Abstract P3-10-14: LIV-1 expression in primary breast cancers in the I-SPY 2 TRIAL

Background: LIV-1 is an estrogen-inducible gene that has been implicated in epidermal-to-mesenchymal transition (EMT) in preclinical models of progression and metastasis. Its expression is associated with node-positivity in breast cancer; and has been detected in a variety of cancer types, including estrogen receptor positive breast cancers. SGN-LIV1A is a novel antibody drug conjugate targeting LIV-1 that is currently being evaluated in the I-SPY 2 TRIAL. In this pilot study, we evaluated LIV-1 levels by IHC within HR/HER2/MammaPrint (MP) defined subtypes among patients screening for the I-SPY 2 TRIAL and its correlation to microarray assessed LIV-1 expression levels.

Method: In a pilot study, LIV-1 IHC staining was performed by Quest Diagnostics on the pre-treatment samples of 38 patients screening for the I-SPY 2 TRIAL. Pre-treatment expression data generated on a custom Agilent 44K platform was also available. We summarized the LIV-1 H-Scores and percent (%)-positivity across the population and within HR/HER2/MP subtypes; and we assessed the Pearson correlation between LIV-1 H-Score and LIV-1 gene expression levels. In addition, we compared the pre-treatment LIV-1 expression levels within HR/HER2/MP subtypes across I-SPY 2 TRIAL patients from completed arms and their relevant controls (n=989) using ANOVA and post-hoc Tukey tests. Our statistics are descriptive rather than inferential; and does not take into account multiplicities of other biomarkers outside of this study.

Results: Of the 38 patients evaluated, 37 have LIV-1 %-positivity > 0; and 18 (47%) have 100% LIV1 positivity. The median LIV-1 H-Score is 200; and 89% of patients (34/38) have moderate/high LIV-1 staining (with H-Score≥100). Of the 34 patients who proceeded onto the trial (and have known HR/HER2/MP status), 9 are triple negative, 19 are HR+HER2-, and 6 are HER2+. Due to our small sample size, we did not further subset the triple negative and HER2+ cases; but within the HR+HER2- patients, 10 are MP1 compared to 9 who are MP2 class. LIV1 H-Score appears highest within the HR+HER2-MP1 cases (median: 290), followed by the HER2+ (median: 216), then the HR+HER2-/MP2 (median: 155), and the TN (median: 120) subtype. LIV1 H-score is significantly correlated with LIV-1 mRNA expression levels (Rp=0.79, p<0.0001). Consistent with these observations, LIV-1 pre-treatment expression levels are significantly higher in the HR+HER2-MP1 group relative to all other HR/HER2/MP defined subtypes (Tukey HSD p < 0.0001) across the I-SPY 2 TRIAL population. The HR+HER2+MP1 group also have high LIV-1 expression levels.

Conclusion: Our result suggest that although LIV-1 expression differs by subtype, it is expressed at a moderate/high level in the majority of patients. The good correlation between IHC and array-based LIV-1 expression levels enables us to leverage the entire existing I-SPY 2 dataset and confirm the high rates of LIV-1 expression across the I-SPY 2 population. Further studies to evaluate LIV-1 expression as a biomarker of response to LIV-1 targeting therapies for the neoadjuvant treatment of breast cancer are warranted and ongoing in I-SPY 2.

Citation Format: Yau C, Brown-Swigart L, Asare S, I-SPY 2 TRIAL Consortium, Esserman L, van' t Veer L, Beckwith H, Forero A, Rugo H. LIV-1 expression in primary breast cancers in the I-SPY 2 TRIAL [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P3-10-14.

Abstract P3-10-02: Identifying breast cancer molecular phenotypes to predict response in a modern treatment landscape: Lessons from ˜1000 patients across 10 arms of the I-SPY 2 TRIAL

Background: The explosion in new treatment options targeting immune checkpoints, HER signaling, DNA repair deficiency, AKT, and other pathways calls for updated breast cancer subtypes beyond HR and HER2 status to predict which patients will respond to which treatments. Here we leverage the I-SPY 2 TRIAL biomarker program over the past 8 years across 10 treatment arms to elucidate a minimal set of biomarkers that may improve response prediction in a modern treatment context, and to investigate which new patient phenotypes are identified by these response-predictive biomarkers.

Methods: 986 patients were considered in this analysis. Treatments included paclitaxel alone (or with trastuzumab (H) in HER2+) or combined with investigational agents: veliparib/carboplatin (VC); neratinib; MK2206; ganitumab; ganetespib; AMG386; TDM1/pertuzumab (P); H/P; and pembrolizumab (Pembro). 24 prospectively defined, mechanism-of-action and pathway-based expression and phospho-protein signatures/biomarkers assayed from pre-treatment biopsies were previously found to be predictive in a particular agent/arm in pre-specified analysis. Here we evaluate these biomarkers in all patients. We assessed association between each biomarker and response in the population as a whole and within each arm and HR/HER2 subtype using a logistic model. To identify optimal dichotomizing thresholds for select biomarkers, 2-fold cross-validation was repeated 500 times. Our analysis is exploratory and does not adjust for multiplicities.

Results: Our initial set of 24 predictive biomarkers reflects DNA repair deficiency (n=2), immune activation (n=7), ER signaling (n=2), HER2 signaling (n=4), proliferation (n=2), phospho-activation of AKT/mTOR (n=2), and ANG/TIE2 (n=1) pathways, among others. Biomarkers reflecting similar biology are correlated and cluster together. We make use of this correlation structure to reduce the dimensionality of the biomarker set to five predictive signals: proliferation, DNA repair deficiency (DRD), immune-engaged (Immune+), luminal/ER (lum), and HER2-activated. These biomarkers, when dichotomized, identify patient groups with differential predicted sensitivities to I-SPY 2 agents and are present at different proportions within receptor subtypes. For instance, in the HER2- subset, Immune+/DRD+ patients are predicted sensitive to both VC and Pembro, and account for 39% of TN, but only 12% of HR+HER2-. On the other end of the spectrum, only 17% of TN are Immune-/DRD-, compared to the majority (56%) of HR+HER2-. There are also subsets of patients positive for only one marker. For the HER2+ subset, 67% are HER2-activated+, and 25% lum+; of these HER2-activated+ patients are more likely to be Immune+ (44%), vs 23% in lum+. HER2-activated+/Immune+ patients have higher predicted sensitivity to HER2-targeted agents than lum+ or Immune- patients.

In all, these molecular phenotypes predict sensitivity to one or more I-SPY 2 investigational agents for 75% of the ˜ 1000 patients.

Conclusion: Molecular phenotypes reflecting proliferation, immune engagement, HER2-activation, luminal/ER-signaling, and DNA repair deficiency may provide a roadmap to guide treatment prioritization for emerging therapeutics.

Citation Format: Wolf DM, Yau C, Wulfkuhle J, Petricoin E, Campbell M, Brown-Swigart L, Hirst G, Asare S, Zhu Z, Lee EP, Delson A, Pohlmann P, I-SPY 2 TRIAL Consortium, Hylton N, Liu MC, Symmans F, DeMichele A, Yee D, Berry D, Esserman L, van 't Veer L. Identifying breast cancer molecular phenotypes to predict response in a modern treatment landscape: Lessons from ˜1000 patients across 10 arms of the I-SPY 2 TRIAL [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P3-10-02.

Bayesian statistical learning for big data biology

Bayesian statistical learning provides a coherent probabilistic framework for modelling uncertainty in systems. This review describes the theoretical foundations underlying Bayesian statistics and outlines the computational frameworks for implementing Bayesian inference in practice. We then describe the use of Bayesian learning in single-cell biology for the analysis of high-dimensional, large data sets.

A descriptive marker gene approach to single-cell pseudotime inference

Motivation

Pseudotime estimation from single-cell gene expression data allows the recovery of temporal information from otherwise static profiles of individual cells. Conventional pseudotime inference methods emphasize an unsupervised transcriptome-wide approach and use retrospective analysis to evaluate the behaviour of individual genes. However, the resulting trajectories can only be understood in terms of abstract geometric structures and not in terms of interpretable models of gene behaviour.

Results

Here we introduce an orthogonal Bayesian approach termed 'Ouija' that learns pseudotimes from a small set of marker genes that might ordinarily be used to retrospectively confirm the accuracy of unsupervised pseudotime algorithms. Crucially, we model these genes in terms of switch-like or transient behaviour along the trajectory, allowing us to understand why the pseudotimes have been inferred and learn informative parameters about the behaviour of each gene. Since each gene is associated with a switch or peak time the genes are effectively ordered along with the cells, allowing each part of the trajectory to be understood in terms of the behaviour of certain genes. We demonstrate that this small panel of marker genes can recover pseudotimes that are consistent with those obtained using the entire transcriptome. Furthermore, we show that our method can detect differences in the regulation timings between two genes and identify 'metastable' states-discrete cell types along the continuous trajectories-that recapitulate known cell types.

Availability and implementation

An open source implementation is available as an R package at http://www.github.com/kieranrcampbell/ouija and as a Python/TensorFlow package at http://www.github.com/kieranrcampbell/ouijaflow.

Supplementary information

Supplementary data are available at Bioinformatics online.

Abstract 2182: DigiPico: A whole-genome sequencing approach to investigate microscopic residual chemotherapy resistance disease in ovarian cancer

Microscopic residual chemotherapy resistance disease (MRCD) following cancer therapy predisposes to recurrence. Genome analysis of MRCDs could, therefore, give insights into evolutionary mechanisms that might govern this process. However, to date, this has largely been restricted to studies utilizing targeted sequencing of known somatic mutations due to the challenges of whole-genome sequencing of subnanogram quantities of DNA. A previous report showed that obtaining accurate whole-genome sequencing is possible from as little as 10 to 20 cells with the added advantage of obtaining reconstructed long fragment reads. Adopting this technology, we were able to optimize a procedure to accurately obtain high-quality whole-genome sequencing data from minimally available clinical samples such as MRCDs, termed DigiPico. Using DigiPico, we were able to perform sequencing on several post-chemotherapy MRCDs and single tumor islets from a high-grade serous ovarian cancer (HGSOC) tumor. The data were found to be highly accurate, with over 96% of the single-nucleotide polymorphisms (SNPs) being detected in DigiPico sequencing results despite the extremely low amount of material available. Moreover, we found that DigiPico can be used for calculation of allele fractions with a consistency rate of up to 84% as opposed to a consistency rate of only 47% for when a standard multiple displacement amplification of similar amount of DNA is used during library preparation. The high-quality allele fraction information from DigiPico was then used to predict the copy number variation (CNV) in these clinical samples. Here, we showed the power of DigiPico for accurate SNP calling and CNV prediction from post-chemotherapy MRCDs and single tumur islets of a HGSOC tumor, the latter of which is extremely important for studying HGSOC as these tumors often show a great degree of structural alterations. This information can now be used for better understanding of how MRCDs might be able to remain unharmed during the course of chemotherapy and initiate recurrence after completion of treatment, which in turn can allow us to design better treatment strategies for HGSOC patients.

Citation Format: Mohammad KaramiNejadRanjbar, Donatien Chedom Fotso, Yuhao Zheng, Christopher Yau, Ahmed Ahmed. DigiPico: A whole-genome sequencing approach to investigate microscopic residual chemotherapy resistance disease in ovarian cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 2182.

The Hamming Ball Sampler

We introduce the Hamming ball sampler, a novel Markov chain Monte Carlo algorithm, for efficient inference in statistical models involving high-dimensional discrete state spaces. The sampling scheme uses an auxiliary variable construction that adaptively truncates the model space allowing iterative exploration of the full model space. The approach generalizes conventional Gibbs sampling schemes for discrete spaces and provides an intuitive means for user-controlled balance between statistical efficiency and computational tractability. We illustrate the generic utility of our sampling algorithm through application to a range of statistical models. Supplementary materials for this article are available online.

Abstract PD6-08: Analysis of immune infiltrates (assessed via multiplex fluorescence immunohistochemistry) and immune gene expression signatures as predictors of response to the checkpoint inhibitor pembrolizumab in the neoadjuvant I-SPY 2 trial

Background: Pembrolizumab (Pembro), an anti-PD-1 immune checkpoint inhibitor, has been approved for the treatment of a variety of cancers including melanoma, non-small cell lung cancer, head and neck squamous cell carcinoma, and urothelial carcinoma. Pembro was recently evaluated in HER2- breast cancer patients in the neoadjuvant I-SPY 2 TRIAL and graduated in the triple negative (TN), HR+HER2-, and HER2- signatures. HER2- patients were randomized to receive Pembro+paclitaxel followed by doxorubicin/cyclophosphamide (P+T -> AC) vs. T -> AC. We and others have shown that TN breast cancers tend to have high numbers of immune infiltrates, including T cells and tumor associated macrophages (TAMs). We evaluated expression signatures representing 14 immune cell types (TILs, T cells, CD8 T cells, exhausted T cells, Th1, Tregs, cytotoxic cells, NK, NK CD56dim, dendritic cells, mast cells, B cells, macrophages, and neutrophils) as specific predictors of response to Pembro.

Methods: Data from 248 patients (Pembro: 69; controls: 179) were available. Pre-treatment biopsies were assayed using Agilent gene expression arrays. Signature scores are calculated by averaging cell type specific genes. All I-SPY 2 qualifying biomarker analyses follow a pre-specified analysis plan. We used logistic modeling to assess biomarker performance. A biomarker is considered a specific predictor of Pembro response if it associates with response in the Pembro arm but not the control arm, and if the biomarker x treatment interaction is significant (likelihood ratio test, p<0.05). This analysis is also performed adjusting for HR status as covariates, and within receptor subsets. For successful biomarkers, we use Bayesian modeling to estimate the pCR rates of 'predicted sensitive' patients in each arm. Our statistics are descriptive rather than inferential and do not adjust for multiplicities of other biomarkers outside this study.

Results: 10 out of the 14 cell-type signatures tested are associated with response in the Pembro arm. Higher expression levels of 9 of these cell-type signatures are associated with higher pCR rates (T cells, exhausted T cells, Th1, cytotoxic cells, NK, NK CD56dim, dendritic cells, B cells, and macrophages), whereas higher mast cell signature expression is associated with non-pCR. Interestingly, many of these same signatures also associate or trend towards association with response in the control arm; and in a model adjusting for HR status, only 3 of these signatures (Th1, B cells and dendritic cells) show significant interaction with treatment. Within the whole population and the TN subtype, the dendritic cell signature is the strongest predictor of specific response to Pembro (OR/1SD: 4.04 and 4.4, LR p < 0.001 overall and in TN). Although other immune signatures (T cells, exhausted T cells, NK, and macrophages) also associate with response in the Pembro arm in the TN subtype, only the dendritic cell and Th1 signatures have a significant interaction with treatment. In contrast, in the HR+HER2- subtype, only 3 signatures (Th1, B cells, and mast cells) associate with response to Pembro; but none of these signatures have significant interaction with treatment. Of note, in both the Pembro and control arms, HR+HER2- patients with higher average mast cell marker expression have lower pCR rates (OR/1SD: 0.33 and 0.51, LRp: 0.006 and 0.04 in Pembro and control arm).

Conclusion: As expected, multiple immune cell expression signatures are predictive of response in the Pembro arm; but only dendritic cells and Th1 cells are specific to Pembro in both the population as a whole and the TN subtype. Interestingly, the presence of mast cells may impede response, especially in HR+HER2- patients. Correlation of these signatures with multiplex-IF immune markers is pending.

Citation Format: Campbell M, Yau C, Borowsky A, Vandenberg S, Wolf D, Rimm D, Nanda R, Liu M, Brown-Swigart L, Hirst G, Asare S, van't Veer L, Yee D, DeMichele A, Berry D, Esserman L. Analysis of immune infiltrates (assessed via multiplex fluorescence immunohistochemistry) and immune gene expression signatures as predictors of response to the checkpoint inhibitor pembrolizumab in the neoadjuvant I-SPY 2 trial [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr PD6-08.

Abstract PD2-14: Participation in a personalized breast cancer screening trial does not increase anxiety at baseline

Introduction: The purpose of this study is to examine whether participation in a personalized screening trial is associated with anxiety or breast cancer worry. The Patient Centered Outcomes Research Institute recently funded WISDOM (Women Informed to Screen Depending On Measures of risk), which is a randomized trial that tests the safety and efficacy of basing starting age, stopping age, frequency and modality of breast cancer screening on individual risk (Clinical Trials Identifier NCT02620852).

Methods: In WISDOM, participants can be randomized to annual screening or personalized screening arm, or self-select an arm an observational cohort. This interim analysis examined the first 1817 participants to determine if the personalized risk arm is acceptable and to explore whether baseline anxiety was associated with study arm. For acceptability our target was to have >60% of participants agree to randomization. Participants completed questions about their Risk Perception, the PROMIS Anxiety short form 8a (total scores 8-40 with higher scores indicating more anxiety), and Breast Cancer Risk Worry (BCRW) survey (total scores 5-20) with higher scores indicating more worry) at baseline and before they were given information on their personal risk or study assignment. For the purposes of these analyses, we defined high anxiety to be the percentage of participants scoring =>22 on the PROMIS and >8 on the BCRW.

Results: The participants were recruited from three sites (UCSD, UCSF, Sanford Health). Of the 1817 initial participants, 1643 completed the baseline questionnaire. Participants has a mean age of 57 years (SD 9). 15.8% felt their chances of developing breast cancer was high, 19.5% felt their chance of developing breast cancer was greater than the average women, and 56.6% felt their lifetime risk of developing breast cancer was >25. Risk perception was not significantly different between women who opted to be randomized versus the observational arm.

The majority of participants were willing to be randomly assigned to an arm (1071/1643, 65.1%). Of those who joined the observational cohort, the majority selected personalized risk arm (474/572, 82.9%). Overall, PROMIS anxiety scores were low at baseline (14.0 MEAN (SD 4.6)) as were the Breast Cancer Risk Worry scores (5.7 MEAN (SD 1.05)). Less than 8% of participants had PROMIS scores >22 and that did not vary across the randomized or observational groups (P=0.2)). About 2% of participants had a BCRW scores >8. Women who worried with breast cancer were more likely to select to be in the observational (3.5%) than randomized (1.7%) arm of the study (P=0.02).

Conclusions: For the women approached to participate in Wisdom, personalized screening was acceptable alternative to annual mammography. Participants in general overestimated their lifetime risk of breast cancer, had very low anxiety and low breast cancer worry. Those who were worried about breast cancer opted more often for the observational arm of the study to allow them to choose between the personalized versus annual arm. Future analyses will follow participants prospectively to determine adherence to assigned or selected arm, and whether anxiety changes after receipt of their personalized risk information.

Citation Format: Naeim A, Sepucha K, Wenger N, Eklund M, Annette S, Madlensky L, van't Veer L, Parker B, Yau C, Cink T, Anton-Culver H, Borowsky A, Petruse A, Sarrafan S, Stover-Fiscalini A, LaCroix A, Adduci K, Wisdom Advocate Partners, Laura E. Participation in a personalized breast cancer screening trial does not increase anxiety at baseline [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr PD2-14.

Abstract PD6-14: Analysis of DNA repair deficiency biomarkers as predictors of response to the PD1 inhibitor pembrolizumab: Results from the neoadjuvant I-SPY 2 trial for stage II-III high-risk breast cancer

Background: Pembrolizumab (P), an anti-PD-1 immune checkpoint inhibitor, has been approved for treatment of microsatellite instability-high and mismatch repair deficient cancers. In I-SPY 2, patients were randomized to receive standard chemotherapy alone or in combination with an experimental agent. P was one of the experimental agents evaluated in HER2- patients in I-SPY 2 and graduated in the TN, HR+HER2-, and HER2- signatures. We hypothesize that a combination of two signatures predicting response to veliparib/carboplatin therapy in I-SPY 2 [MammaPrint High2 (MP2)/PARPi7-high] and reflecting DNA damage repair deficiency, may also predict response to P. In addition, we also tested 9 gene expression signatures reflecting different aspects of DNA damage and repair: FA, MMR, BER, HR, TLS, NER, NHEJ, DR, and DNA damage sensing (DDS) pathways.

Methods: Data from 249 patients (P: 69 and controls: 180) were available. Pre-treatment biopsies were assayed using Agilent gene expression arrays. All I-SPY 2 qualifying biomarker analyses follow a pre-specified analysis plan. We used logistic modeling to assess biomarker performance. A biomarker is considered a specific predictor of P response if it associates with response in the P arm but not the control arm, and if the biomarker x treatment interaction is significant (likelihood ratio test, p<0.05). This analysis is also performed adjusting for HR status as a covariate, and within receptor subsets, sample size permitting. For successful biomarkers, we use Bayesian modeling to estimate the pCR rates of 'predicted sensitive' patients in each arm. Our statistics are descriptive rather than inferential and do not adjust for multiplicities of other biomarkers outside this study.

Results: MP2 status associates with pCR in P (OR=7.7; p=0.00021), but also to a lesser extent in the control arm (OR=2.4:p=0.045), with an OR ratio of 3.3 which trends toward significance, even after adjusting for HR status (LR p=0.083). A majority of TN patients are MP2; and TN/MP2 patients have an estimated pCR rate of 67% in P (vs. 23% in control). Although only ~30% of HR+HER2- patients were MP2, their estimated pCR rate in P is 61%, compared to 29% in unselected HR+/HER2- patients. PARPi7 predicted response in the P arm only in the HR+HER2- group (LR p= 0.025), but not in the population as a whole or the TN subtype. Combining MP2 and PARPi7 into MP2/PARPi7-high did not improve performance over MP2 as a single biomarker. Of the 9 DDR pathway signatures tested, both BER and DDS associate with pCR in P, but only DDS (which includes ATM, ATR, CHEK1-2) associates with pCR in the P arm (LR p=0.00029), and not the control arm (LR p=0.53), with a significant interaction with treatment (LR p=0.0064) that retains significance in a model adjusting for HR status. When dichotomized to optimize the biomarker x treatment interaction, the estimated pCR rate is 75% in P vs 18% in control, in the DDS+ subset.

Conclusion: In this small study, MP2 status and a DNA damage sensing pathway but not the PARPi7 or other repair pathways show promise as predictive biomarkers for immune checkpoint inhibition therapy in breast cancer.

Citation Format: Yau C, Wolf D, Brown-Swigart L, Hirst G, Sanil A, Singhrao R, I-SPY 2 TRIAL Investigators, Asare S, DeMichele A, Berry D, Esserman L, van 't Veer L, Nanda R, Liu M, Yee D. Analysis of DNA repair deficiency biomarkers as predictors of response to the PD1 inhibitor pembrolizumab: Results from the neoadjuvant I-SPY 2 trial for stage II-III high-risk breast cancer [abstract]. In: Proceedings of the 2017 San Antonio Breast Cancer Symposium; 2017 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2018;78(4 Suppl):Abstract nr PD6-14.

Tuning microtubule dynamics to enhance cancer therapy by modulating FER-mediated CRMP2 phosphorylation

Though used widely in cancer therapy, paclitaxel only elicits a response in a fraction of patients. A strong determinant of paclitaxel tumor response is the state of microtubule dynamic instability. However, whether the manipulation of this physiological process can be controlled to enhance paclitaxel response has not been tested. Here, we show a previously unrecognized role of the microtubule-associated protein CRMP2 in inducing microtubule bundling through its carboxy terminus. This activity is significantly decreased when the FER tyrosine kinase phosphorylates CRMP2 at Y479 and Y499. The crystal structures of wild-type CRMP2 and CRMP2-Y479E reveal how mimicking phosphorylation prevents tetramerization of CRMP2. Depletion of FER or reducing its catalytic activity using sub-therapeutic doses of inhibitors increases paclitaxel-induced microtubule stability and cytotoxicity in ovarian cancer cells and in vivo. This work provides a rationale for inhibiting FER-mediated CRMP2 phosphorylation to enhance paclitaxel on-target activity for cancer therapy.

Some anticancer drugs target cell microtubules inhibiting mitosis and cell division. Here, the authors show that CRMP2 induces microtubule bundling and that this activity is regulated by the FER kinase, thus providing a rationale for targeting FER in combination with microtubule-targeting drugs.

Associations between birth at, or after, 41 weeks gestation and perinatal encephalopathy: a cohort study

BACKGROUND

Preterm birth causes long-term problems, even for infants born 1 or 2 weeks early. However, less is known about infants born after their due date and over a quarter of infants are born over 1 week late, and many still remain undelivered after 2 weeks. The aim of this work is to quantify the risks of infants developing encephalopathy when birth occurs after the due date, and if other proposed risk factors modify this relationship.

METHODS

The dataset contain information on 4 036 346 infants born in Sweden between 1973 and 2012. Exposure was defined as birth 7, or more, days after the infants' due date. The primary outcome was the development of neonatal encephalopathy (defined as seizures, encephalopathy or brain injury caused by asphyxia or with unspecified cause). Covariates were selected as presumed confounders a priori.

RESULTS

28.4% infants were born 1 or more weeks after their due date. An infant's risk of being born with encephalopathy was higher in the post 41 weeks group in the unadjusted (OR 1.40 (95% CI 1.32 to 1.49)) and final model (OR 1.38 (95% CI 1.29 to 1.47)), with the relative odds of encephalopathy increasing by an estimated 20% per week after the due date, and modified by maternal age (P=0.022).

CONCLUSIONS

Singleton infants born at, or after, 41 weeks gestation have lower Apgar scores and higher risk of developing encephalopathy in the newborn period, and the association appeared more marked in older mothers. These data could be useful if provided to women as part of their decision-making.

Separation of Dual Oxidase 2 and Lactoperoxidase Expression in Intestinal Crypts and Species Differences May Limit Hydrogen Peroxide Scavenging During Mucosal Healing in Mice and Humans

BACKGROUND

DUOX2 and DUOXA2 form the predominant H2O2-producing system in human colorectal mucosa. Inflammation, hypoxia, and 5-aminosalicylic acid increase H2O2 production, supporting innate defense and mucosal healing. Thiocyanate reacts with H2O2 in the presence of lactoperoxidase (LPO) to form hypothiocyanate (OSCN-), which acts as a biocide and H2O2 scavenging system to reduce damage during inflammation. We aimed to discover the organization of Duox2, Duoxa2, and Lpo expression in colonic crypts of Lieberkühn (intestinal glands) of mice and how distributions respond to dextran sodium sulfate (DSS)-induced colitis and subsequent mucosal regeneration.

METHODS

We studied tissue from DSS-exposed mice and human biopsies using in situ hybridization, reverse transcription quantitative polymerase chain reaction, and cDNA microarray analysis.

RESULTS

Duox2 mRNA expression was mostly in the upper crypt quintile while Duoxa2 was more apically focused. Most Lpo mRNA was in the basal quintile, where stem cells reside. Duox2 and Duoxa2 mRNA were increased during the induction and resolution of DSS colitis, while Lpo expression did not increase during the acute phase. Patterns of Lpo expression differed from Duox2 in normal, inflamed, and regenerative mouse crypts (P < 0.001). We found no evidence of LPO expression in the human gut.

CONCLUSIONS

The spatial and temporal separation of H2O2-consuming and -producing enzymes enables a thiocyanate- H2O2 "scavenging" system in murine intestinal crypts to protect the stem/proliferative zones from DNA damage, while still supporting higher H2O2 concentrations apically to aid mucosal healing. The absence of LPO expression in the human gut suggests an alternative mechanism or less protection from DNA damage during H2O2-driven mucosal healing.

Colorectal Cancer Cell Line Proteomes Are Representative of Primary Tumors and Predict Drug Sensitivity

BACKGROUND AND AIMS

Proteomics holds promise for individualizing cancer treatment. We analyzed to what extent the proteomic landscape of human colorectal cancer (CRC) is maintained in established CRC cell lines and the utility of proteomics for predicting therapeutic responses.

METHODS

Proteomic and transcriptomic analyses were performed on 44 CRC cell lines, compared against primary CRCs (n=95) and normal tissues (n=60), and integrated with genomic and drug sensitivity data.

RESULTS

Cell lines mirrored the proteomic aberrations of primary tumors, in particular for intrinsic programs. Tumor relationships of protein expression with DNA copy number aberrations and signatures of post-transcriptional regulation were recapitulated in cell lines. The 5 proteomic subtypes previously identified in tumors were represented among cell lines. Nonetheless, systematic differences between cell line and tumor proteomes were apparent, attributable to stroma, extrinsic signaling, and growth conditions. Contribution of tumor stroma obscured signatures of DNA mismatch repair identified in cell lines with a hypermutation phenotype. Global proteomic data showed improved utility for predicting both known drug-target relationships and overall drug sensitivity as compared with genomic or transcriptomic measurements. Inhibition of targetable proteins associated with drug responses further identified corresponding synergistic or antagonistic drug combinations. Our data provide evidence for CRC proteomic subtype-specific drug responses.

CONCLUSIONS

Proteomes of established CRC cell line are representative of primary tumors. Proteomic data tend to exhibit improved prediction of drug sensitivity as compared with genomic and transcriptomic profiles. Our integrative proteogenomic analysis highlights the potential of proteome profiling to inform personalized cancer medicine.

The Influence of the Microbiome on Type 1 Diabetes

Type 1 diabetes (T1D) is characterized by the autoimmune destruction of pancreatic β cells. The rapid rise in T1D incidence during the past 50 y suggests environmental factors contribute to the disease. The trillion symbiotic microorganisms inhabiting the mammalian gastrointestinal tract (i.e., the microbiota) influence numerous aspects of host physiology. In this study we review the evidence linking perturbations of the gut microbiome to pancreatic autoimmunity. We discuss data from rodent models demonstrating the essential role of the gut microbiota on the development and function of the host's mucosal and systemic immune systems. Furthermore, we review findings from human longitudinal cohort studies examining the influence of environmental and lifestyle factors on microbiota composition and pancreatic autoimmunity. Taken together, these data underscore the requirement for mechanistic studies to identify bacterial components and metabolites interacting with the innate and adaptive immune system, which would set the basis for preventative or therapeutic strategies in T1D.

A pan-cancer genome-wide analysis reveals tumour dependencies by induction of nonsense-mediated decay

Nonsense-mediated decay (NMD) eliminates transcripts with premature termination codons. Although NMD-induced loss-of-function has been shown to contribute to the genesis of particular cancers, its global functional consequence in tumours has not been characterized. Here we develop an algorithm to predict NMD and apply it on somatic mutations reported in The Cancer Genome Atlas. We identify more than 73 K mutations that are predicted to elicit NMD (NMD-elicit). NMD-elicit mutations in tumour suppressor genes (TSGs) are associated with significant reduction in gene expression. We discover cancer-specific NMD-elicit signatures in TSGs and cancer-associated genes. Our analysis reveals a previously unrecognized dependence of hypermutated tumours on hypofunction of genes that are involved in chromatin remodelling and translation. Half of hypermutated stomach adenocarcinomas are associated with NMD-elicit mutations of the translation initiators LARP4B and EIF5B. Our results unravel strong therapeutic opportunities by targeting tumour dependencies on NMD-elicit mutations.

switchde: inference of switch-like differential expression along single-cell trajectories

Motivation

Pseudotime analyses of single-cell RNA-seq data have become increasingly common. Typically, a latent trajectory corresponding to a biological process of interest—such as differentiation or cell cycle—is discovered. However, relatively little attention has been paid to modelling the differential expression of genes along such trajectories.

Results

We present switchde, a statistical framework and accompanying R package for identifying switch-like differential expression of genes along pseudotemporal trajectories. Our method includes fast model fitting that provides interpretable parameter estimates corresponding to how quickly a gene is up or down regulated as well as where in the trajectory such regulation occurs. It also reports a P-value in favour of rejecting a constant-expression model for switch-like differential expression and optionally models the zero-inflation prevalent in single-cell data.

Availability and Implementation

The R package switchde is available through the Bioconductor project at https://bioconductor.org/packages/switchde.

Supplementary information

Supplementary data are available at Bioinformatics online.

Probabilistic modeling of bifurcations in single-cell gene expression data using a Bayesian mixture of factor analyzers

Modeling bifurcations in single-cell transcriptomics data has become an increasingly popular field of research. Several methods have been proposed to infer bifurcation structure from such data, but all rely on heuristic non-probabilistic inference. Here we propose the first generative, fully probabilistic model for such inference based on a Bayesian hierarchical mixture of factor analyzers. Our model exhibits competitive performance on large datasets despite implementing full Markov-Chain Monte Carlo sampling, and its unique hierarchical prior structure enables automatic determination of genes driving the bifurcation process. We additionally propose an Empirical-Bayes like extension that deals with the high levels of zero-inflation in single-cell RNA-seq data and quantify when such models are useful. We apply or model to both real and simulated single-cell gene expression data and compare the results to existing pseudotime methods. Finally, we discuss both the merits and weaknesses of such a unified, probabilistic approach in the context practical bioinformatics analyses.

Abstract P6-11-04: The evaluation of ganitumab/metformin plus standard neoadjuvant therapy in high-risk breast cancer: Results from the I-SPY 2 trial

Background: I-SPY 2 is a multicenter, phase 2 trial using response-adaptive randomization within biomarker subtypes to evaluate novel agents when added to standard neoadjuvant therapy for women with high-risk stage II/III breast cancer - investigational agent(I) +paclitaxel(T) qwk, doxorubicin & cyclophosphamide(AC) q2-3 wk x 4 vs. T/AC (control arm). The primary endpoint is pathologic complete response (pCR) at surgery. The goal is to identify/graduate regimens that have ≥85% Bayesian predictive probability of success (statistical significance) in a 300-patient phase 3 neoadjuvant trial defined by hormone-receptor (HR) & HER2 status & MammaPrint (MP). Regimens may also leave the trial for futility (&lt; 10% probability of success) or following accrual of maximum sample size (10%&lt; probability of success &lt;85%). We report the results for experimental arm Ganitumab, a type I insulin-like growth factor receptor (IGF1R) inhibitor. IGF1R inhibitors are known to induce insulin resistance and all patients assigned to Ganitumab received metformin.

Methods: Women with tumors ≥2.5cm were eligible for screening. MP low/HR+ and HER2+ tumors were ineligible for randomization. Hemoglobin A1C≥ 8.0% were ineligible. MRI scans (baseline, 3 cycles after start of therapy, at completion of weekly T and prior to surgery) were used in a longitudinal statistical model to improve the efficiency of adaptive randomization. Ganitumab was given at 12mg/kg q2 weeks and metformin at 850mg PO BID, while receiving ganitumab. Analysis was intention to treat with patients who switched to non-protocol therapy counted as non-pCRs. Ganitumab/metformin was open only to HER2- patients, and eligible for graduation in 3 of 10 pre-defined signatures: HER2-, HR+HER2- and HR-HER2-.

Results: Ganitumab/metformin did not meet the criteria for graduation in the 3 signatures tested. When the maximum sample size was reached, accrual to this arm stopped. Ganitumab/metformin was assigned to 106 patients; there were 128 controls. We report probabilities of superiority for Ganitumab/metformin over control and Bayesian predictive probabilities of success in a neoadjuvant phase 3 trial equally randomized between Ganitumab/metformin and control, for each of the 3 biomarker signatures, using the final pathological response data from all patients. Safety data will be presented.

SignatureEstimated pCR Rate (95% probability interval)Probability Ganitumab/ Metformin Is Superior to ControlPredictive Probability of Success in Phase 3 Ganitumab/ Metformin N = 106Control N = 128  All HER2-22% (13%-31%)16% (10%-23%)89%33%HR+/HER2-14% (4%-24%)12% (4%-19%)66%21%HR-/HER2-32% (17%-46%)21% (11%-32%)91%51%

Conclusion: The I-SPY 2 adaptive randomization study estimates the probability that investigational regimens will be successful in a phase 3 neoadjuvant trial. The value of I-SPY 2 is to give insight about the performance of an investigational agent's likelihood of achieving pCR. For Ganitumab/metformin, no subtype came close to the efficacy threshold of 85% likelihood of success in phase 3, and this regimen does not appear to impact upfront reduction of tumor burden. Our data do not support its continued development for the neoadjuvant treatment of breast cancer.

Citation Format: Yee D, Paoloni M, van't Veer L, Sanil A, Yau C, Forero A, Chien AJ, Wallace AM, Moulder S, Albain KS, Kaplan HG, Elias AD, Haley BB, Boughey JC, Kemmer KA, Korde LA, Isaacs C, Minton S, Nanda R, DeMichele A, Lang JE, Buxton MB, Hylton NM, Symmans WF, Lyandres J, Hogarth M, Perlmutter J, Esserman LJ, Berry DA. The evaluation of ganitumab/metformin plus standard neoadjuvant therapy in high-risk breast cancer: Results from the I-SPY 2 trial [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P6-11-04.

Abstract P6-11-02: Efficacy of Hsp90 inhibitor ganetespib plus standard neoadjuvant therapy in high-risk breast cancer: Results from the I-SPY 2 trial

Background:Pathologic complete response(pCR) after neoadjuvant therapy is an established prognostic biomarker for high-risk breast cancer(BC). Improving pCR rates may identify new therapies that improve survival. I-SPY 2 uses response-adaptive randomization within biomarker subtypes to evaluate novel agents when added to standard neoadjuvant therapy for women with high-risk stage II/III breast cancer; the goal is to identify regimens that have ≥85% Bayesian predictive probability of success (statistical significance) in a 300-patient phase 3 neoadjuvant trial defined by hormone-receptor (HR), HER2 status and MammaPrint (MP). We report the results for Ganetespib, a selective inhibitor of Hsp90 that induces the degradation/deactivation of key drivers of tumor initiation, progression, angiogenesis, and metastasis.Ganetespib + taxanes previously have resulted in a superior therapeutic response compared to monotherapy in multiple solid tumor models including BC.

Methods:Women with tumors ≥2.5cm were eligible for screening and participation. MP low/HR+ tumors were ineligible for randomization. QTcF &gt;470msec and HbA1C &gt;8.0% were ineligible. MRI scans (baseline, +3 cycles, following weekly paclitaxel, T, and pre-surgery) were used in a longitudinal statistical model to improve the efficiency of adaptive randomization. Ganetespib was given with weekly T at 150 mg/m2 IV weekly (3 weeks on, 1 off). Patients were premedicated (dexamethasone 10mg and diphenhydramine HCl 25-50 mg, or therapeutic equivalents). Analysis was intention to treat with patients who switched to non-protocol therapy counted as non-pCRs. The Ganetespib regimen was open only to HER2- patients, and eligible for graduation in 3 of 10 pre-defined signatures: HER2-, HR+/HER2- and HR-/HER2-.

Results:Ganetespib did not meet the criteria for graduation in the 3 signatures tested. When the maximum sample size was reached, accrual stopped. Ganetespib was assigned to 93 patients; there were 140 controls. We report probabilities of superiority for Ganetespib over control and Bayesian predictive probabilities of success in a neoadjuvant phase 3 trial equally randomized between Ganetespib and control, for the 3 biomarker signatures, using the final pCR data from all patients. Safety data will be presented.

SignatureEstimated pCR Rate (95% probability interval)Probability Ganetespib Is Superior to ControlPredictive Probability of Ganetespib Success in a Phase 3 Trial Ganetespib N = 93Control N = 140  All HER2-26% (16%-37%)18% (8%-28%)91%47%HR+/HER2-15% (4%-27%)14% (4%-24%)60%19%HR-/HER2-38% (23%-53%)22% (9%-35%)96%72%

Conclusion:The I-SPY 2 adaptive randomization model efficiently evaluates investigational agents in the setting of neoadjuvant BC. The value of I-SPY 2 is that it provides insight as to the regimen's likelihood of success in a phase 3 neoadjuvant study. Although no signature reached the efficacy threshold of 85% likelihood of success in phase 3, we observed the most impact in HR-/HER2- patients, with a 16% improvement in pCR rate. While our data do not support the continued development of Ganetespib alone for neoadjuvant BC, combinations with Ganetespib, which could potentiate its effect, may be worth pursuing in I-SPY 2 or similar trials.

Citation Format: Forero A, Yee D, Buxton MB, Symmans WF, Chien AJ, Boughey JC, Elias AD, DeMichele A, Moulder S, Minton S, Kaplan HG, Albain KS, Wallace AM, Haley BB, Isaacs C, Korde LA, Nanda R, Lang JE, Kemmer KA, Hylton NM, Paoloni M, van't Veer L, Lyandres J, Perlmutter J, Hogarth M, Yau C, Sanil A, Berry DA, Esserman LJ. Efficacy of Hsp90 inhibitor ganetespib plus standard neoadjuvant therapy in high-risk breast cancer: Results from the I-SPY 2 trial [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P6-11-02.

Abstract PD7-02: Identification of breast cancers with an indolent disease course: 70 gene indolent threshold validation in a Swedish randomized trial of tamoxifen vs. not, with 20 year outcomes

Importance: The frequency of cancers with indolent behavior has increased with screening. We asked whether an ultralow risk threshold on a multigene classifier would identify women whose cancers had an indolent course over 2 decades of follow-up, and which features were most predictive of outcome.

Methods: An ultralow risk threshold of the FDA-cleared MammaPrint 70-gene expression score was set to predict long-term absence of breast cancer-specific mortality in the absence of systemic therapy. The Stockholm Tamoxifen (STO) trial conducted between 1976 and 1990, where postmenopausal women with clinically detected node-negative breast cancers &lt;3cm were randomized to receive tamoxifen versus not, was used for validation. Immunohistochemistry markers (n=727) and Agilent microarrays for MammaPrint risk scoring (n=652) were performed from formalin-fixed paraffin-embedded primary tumor blocks. Recursive partitioning was performed using the rpart package in R to select variables and construct a regression tree that best predicts 20-year breast cancer specific survival. Input variables include: age, period of diagnosis, grade, hormone receptor status, HER2 and Ki69 status, 70-gene risk categories (high, low but not ultra, or ultralow), treatment arm and tumor size; and cross-validation was used to select the final regression tree model.

Results: In this trial conducted in the era before mammographic screening, 58% and 42% were MammaPrint low and high risk, respectively, while 15% were above the ultralow threshold. In the tamoxifen treated arm, women with tumors above the ultralow threshold had no deaths at 15 years and their 20-year disease-specific survival rates of 97%; whereas if untreated, their survival rates were 94%. Recursive partitioning identified the ultralow threshold classification as the first primary split in the model. Once the indolent tumors were partitioned out, among women with tumors below the ultralow threshold, the next most prognostic feature was size, where patients with tumors &gt;20mm have worse breast cancer specific survival. The last split in the model divides the patients with tumors ≤20mm into 70-gene high risk vs low but not ultralow risk groups.

Conclusions and Relevance: A threshold of the 70-gene MammaPrint assay can identify patients with indolent disease whose long-term risk of death from breast cancer after surgery alone is exceedingly low. This threshold emerged as the most prognostic variable, followed by tumor size, and mammaprint high vs. low but not ultralow in our recursive partitioning analysis. This suggests that finding indolent tumors early at a small size may not have much impact on patient outcome. Determining the presence of an ultralow risk breast cancer may prevent overtreatment. Conversely, once the indolent tumors are taken out of consideration, both biology and size impact outcome, and finding these tumors at a small size is likely still important and supports screening in this postmenopausal node negative population.

Citation Format: Esserman LJ, Yau C, Thompson CK, van't Veer LJ, Borowsky AD, Hoadley KA, Tobin NP, Nordenskjöld B, Fornander T, Stål O, Benz CC, Lindström LS. Identification of breast cancers with an indolent disease course: 70 gene indolent threshold validation in a Swedish randomized trial of tamoxifen vs. not, with 20 year outcomes [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr PD7-02.

Abstract P2-05-03: Intra-tumor heterogeneity of the estrogen receptor predicts less benefit from tamoxifen therapy and poor long-term breast cancer patient survival – Retrospective analyses of the STO-3 randomized trial

Background

We and others have shown that the clinically used breast cancer markers alter their expression throughout tumor progression, influencing patient survival (Lindström et al, JCO 2012). What are the likely explanations to our findings? Here, we aimed to determine whether breast cancer intra-tumor heterogeneity of the estrogen receptor (ER) is a marker of tumor aggressiveness and benefit of tamoxifen therapy in a large randomized trial.

Material and methods

The Stockholm Tamoxifen (STO-3) trial enrolled postmenopausal lymph node negative breast cancer patients with a tumor size of less than 30 mm, between 1976 and 1990, to be randomized to receive adjuvant tamoxifen versus not. From the original randomized trial cohort approximately half of the patients (778 patients) had primary tumor formalin-fixed paraffin-embedded blocks available and were included in our study. No significant differences in age and period of diagnosis, type of surgery received, receptor status, tumor grade and size were observed between the treatment arms.

All tumor slides were immunostained in a central laboratory using the SP1 antibody. ER slides were scored by two independent breast cancer pathologists assessing the fraction of cancer cells for each ER intensity level (0, +1, +2 or +3) compared to established standards. The resulting distribution of ER stained tumor cells defines intra-tumor heterogeneity of ER (Rao's quadratic entropy (QE),Potts et al, Lab Invest 2012). Intra-tumor heterogeneity was categorized using the third tertile as cut-off for high heterogeneity (726 patients).

Analyses of long-term breast cancer specific survival (25 years) by intra-tumor heterogeneity of ER were performed using univariate Kaplan-Meier and multivariate Cox proportional hazard modeling adjusting for treatment arm, age and period of diagnoses, ER, progesterone receptor (PR), HER2, Ki-67, tumor grade, and tumor size. Further, a test of correlation was performed to investigate whether intra-tumor heterogeneity of ER was correlated to the percentage of ER positive cells, the H-Score or the Luminal A and B subtype (PAM50).

Results

In the univariate Kaplan-Meier analyses, a statistically significant difference in long-term survival by intra-tumor heterogeneity of ER was seen for all patients (log rank, P=0.018), tamoxifen treated arm (log rank, P=0.0033), but not untreated arm (log rank, P=0.19). However in the multivariate analysis, patients with high intra-tumor heterogeneity of ER in the treated arm as well as in the untreated arm had an almost two-fold increased long-term risk of fatal breast cancer disease as compared to patients with low or intermediate heterogeneity (Treated arm: HR, 2.06; 95% CI, 1.04-4.07 and Untreated arm: HR, 1.71; 95% CI, 1.01-2.87).

No significant correlation of intra-tumor heterogeneity to the tested variables was seen.

Conclusions

Patients with high intra-tumor heterogeneity of ER had less benefit from tamoxifen therapy and an increased long-term risk of fatal breast cancer disease. Our findings should be clinically relevant since therapy benefit was evaluated in a randomized trial with long-term follow-up.

Citation Format: Lindström LS, Yau C, Czene K, Thompson CK, van't Veer LJ, Nordenskjöld B, Stål O, Fornander T, Benz CC, Borowsky AD, Esserman LJ. Intra-tumor heterogeneity of the estrogen receptor predicts less benefit from tamoxifen therapy and poor long-term breast cancer patient survival – Retrospective analyses of the STO-3 randomized trial [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P2-05-03.

Abstract P3-05-02: Quantitative ERα measurements in TNBC from the I-SPY 2 TRIAL correlate with HER2-EGFR co-activation and heterodimerization

Background: We have previously described that TNBC patients whose tumors have both HER2 Y1248 phosphorylation (pHER2) “high” and phospho-EGFR Y1173 (pEGFR) “high” have increased response (pCR) to neratinib in the I-SPY2 TRIAL. We hypothesize that the paradoxical finding of a response prediction signature comprised of HER2 activation in a HER2 IHC/FISH-negative population means there must be a ligand-driven biochemical event responsible for the HER2 phosphorylation because HER2 mutations were also not found to be significant. Exploratory analysis of additional cellular signaling events and protein expression levels in pre-treatment, LCM-purified tumor epithelium by reverse phase protein microarray (RPPA) included semi-quantitative measurement of total levels of estrogen receptor alpha (ERα), which has been previously shown to be able to act as a membrane non-genomic signaling molecule through direct interaction with various tyrosine kinases including EGFR and HER2. Since ERα has been previously shown to act as a ligand and co-stimulate (activate) HER2 and EGFR when present at low levels, we investigated whether or not RPPA-measured ERα levels in the TNBC cohort analyzed to date were higher in tumors with both pHER2 “high” and pEGFR “high” levels and thus provide evidence explaining how HER2-EGFR activation is occurring in TNBC.

Methods: Using RPPA analysis, we measured 118 analytes in lysates of LCM tumor epithelium obtained from the pre-treatment biopsy samples of 86 TNBC (Allred=0) patients in the I-SPY2 TRIAL analyzed to date. Cutpoints for pEGFR and pHER2 were determined previously by ROC analysis for pCR correlation in the neratinib treated TNBC population, and used here to dichotomize the pHER2 and pEGFR data in the larger TNBC population. Wilcoxon Rank Sum testing was performed using the continuous variable total ERα data and compared the TNBC that were both pHER2 and pEGFR “high” (N=39) to the rest of the TNBC population (N=47). Total ERα values were then divided into “high” and “low” groups based on the TNBC population median value in order to determine frequency/percentages within each class. Our study is exploratory with no claims for generalizability of the data, and calculations are descriptive (e.g. p-values are measures of distance with no inferential content).

Results: Total ERα values were obtained in 84/86 TNBC tumors analyzed. Total levels of ERα were higher (p&lt; 0.006) in TNBC tumors with pHER2 and pEGFR “high” levels. 68% (26/38) of tumors in the pHER2 and pEGFR “high” group had ERα levels above the population median compared to 35% (16/46) in the rest of the TNBC population.

Conclusion: Our exploratory analysis reveals that ERα levels are significantly higher in TNBC with pHER2 and pEGFR activation and may be behaving as a direct signaling ligand in TNBC and driving HER2-EGFR signaling. This ERα-pHER2/pEGFR association was missed by current ER and HER2 clinical laboratory testing techniques, and if validated in larger independent study sets could suggest that utilization of new protein-based techniques defining ER more quantitatively could be helpful to understand tumor biology and therapeutic response prediction, especially in the context of TNBC that are ostensibly ER negative.

Citation Format: Gallagher RI, Yau C, Wolf DM, Dong T, Hirst G, Brown-Swigart L, ISPY-2 TRIAL Investigators, Buxton M, DeMichele A, van't Veer L, Yee D, Paoloni M, Esserman L, Berry D, Park J, Petricoin EF, Wulfkuhle JD. Quantitative ERα measurements in TNBC from the I-SPY 2 TRIAL correlate with HER2-EGFR co-activation and heterodimerization [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P3-05-02.

Abstract P5-11-18: Trajectory of patient (Pt) reported physical function (PF) during and after neoadjuvant chemotherapy in the I-SPY 2 trial

Background

Patients (pts) receiving chemotherapy for breast cancer experience toxicities impacting short and long-term quality of life (QOL). Within I-SPY 2, a trial adaptively randomizing stage II/III breast cancer pts to neoadjuvant chemotherapy +/- an investigational agent, we are collecting pt reported outcome (PRO) data to understand the impact of investigational agents on QOL. This PRO sub-study provides a unique opportunity to study QOL longitudinally and explore how pt and tumor characteristics, exposure to investigational therapies, and surgical outcome impact QOL.

Methods

Pts enrolled in this trial receive paclitaxel (T) +/- an investigational agent for 12 weeks followed by 4 cycles of doxorubicin and cyclophosphamide (AC). Surveys include the EORTC QLQ-C30 and BR-23, and PROMIS measures for QOL metrics including but not limited to physical function (PF), anxiety, and depression. Surveys are administered pre-chemotherapy to 2 years post-surgery. PF data from the EORTC and PROMIS instruments was analyzed for 238 pts at 5 sites (UCSF, UCSD, U of Pennsylvania, U of Minnesota, and Swedish Cancer Center). 48 pts completed baseline, inter-regimen (between T and AC), pre-operative and post-surgery surveys. Of the 48 pts 32 completed a 6-month follow up (FUP) and 31 completed a 1-year FUP survey. A linear mixed effect model, adjusting for HER2 status and treatment type was used to evaluate changes in PF over time. Sample size is small and statistics are descriptive rather than inferential.

Results

Median age of pts in this analysis was 50 (range 27-72).

Table 1 shows PROMIS & EORTC PF scores in this cohort.Time Point  PROMISEORTC  nMeanSEMeanSEPre-TreatmentAll4852.51.092.02.0 HER2+1553.51.594.12.2 HER2-3352.11.391.12.8Inter-RegimenAll4845.51.282.22.7 HER2+1548.62.384.44.2 HER2-3344.11.381.23.4Pre-SurgeryAll4843.91.179.42.3 HER2+1545.12.275.34.1 HER2-3343.41.381.32.86-Month FUPAll3248.11.487.41.9 HER2+1247.52.285.03.3 HER2-2048.41.888.92.41 Year FUPAll3148.91.488.43.1 HER2+949.12.988.95.4 HER2-2248.81.788.33.8

At baseline, mean PROMIS PF scores were higher than the US average (mean = 50) but declined as expected throughout treatment. HER2+ patients experienced a similar degree of recovery as HER2- pts post-surgery despite adjuvant treatment with Herceptin. Analysis of post-operative PROMIS PF indicated an average score within the U.S. general population (mean =50) but did not return to higher functioning seen at baseline levels (mean 52.5, p-value &lt; 0.05). Analysis of the EORTC PF sub-scale demonstrated a similar trend; however, the baseline and post-operative difference was not significant (p-value=0.15 for both FUP). Finding supports PROMIS PF ability to measure high functioning cancer patients.

Conclusions: Among a subset of pts who completed all surveys in the I-SPY 2 QOL substudy, PF did not return to baseline at 6-12 months post-operatively. Through transition to an electronic platform of data collection we hope to improve compliance with survey completion. We continue to analyze other QOL measures and plan to correlate QOL data with treatment arm, adverse events, comorbidities, and response to neoadjuvant treatment.

Citation Format: Shah M, Jensen R, Yau C, Straehley I, Berry DA, DeMichele A, Buxton MB, Hylton NM, Perlmutter J, Symmans WF, Tripathy D, Yee D, Wallace A, Kaplan HG, Clark A, Chien AJ, I-SPY 2 Investigators, Esserman LJ, Melisko ME. Trajectory of patient (Pt) reported physical function (PF) during and after neoadjuvant chemotherapy in the I-SPY 2 trial [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P5-11-18.