hENT1 as a Predictive Biomarker in PDAC—Response

We would like to thank Boyd and colleagues for their thoughtful comments on our work demonstrating hENT1 as a predictive biomarker for gemcitabine and nab-paclitaxel (GnP) in advanced pancreatic ductal carcinoma (PDAC) in the COMPASS trial (NCT02750657; ref. 1). These findings will be further validated in the ongoing phase II PASS-01 trial (ref. 2; NCT04469556), randomizing patients to either modified FOLFIRINOX (mFFX) or GnP.The authors correctly highlight RNA expression after laser capture microdissection (LCM) as a challenging diagnostic to implement in current clinical practice and propose instead, multimodal IHC or PCR and mass spectrometry–based approaches. We agree entirely with these comments and hope that the COMPASS trial through its design provides a rich resource for biomarker discovery, validation, and the future development of diagnostics. The main objectives of the COMPASS trial were to establish whether whole-genomic sequencing (WGS) and RNA sequencing (RNA-seq) could be achieved within 8 weeks of a biopsy in treatment-naïve patients with advanced PDAC (3). Secondary objectives included evaluating RNA-based classifiers and ascertaining predictive biomarkers for the first-line regimens. In this regard, tumor enrichment through LCM was integrated into our protocol to resolve issues related to poor tumor cellularity which limits WGS interpretation and obviates the potential for stromal infiltration which can confound molecular classifiers as well as other biomarkers (3–5). Given the very high success rate of RNA-seq on COMPASS, we are able to address biomarker discovery in a real-world advanced PDAC population rather than a selected cohort.It is notable that FOLFIRINOX and perhaps NALIRIFOX are increasingly used as a first-line approach for patients with advanced PDAC, yet many trials utilize GnP as a backbone and only 50% of patients will receive second-line treatment in advanced PDAC (6). In addition, mFFX may not be a suitable regimen for many patients. This highlights the need for optimal first-line regimens and biomarkers beyond homologous recombination deficiency and targets within the KRAS wild-type population.The varying assays (IHC/mRNA/PCR), techniques (LCM vs. non-LCM), scoring methods (H-score, median, tertiles), and samples [microarray vs. formalin-fixed paraffin-embedded (FFPE) and primaries vs. metastases] used in evaluating hENT1 have contributed to its slow development as a biomarker (refs. 7–15; Table 1). Tumor enrichment has likely contributed to the success of hENT1 in our dataset. It has been long known that microdissected epithelial cells in PDAC result in significant differences in mRNA expression levels of hENT1 together with other proteins involved in gemcitabine metabolism when compared with non-microdissected tissue and normal pancreas (16). This variation will widely influence scoring methods and is in line with our own findings that stromal expression of hENT1 is significantly lower than in the epithelial compartment (Fig. 1A). Accordingly, the common case-specific variations in stromal content would thus likely lead to an overestimation and underestimation of epithelial hENT1 expression in bulk samples. In fact, Giovannetti reported one of the first transcriptional associations with hENT1 and gemcitabine efficacy in resected PDAC using qRT-PCR after LCM in resected tissue (ref. 15; Table 1). Other studies have noted the need to increase threshold levels for mRNA expression in the absence of microdissection due to contamination of the tumor microenvironment/normal tissue (14). Furthermore when comparing epithelial expression and stromal expression of hENT1 in a small series, it was only epithelial expression that impacted disease-free survival (DFS) and overall survival (OS) which likely reflects the putative mechanism, namely that epithelial gemcitabine uptake, mediated via hENT1 promotes an effective drug response (17).We agree that multigene sets and signatures will be crucial to biomarker development, however this will require robust validation in clinical trials. The COMPASS trial first confirmed the basal-like and classical signatures as prognostic in advanced PDAC and highlighted the possibility that basal-like patients do particularly poorly with mFFX; this will be validated in PASS-01 trial (NCT04469556). Encouragingly, the GemPred signature evaluated retrospectively in the PRODIGE-24 trial, demonstrated that patients considered GemPred+ve had a similar OS with gemcitabine adjuvant therapy compared with those who received mFFX and we look forward to further studies validating this signature (18, 19). This highlights the need to find biomarkers predicting gemcitabine ± taxane response.It is clear that single-gene IHC analyses alone may be challenging given mixed outcomes, although the 10D7G2 clone is most predictive (20). We agree that multiplex assays are increasingly utilized and perhaps more readily translatable to the clinic. For instance, initial single-marker IHC and ISH analyses on the COMPASS trial together with single-cell analyses have demonstrated the coexistence of subtypes in PDAC (4, 21). More recently, Williams and colleagues thoroughly dissected this coexistence of subtypes through multiplex immunofluorescence, incorporating six key markers and such assays may be important for therapy selection (22).In this regard, hENT1 metabolism is known to be regulated by a number of proteins including deoxycytidine kinase (dCK), cytidine deaminase, 5′-nucleotidase, and ribonucleotide reductase subunit M1 (RRM1). We believe hENT1 to be most critical and although other groups have suggested a role for dCK protein expression (9, 23), we found no correlation with GnP response when comparing high versus low expression of dCK (P > 0.9) or OS to GnP (P = 0.371; Fig. 1B and C). In addition, dCK expression did not affect mFFX response.Understanding the influence of tumor and stromal expression of these proteins as an amalgamation of translational research will help understand sensitivity and resistance to gemcitabine-based regimens. For example, while epithelial hENT1 expression is likely to determine therapy response, it appears that stromal factors such as TGFβ can regulate such hENT1 expression levels in epithelial cells (24). In COMPASS, the stromal elements and tumor epithelium were separated to facilitate the accurate analysis of these compartments, which do share a significant fraction of expressed genes and are therefore difficult to accurately deconvolute. We have recently begun to collect and deeply profile stroma fractions from PDAC tumors under careful consideration of the extensive spatial and biological heterogeneity in this multicellular compartment (21). While hENT1 protein was not abundant enough for detection in our nontargeted proteomics approach, the respective transcriptomic datasets will facilitate an analysis of the interplay between stromal programs and epithelial hENT1 expression to shed light on regulating mechanisms that will possibly also help us better understand how hENT1 expression is altered in chemotherapy-treated patients.The challenges highlighted by Boyd and colleagues are important especially with increasing use of neoadjuvant therapy. The SWOG S1505 trial provided interesting results in a pick the winner design of perioperative (12 weeks preoperative, 12 weeks postoperative) mFFX versus GnP (25). No differences in 2-year OS were found and although no significant differences in pathologic response rates were identified, more patients had a moderate or complete pathologic response rate in the GnP versus mFFX arm (40% vs. 25%). Identifying biomarkers from endoscopic biopsies will thus be important and the NEOPANCONE (NCT04472910) study will evaluate GATA6 in patients receiving mFFX. However, given the aforementioned study, predictors of neoadjuvant GnP remain important. Neoadjuvant studies also provide analyses of matched samples which will be critical to understanding the heterogeneity of hENT1 expression and to understand resistance and evolutionary changes of hENT1 expression under the influence of chemotherapy. It is difficult to know from our study whether the small numbers of cases (4/14) where expression levels changed was due to heterogeneity or the influence of chemotherapy. We therefore agree that analyzing matched samples should be undertaken.The COMPASS trial utilized most tissue for WGS/RNA-seq in a rapidly evolving field and as such future studies may seek to validate such multiplex assays on FFPE samples. However, we believe the purity of samples with COMPASS likely identify accurate tumor biomarkers that can be further developed for routine clinical application.The randomized PASS-01 trial will be important in validating our findings; samples within this trial do undergo microdissection. We expect however that additional FFPE material and organoid models will provide additional resources to evaluate key questions of chemotherapy sensitivity and resistance in PDAC.See the original Letter to the Editor, p. 2944S. Perera reports personal fees from Eisai Canada and Taiho Pharmaceuticals outside the submitted work. R. Grant reports other support from Pfizer; personal fees from AstraZeneca, Eisai, Incyte, and Knight Therapeutics outside the submitted work; and unpaid consulting with Tempus. J.J. Knox reports grants and personal fees from Roche and AstraZeneca; grants from Ibsen and Merck outside the submitted work. G.M. O'Kane reports personal fees from AstraZeneca, MSD, Incyte, Servier and grants and personal fees from Roche outside the submitted work. No disclosures were reported by the other authors.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.