claims, wherein the tumour or tumour sample is selected from the list consisting of: a solid tumour, prostate cancer, bladder cancer, renal cancer, small cell or non-small cell lung cancer, colon cancer, pancreatic cancer or melanoma.

28. The method of any one of the preceding claims, wherein the tumour or tumour sample is selected from the list consisting of: breast cancer, Ewing sarcoma and head and neck cancer.

29. The method of any one of the preceding claims, wherein said high level of biomarker determined by an Intensity X Percent Score of greater than 5.

30. The method of any one of the preceding claims, wherein said low level of biomarker is determined by an Intensity X Percent Score of less than 5.