Immune and growth factor signaling pathways are associated with pathologic complete response to an anti-type I insulin-like growth factor receptor regimen in patients with breast cancer

Abstract

Purpose:

Pre-treatment specimens from patients treated on the I-SPY2 neoadjuvant breast cancer trial were studied to identify pre-specified biomarkers associated with response to the regimen of paclitaxel, an anti-type I IGF receptor (IGF-1R) antibody ganitumab, and metformin (PGM) followed by doxorubicin and cyclophosphamide (AC) compared to control therapy (paclitaxel followed by AC). The primary endpoint of this trial is pathologic complete response (pCR).

Experimental Design:

106 patients treated with PGM and 119 contemporary controls were evaluated by laser capture microdissection and reverse phase protein array to evaluate 32 pre-specified potential predictive biomarkers in the IGF-1R pathway and 109 additional exploratory endpoints.

Results:

Total levels of IGF-1R were poorly correlated with phosphorylated IGF-1R/Insulin receptor (pIGF-1R/IR). Higher levels of phospho-IGF-1R/IR were associated with increased likelihood of obtaining pCR especially in the hormone receptor positive (HR+) subgroup. Markers of immune response also showed an association with pCR but differed between HR+ and HR− subgroups. In HR− tumors, phospho-STAT1 Y701 and low levels of phospho-p27 associated with pCR. These relationships were not observed in patients treated with control chemotherapy.

Conclusions:

Activation status of IGF-1R/IR associated with increased pCR to PGM in HR+ breast cancers. Immune activation markers were also associated with response in HR+ and HR− subgroups. Thus, IGF-1R may directly regulate tumor biology and associate with immune response to therapy.

Introduction

The type I insulin-like growth factor receptor (IGF-1R) has emerged as a potential target in cancer therapy, supported by population, preclinical, and translational studies. This signaling system is responsible for normal tissue growth during development. Preclinical data suggest that IGF-IR signaling is required for normal mammary gland development (1) and in an analogous fashion to estrogen receptor (ER), inhibition of receptor function could be a therapeutic strategy. Multiple IGF-1R monoclonal antibodies (moAbs), tyrosine kinase inhibitors that disrupt the intracellular kinase of the receptor, and neutralizing antibodies to the two ligands (IGF-1 and IGF-2) have been tested in multiple clinical trials. To date, the addition of IGF-1R inhibitors to standard-of-care therapy has not resulted in improvements in clinical outcomes and essentially all drugs have been abandoned (2). Although generally deemed safe, hyperglycemia was a common adverse effect, presumably due to insulin resistance induced by a reflex increase in growth hormone caused by IGF-1R signaling disruption. However, it is notable that one IGF-1R monoclonal antibody (moAb), teprotumumab, has been approved for thyroid eye disease, a proliferative disease of orbital soft tissue (3), proving the clinical relevance of IGF-1R targeting in a proliferative disease.

While IGF-1R inhibitor trials in breast cancer have not required expression of the target (as in vitro levels of IGF-1R do correlate with its biological effects), post hoc analysis of IGF-R1 mRNA in one trial showed very low levels of IGF-1R in endocrine resistant tumors (4). Many have hypothesized that response is tied to higher levels of IGF-1R mediated gene expression (5,6), Since IGF-1R requires ligand activation to signal, serum evaluation of IGF-1 and IGF-2 has also been explored with mixed results (7,8). However, there have been no direct measurements of receptor activation and its relationship to response to anti-IGF therapies.

Previous investigators found that phosphorylated IGF-1R and insulin receptor (IR) in breast cancer associated with poor prognosis while total levels of IGF-1R were not prognostic (9). This study, along with the lack of correlation of mRNA and protein expression with response prediction, suggested that the activation status of the receptor may be more informative than the total levels of receptor protein expression in association with response. This would be in keeping with previous results wherein phosphorylated AKT, phosphorylated HER2 and phosphorylated TIE2 associated with response to AKTi, HER2 TKIs and a TIE2 peptibody respectively, while the total protein expression levels and DNA/RNA cognates did not (10–12). We hypothesized that the activation state of IGF-1R and downstream signaling molecules would be informative. To test if phosphoprotein analysis would be useful for therapy with an IGF-1R inhibitor, we examined pre-treatment tumor specimens from patients treated with paclitaxel, ganitumab (IGF-1R moAb) and metformin (PGM) in the neoadjuvant I-SPY2 trial for early breast cancer (13), in which there was evidence of numerically improved pathologic complete response (pCR) rates in triple negative breast cancer, but the arm did not meet the primary endpoint. We evaluated several putative mRNA-based predictive biomarkers of ganitumab response mostly developed in cell lines (e.g., IGF-1 ligand score (5), IGF-1R signature (6), IGFBP5 expression, baseline HbA1c) (13). None were specific predictors of response to PGM, although several signatures were associated with pCR in both the control and PGM arms.

We have recently shown that evaluating the activation state of signaling pathways by reverse phase protein array (RPPA) of laser capture microdissection (LCM)-enriched tumor epithelium provides important biomarker information (14,15). Moreover our previous work has consistently demonstrated that the LCM-RPPA data produced in our CLIA-CAP accredited proteomics laboratory retains high fidelity of the in vivo state of the tumor signaling architecture, and retains accurate in vivo information about ongoing phosphoprotein-driven functional signaling including identification of specific predictive markers (i.e. phosphoproteins) that match the mechanism of action of the targeted agents used such as HER2 TKI agents such as neratinib (11), TDM-1/Pertuzumab (11,16), AKT inhibitors (10,17), TIE2 inhibitors (12). Moreover, we have previously shown that specific highly actionable functional HER2 and HER family protein and phosphoprotein data produced by the LCM-RPPA workflow in our laboratory is highly concordant with FDA IHC and FISH HER2 determination in HER2+ BC (15,16).

Thus, in this report, we have applied functional RPPA analysis to identify predictive biomarkers for ganitumab response. We demonstrate that in pre-treatment biopsy samples with high levels of epithelium, activation status of key signaling proteins identified tumors that were more responsive to PGM on a subtype-specific basis. These included IGF-1R itself, those involved in downstream target activation and cell cycle control, and those involved in immune signaling/activation.

Material and Methods

Patients and trial schema

The study was approved by the institutional review boards at each participating clinical site, and an independent Data Safety Monitoring Board met monthly to review patient safety and study progress. The I-SPY2 trial complies with all local and national regulations regarding the use of human study participants and was conducted in accordance to the criteria set by the Declaration of Helsinki. All participants provide written informed consent prior to screening and again upon randomization, prior to treatment.

Figure 1A demonstrates the adaptive randomization of patients on the multi-center, multi-arm neoadjuvant I-SPY2 trial (NCT01042379; IND 105139). I-SPY2 is an adaptive phase 2 platform trial designed to test novel drug combinations for neoadjuvant therapy of early breast cancer. The primary endpoint is pathologic complete response (pCR) and drug combinations are successful if they exceed the pCR rate of control standard-of-care chemotherapy using with a prespecified threshold of graduation (≥85% probability of success in a hypothetical phase 3 controlled trial). This correlative study involved an arm consisting of 106 patients treated with the experimental regimen of paclitaxel, ganitumab, and metformin (PGM) and 119 contemporary control patients. HER2-positive patients were not enrolled and Figure 1B shows the distribution of patients assigned to PGM versus control. Detailed descriptions of the design, eligibility, and study assessments in I-SPY2 have been reported previously (18), including the efficacy of investigational agent ganitumab and metformin plus standard chemotherapy (PGM; PGM+T->AC) (13). PGM was active in the trial from July 2012 to February 2015. Pre-treatment specimens from 102 of the 106 patients randomized to the investigational arm PGM plus standard chemotherapy were available to study (Figure 1B). 119 control specimens from HER2-negative patients were available for comparison.

Figure 1 –

Ganitumab schema with qualifying biomarkers

A. I-SPY 2 trial schema

B. Treatment subtypes included in this study

C. Qualifying phospho-biomarkers evaluated for ganitumab response

Pre-treatment processing and protein/phospho-protein profiling

Core needle biopsies of 16-gauge were taken from the primary breast tumor before treatment. Collected tissue samples are immediately frozen in Tissue-Tek^® O.C.T.^™ embedding media and then stored in −80°C until further processing. An 8μM section was stained with hematoxylin and eosin (H&E) and pathologic evaluation was performed to confirm the tissue contains at least 30% tumor. A tissue sample meeting the 30% tumor requirement was further cryosectioned at 30 μM. Details of the tissue processing were previously described (14).

In addition to gene expression profiling (mRNA profiling data for first 10 arms of I-SPY 2 including PGM) were previously deposited in NCBI’s Gene Expression Omnibus (GEO) GSE194040 (18), LCM was performed on pre-treatment biopsy specimens to isolate and enrich tumor epithelium for protein and phospho-protein profiling by reverse phase protein arrays (RPPA) as previously published (11). Approximately 10,000 cells were captured per sample. RPPA samples from PGM were assayed on an array quantifying 141 protein/phospho-protein endpoints (GPL28470). Samples from the control arm were previously assayed on three earlier arrays (also GPL28470) quantifying 141 analytes on hundreds of samples from other arms, then batch adjusted and combined as previously described (NCBI’s Gene Expression Omnibus (GEO) GSE196093 (18)). RPPA data for the PGM arm have been deposited in GEO (GSE 298828), along with clinical data for patients in this study including response (pCR yes/no).

‘Qualifying’ protein/phospho-protein biomarkers assessed

Based upon ganitumab mechanism of action and our previous RPPA-based biomarker studies, we hypothesized that patients with tumors expressing high levels of IGF-1R activation and its downstream signaling molecules are more likely to respond to ganitumab (Figure 1C and Table 1). Thus, our hypothesis-driven pre-specified selection of 32 ‘qualifying’ biomarkers for testing as specific PGM response-predictive markers included proteins/phospho-proteins from the following pathways: IGF1R/AMPK (n=9) including phosphorylated IGF1R/IR and IRS1, AKT (n=9), ERK (n=3), and mTOR (n=11). Notably, the antibody used to detect the tyrosine phosphorylation status of IGF-1R also detects the analogous sites on the insulin receptor (IR) (Supplementary Table S1).

Table 1 –

Correlation between 32 qualifying biomarkers (QB) measured by reverse phase protein array (RPPA) in the entire ganitumab population, HR+/HER2−negative subset, and triple negative (TN) subset.

RPPA endpoint (n=32 QBs)	Entire Ganitumab population (n=102)						HR+HER2− subset (n=58)			TN subset (n=44)
	pCR~RPPA			pCR~RPPA + HR (adjusting for HR)			pCR~RPPA (HR+HER2− only)			pCR~RPPA (TN only)
	OR/1SD	LR p	BH LR p	OR/1SD	LR p	BH LR p	OR/1SD	LR p	BH LR p	OR/1SD	LR p	BH LR p
IGF-1 Rec Y1135/Y1136_Insulin Rec Y1150/Y1151 (pIGF1-Rec)	2.13	0.0091	0.146	2.68	0.00191	0.0314	5.1	0.000523	0.0167	1.28	0.697	0.934
p27 T187	0.41	0.00203	0.065	0.419	0.00196	0.0314	0.424	0.104	0.522	0.417	0.0084	0.269
IGF-1 Rec beta total	1.59	0.0785	0.359	2.23	0.0205	0.219	2.69	0.0144	0.154	0.222	0.535	0.934
AKT T308	0.637	0.104	0.416	0.594	0.0696	0.557	0.564	0.246	0.541	0.609	0.161	0.708
FOXO1 T24/FOXO3a T32	1.56	0.0478	0.306	1.4	0.166	0.891	1.96	0.0656	0.522	1.08	0.802	0.934
IRS1 S612	1.62	0.0446	0.306	1.42	0.167	0.891	1.02	0.968	0.968	1.73	0.0887	0.661
SGK S78	0.937	0.786	0.992	0.739	0.263	0.989	1.05	0.898	0.927	0.474	0.0764	0.661
AMPKb1 S108	1.49	0.0706	0.359	1.3	0.283	0.989	1.68	0.159	0.522	1.07	0.817	0.934
p90RSK S380	1.58	0.04	0.306	1.31	0.285	0.989	1.74	0.321	0.541	1.22	0.478	0.934
Insulin Rec beta total	1.2	0.386	0.81	1.23	0.369	0.989	1.38	0.196	0.541	0.801	0.64	0.934
LKB1 S334	1.32	0.253	0.674	1.23	0.413	0.989	4.08	0.00182	0.0291	0.586	0.124	0.661
PTEN S380	0.754	0.274	0.674	0.816	0.43	0.989	0.736	0.479	0.676	0.864	0.651	0.934
p70S6K T389	1.24	0.323	0.738	1.19	0.451	0.989	1.71	0.082	0.522	0.76	0.448	0.934
FOXO1 S256	1.06	0.798	0.992	0.84	0.515	0.989	1.26	0.673	0.798	0.713	0.325	0.934
eIF4E S209	1.15	0.553	0.93	1.15	0.575	0.989	1.28	0.486	0.676	1.04	0.907	0.936
PTEN total	0.81	0.405	0.81	0.873	0.596	0.989	0.76	0.514	0.676	0.95	0.875	0.936
p70S6K S371	0.958	0.859	0.992	1.13	0.637	0.989	1.27	0.383	0.613	0.591	0.426	0.934
MEK1/2 S217/S221	1.35	0.189	0.605	1.12	0.639	0.989	1.42	0.431	0.657	1.01	0.96	0.96
4EBP1 S65	1.34	0.186	0.605	1.12	0.657	0.989	1.84	0.131	0.522	0.846	0.599	0.934
FOXO3a S253	0.874	0.581	0.93	0.901	0.682	0.989	1.64	0.163	0.522	0.411	0.0375	0.6
mTOR total	0.906	0.685	0.992	0.92	0.744	0.989	0.781	0.528	0.676	1.04	0.898	0.936
eIF4G S1108	1.29	0.265	0.674	1.07	0.795	0.989	1.48	0.296	0.541	0.829	0.58	0.934
mTOR S2448	0.982	0.939	0.992	0.953	0.847	0.989	1.43	0.316	0.541	0.657	0.236	0.839
SHC Y317	1.15	0.547	0.93	1.04	0.862	0.989	1.2	0.597	0.735	0.917	0.796	0.934
ERK1/2 T202/Y204	1.02	0.924	0.992	0.957	0.863	0.989	0.555	0.29	0.541	1.19	0.59	0.934
AMPK alpha T172	0.998	0.992	0.992	0.966	0.89	0.989	1.12	0.764	0.873	0.866	0.661	0.934
S6RP S240/S244	0.962	0.875	0.992	0.974	0.916	0.989	1.48	0.208	0.541	0.45	0.115	0.661
GSK3aB S21/S9	1.04	0.853	0.992	0.979	0.933	0.989	1.06	0.865	0.923	0.893	0.763	0.934
S6RP S235/S236	1.01	0.959	0.992	1.01	0.952	0.989	1.59	0.155	0.522	0.483	0.177	0.708
p70S6K T412	1.19	0.446	0.84	1.01	0.966	0.989	1.51	0.295	0.541	0.808	0.491	0.934
AKT S473	1	0.99	0.992	0.994	0.98	0.989	1.07	0.841	0.923	0.916	0.809	0.934
IGF1R Y1131/IR Y1146	0.991	0.969	0.992	1	0.989	0.989	1.39	0.32	0.541	0.688	0.318	0.934

OR/1SD = odds ratio of pCR per 1-standard deviation in the marker, LR p = likelihood ratio test p-value, BH LR p = Benjamini-Hochberg adjusted p-value.

We also performed exploratory pCR association analysis over the entire set of 141 RPPA endpoints, which include proteins/phospho-proteins from ER, proliferation, immune, DNA damage and repair, angiogenesis, and other pathways (Supplementary Table S1).

Statistical analysis

To assess association with pCR in the PGM arm we used logistic regression overall, in a model adjusting for hormone receptor (HR) status, and within HR+HER2− and triple negative (TN) subsets. The Benjamini-Hochberg (BH) method was used to adjust p-values for multiple hypothesis testing. Receiver Operator Curve (ROC) analysis was used to further quantify performance and to identify dichotomizing thresholds for select biomarkers (Youden criterion).

Data availability statement

All clinical data and qualifying biomarker data analyzed in this manuscript appear in Supplemental Data (Tables S1–S4). The complete RPPA dataset (including exploratory endpoints) for the PGM + control arms is publicly available on the Gene Expression Omnibus (GEO) GSE298828 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE298828). Normalized data can be found in the Series Matrix File for GSE298828, and raw data is available in the same record as a Supplementary File (GSE298828_ISPY2_gan_controls_RPPA_rawDat_RIDsonly.csv.gz).

Results

Hypothesis-driven ‘Qualifying’ biomarker analysis identifies phosphorylated IGF-1R and p27 as predictive

All 32 qualifying protein biomarkers (Table 1, Figure 2A) were examined for their ability to associate with pCR. There was a weak correlation between phospho-IGF1-R/IR and IGF-1R levels in the overall group (Figure 2B) and no correlation in the HR+ subgroup (Figure 2C). The lack of strong correlation between total levels of IGF-1R and benefit from PGM is consistent with other findings showing that total IGF-1R is not a predictive biomarker for anti-IGF-1R therapies (19).

Figure 2 –

Qualifying biomarker analysis. pIGF-1 Rec* represents phospho-IGF-1R and phospho IR

A. Heatmap showing relationship between response (pCR or not), therapy (ganitumab or control), and biomarker expression

B. Correlation between total IGF-1R and phospho-IGF-1R/IR in all ganitumab treated patients.

C. Correlation between total IGF-1R and phospho-IGF-1R/IR in HR+/HER2− subgroup ganitumab treated patients.

In contrast, higher levels of phosphorylated IGF-1R (Y1135/Y1136)/IR (Y1150/Y1151) (OR=2.13 p=0.002; BH p<0.05; pIGF-1R/IR) were significantly associated with an increased likelihood of obtaining pCR in the PGM treated patients in a model adjusting for HR status (Figure 3A, Table 1) although this observation was seen most strongly in HR+ subgroup (OR=4.08 p=0.00052; LR p<0.05; Figure 3C and D) with an AUC of 0.79 [0.56–1]. Several other activated proteins (p90RSK, IRS-1, FOXO1) were nominally associated with increased pCR and these molecules all lie downstream of activated IGF-1R/IR (Figure 3A, Table 1).

Figure 3 -.

Predictive biomarkers in ganitumab treated and control patients. pIGF-1 Rec* represents phospho-IGF-1R and phospho IR

A. Association of individual signaling molecules and pCR. Red circles represent a positive correlation, blue circles represent a negative correlation. The size of the circle represents the strength of the association.

B. Relationship between pCR and ganitumab therapy in TNBC

C. Relationship between pCR and ganitumab or control therapy in HR+/HER2− patients

D. ROC curve for phospho-IGF-R in HR+/HER2− patients

Dichotomized using the Youden criterion, 16% of HR+HER2− were phospho-IGF-1R/IR-High. 56% (5/9) of phospho-IGF-1R/IR-High patients in the HR+ subgroup achieved pCR in PGM. In contrast only 3 of 49 (6%) HR+ patients with low p phospho-IGF-1R/IR achieved pCR. Thus, high levels of phospho-IGF-1R/IR positively predict response to PGM while low levels associated with low chance of pCR in HR+ tumors.

For TNBC, phospho-IGF-1R/IR did not associate with response to PGM (Figure 3B; OR=1.28 p>0.05). In contrast, low levels of phospho-p27 (T127) nominally associated with higher levels of pCR in PGM treated patients with TNBC (OR=0.42 p=0.0048; BH p>0.05), with an AUC of 0.74 [0.58–0.89] (not shown). Dichotomized using the Youden criterion, 59% (25/42) of TN were phospho-p27-Low. 52% (13/25) of TN patients with low p27 T187 had pCR, whereas only 12% (2/17) of TN with high p27 T187 achieved pCR.

Figure 3A and Supplemental Table S2 shows that none of the 32 qualifying protein biomarkers associated with pCR in the paclitaxel-only treated control group.

Exploratory biomarker analysis reveals immune participation in response

We wondered to what extent pathways other than those clearly related to ganitumab’s mechanism of action may play a role in response to PGM. Therefore, we performed exploratory pCR association analysis over the entire set of 141 protein/phospho-protein analytes we measured by RPPA, and found that mostly immune, and a few other pathways contribute to response (Supplemental Table S1 and Figure 4). High tumor epithelium activation of STAT1 at Y701 significantly associated with pCR (OR=2.64, p=0.00024, BH p<0.05) in the population as a whole in a model adjusting for HR status, and nominally associated with pCR in both TN (OR=3.5 p=0.00128, BH p>0.05) and HR+ (OR=2 p=0.0483, BH p>0.05) subsets (see Supplemental Table 1 and Figure 4A–C). STAT1 signaling is also implicated in IGF-1R signaling pathways (20) along with other growth factor and cytokine signaling pathways (21). Because STAT1 is involved in many different signaling pathways including cytokine signaling, we examined other proteins/phospho-proteins associated with immune function.

Figure 4 -.

Exploratory biomarker analysis for ganitumab and control therapy. pIGF-1 Rec* represents phospho-IGF-1R and phospho IR

A. Association dot plot of additional biomarkers. Red circles represent a positive correlation, blue circles represent a negative correlation. The size of the circle represents the strength of the association.

B. Relationship between phospho-STAT1 Y701 in ganitumab and control therapy in all patients

C. Relationship between phospho-STAT1 Y701 in ganitumab and control therapy in TNBC

D. Correlation between p27 T187 and STAT1 Y701 in ganitumab therapy in TNBC patients

E. Correlation between p27 T187 and phospho-IGF-1R/IR and CD3 zeta in ganitumab therapy in HR+, HER2− patients

Several other immune protein/phospho-protein analytes showed at least nominal associations to response in one or more subsets. For instance, within HR+, CD3 zeta significantly associated with response (OR=4.38 p=0.00028, BH p<0.05) and HLA-DR, HLA-A total, PD-L1 total (E1L3N) and CTLA-4 total nominally associated with pCR (uncorrected p<0.05, BH p>0.05; Table S1 and Figure 4A,E). Within TN, in addition to nominal associations of pCR with high levels of STAT1 Y701, and low levels of phospho-p27 per above; low levels of ER, phospho-EGFR, phospho-beta catenin, and phospho-PI3K were nominally associated with pCR (uncorrected p<0.05; BH p>0.05; Table S1, Figure 4A,D). With the exception of STAT1 Y701 in the HR+ subset, these associations were not observed in the control treated patients (Figure 4A, C).

To further investigate the exploratory RPPA results suggesting that an immune-enriched phenotype associates with response to PGM, we performed a similar exploratory analysis using selected mRNA-based immune biomarkers. Continuous immune signatures assessed include B cell and Dendritic cell immune signatures (22), a STAT1-chemokine/cytokine signature (23), and a chemokine/cytokine signature associated with tertiary lymphoid structures (24), all of which were previously shown to associate with the pembrolizumab IO arm in I-SPY2(18). We also assessed association between PGM response and the dichotomous immune biomarker (Immune+/Immune-) used in the research-grade Response Predictive Subtypes developed in I-SPY2 to better match tumor biology to a modern treatment landscape(18).

Figure 5 demonstrates the four continuous immune signatures’ association with response to PGM therapy. Figure 5A shows all the immune signatures (4/4) significantly associated with response to PGM in the population as a whole; and of these Bcells sig, STAT1_sig, and CK12 were statistically significantly associated with response in the HR+HER2− subset. The Bcells sig was associated with improved response in PGM treated patients, but not in patients treated with control (Figure 5B and C). None of these signatures significantly associated with PGM response in the TN subset. Considering the dichotomous Immune marker [Supplementary data file, (18)], there were numerically higher numbers of PGM responders in Immune+ compared to Immune− in the population as a whole (Immune+: 32% pCR; Immune-: 16% pCR) and within HR/HER2 subsets; though these differences did not reach statistical significance (Figures D–F). In contrast, there were no statistically significant associations between immune signatures and response in the control arm. These results corroborate the immune-related results seen in the phosphoprotein data in the population as a whole and in HR+HER2−; and suggest that immune participation in TN response to PGM observed on the phospho-protein level (p-STAT1) may not be observed on the total protein or mRNA levels. Thus, immune activation appears to be the dominant “missing” response-related signal outside the known mechanism of action of IGF-1R inhibition.

Figure 5 –

Association of immune signatures and response to PGM A. Heatmap showing relationship between four immune signatures (Bcells, STAT1, CK12, and Dendritic Cells) and treatment groups.

B. Correlation between BCells signature in PGM (ganitumab) and control chemotherapy and pCR patients

C. Correlation between Bcells signature in PGM (ganitumab) and control chemotherapy and pCR in HR+, HER2− tumors

D. Relationship between PGM (ganitumab) in all patients by immune signature

E. Relationship between PGM (ganitumab) in HER+, HER2− tumors by immune signature

F. Relationship between PGM (ganitumab) in triple negative (TN) tumors by immune signature

Discussion

Decades of research pointed to a role for IGF-1R activation in cancer biology (25), yet drugs designed to inhibit this pathway in cancer were failures. While numerous reasons for the failure of these drugs have been discussed (2), one important reason has been a lack of biomarkers in predicting response. Our previous results showed that IGF-1R gene activation measured by mRNA profiling correlated with improved response to both PGM and control therapy (13) indicating a lack os specificity for PGM response and potentially identifying more chemotherapy responsive tumors. However, when functional protein biomarkers of IGF-1R activation were measured, treatment-specific predictive associations were found. Tyrosine phosphorylation of the IGF-1R intracellular kinase domain is required for downstream signaling (26) and we studied functional biomarkers such as phospho-proteins directly related to the mechanism of action of ganitumab as a way to discover predictive biomarkers. Since autophosphorylation is the obligate first step in receptor activation (27), we first studied phospho-IGF-1R/IR. Given the high degree of homology between IGF-1R and IR, it is not possible to distinguish between the activation states of the two receptors. Further complicating this analysis is that IGF-1R and IR can form hybrid receptors and hybrid receptor activation will be detected in our assays.

In HR+/HER2− breast cancers, we found that higher levels of phospho-IGF-1R were associated with increased pCR rates after PGM and AC therapy. IGF-1R and ER signaling have extensive signaling crosstalk (28) and IGF-1R activation alone stimulates ER-transcriptional pathways (29) thus, inhibiting activated IGF-1R may further reduce survival signaling mediated by ER and make luminal B breast cancers more susceptible to cytotoxic chemotherapy like paclitaxel. Notably, other studies have suggested that extracellular and circulating factors, such as the IGF binding proteins, may influence ligand availability and downstream receptor activation (7,30). It also must be acknowledged that the phospho-antibody we used to detect IGF-1R also detects phosphorylated insulin receptor (pIR). While ganitumab does not bind IR, IGF-1R and IR can form hybrid receptors composed of one chain from each receptor and anti-IGF-1R moAbs inhibit hybrid receptor activation (31) with preclinical data suggesting IR may also be inhibited by anti-IGF-1R moAbs. Several other IGF-1R downstream signaling molecules were nominally associated with pCR. Importantly, phospho-IGF-1R was not associated with increased pCR in the control group. Taken together, our data suggest that pre-treatment levels phospho-IGF-1R may be a predictive biomarker for PGM in HR+ breast cancers.

In TNBC, there was not a clear association between phospho-IGF-1R and pCR. There was an inverse relationship between phosphorylated p27 and pCR in this subgroup, but none of the IGF signaling components were associated with pCR in PGM treated patients. p27 inhibits cyclin-CDK to arrest progression through the cell cycle but also has additional effects on several transcriptional targets (32). p27 is phosphorylated by the PI3K signaling system and its proteolysis is mediated by several signaling pathways (33). Lower levels of total p27 are associated with diminished overall survival in breast cancer, but phospho-p27 has not been studied in detail. It is possible that p27 phosphorylation identifies a subset of TNBC with differential responses to IGF-1R inhibition and deserves further study.

Because STAT1 is involved in immune function, we explored mRNA signatures of immune activation and found that in HR+, HER2− tumors that achieved a pCR after PGM also had higher expression of the immune signatures. This relationship was not seen in patients treated with control chemotherapy arguing for a role for IGF-1R in immune response. The suggestion that suppression of IGF signaling might alter the immunogenicity of cancers was suggested by work in rat glioma cells showing that downregulation of the IGF-1 ligand increased immunogenicity of the tumor (34) with upregulation of MHC-1 and B-7(35). IGF signaling also regulates components of the immune system. IGF-I stimulates murine regulatory T cells (Treg) and this function increases their immunosuppressive effects (36). Notably, suppression of IGF-1R reverses these effects suggesting that this receptor’s function in the immune microenvironment may be immunosuppressive. Taken together, our results support the idea that ganitumab could affect immune recognition of cancer cells and this effect is most pronounced in tumors with evidence of baseline immune activation. Combination of an IGF-1R inhibitory strategy with an immune checkpoint inhibitor could have activity.

This study is limited by its small population size. However, our findings show that subsets of breast cancers with evidence of enhanced IGF-1R phosphorylation have improved response rates to ganitumab. This observation was seen primarily in HR+ cells and is consistent with the idea that treatment naïve HR+ cells have elevated levels of the receptor and downstream signaling pathways (37). These findings are consistent with the initial concept of targeting IGF-1R, inhibition of signaling would result in inhibition of tumor growth. Additional drugs targeting signaling pathways downstream of IGF-1R activation, such as CDK4/6 and PI3K pathway inhibitors, have been developed. Perhaps a combination of drugs, other than cytotoxic chemotherapy, might yield a higher rate of response especially in these HR+, high risk tumors. Our studies are the first to directly measure IGF-1R/IR activation in treatment-naïve patients who received anti-IGF-1R therapy. Our results also suggest the immune activation status of a subset of TNBCs could identify another subset of breast cancers susceptible to this approach. Since IGF-1R activation requires interaction with its cognate ligands (26), other extracellular proteins (IGF ligands and IGF binding proteins) have also been found to be associated with improved response to ganitumab ((7,38,39). Thus, more detailed analysis of IGF-1R signaling activation has yielded putative biomarkers for this class of inhibitors. Further improvement in the development of predictive biomarkers for IGF-1R dependent tumors could depend on measurement of circulating ligand, binding proteins, and direct measurement of receptor phosphorylation status.

Translational Relevance.

Inhibitors of the type I insulin-like growth factor receptor (IGF-1R) have been tested in multiple clinical trials. Drugs targeting this receptor have failed to provide benefit in phase 3 clinical trials and no biomarkers have been identified that correlate with benefit. We studied the IGF-1R monoclonal antibody ganitumab in the neoadjuvant phase 2 I-SPY 2 trial in combination with paclitaxel and metformin. We find that phosphoprotein analysis identified predictive biomarkers of response in a subtype specific way including phospho-IGF-1R, p27/Kip1, and pSTAT1. These data show that interrogation of activated signaling pathways yields more predictive information for an IGF-1R inhibitor compared to other genomic/transcriptomic approaches.