In this issue, Miyabayashi and colleagues describe a novel intraductal model of pancreatic cancer that allows modeling of the transcriptional subtypes of pancreatic cancer. Using this model, they are able to observe subtype switching driven by the microenvironment, a process at least partially mediated by RAS signaling.
See related article by Miyabayashi et al., p. 1566.
Pancreatic ductal adenocarcinoma (PDAC) is predicted to be the second most common cause of cancer-related death by 2030. The disease is aggressive and current therapies are largely ineffective. Together, these factors translate to a 5-year overall survival of less than 10%. In vivo models that recapitulate disease progression, response to therapy, and genomic and transcriptomic diversity are vital for our understanding of the disease and for therapeutic testing, but, despite years of research, there are drawbacks associated with all the models developed thus far. In this issue of Cancer Discovery, Miyabayashi and colleagues present a novel transplant model of the disease which addresses some of these concerns (1).
At the simplest level, in vivo models can be separated into transplant models and genetically engineered mouse models (GEMM). Most GEMMs of pancreatic cancer use pancreas-specific oncogenic KRAS expression as a backbone on which to layer further genetic aberrations under the control of Cre-Lox technology driven by pancreas-specific promoters. These models are valuable, given that they mimic the full spectrum of human disease from precursor lesions to invasive and often metastatic adenocarcinoma. The KC model (originally LSL-KrasG12D; Pdx1-Cre, but now the term is also used to describe Ptf1a-Cre-driven models) has been successfully used to evaluate the contribution of additional genetic lesions to tumor progression. The most widely used of these models is the KPC model [LSL-KrasG12D; Trp53R172H; Pdx1-Cre (or Ptf1a-Cre)] in which Kras and Trp53 mutations are directed to the pancreas (2). Tumors arising in these models recapitulate the histopathology and dense desmoplastic stroma of human PDAC and, as with patients, are refractive to therapy. However, an obvious limitation is that the activation and inactivation of genes occurs within the embryonic pancreas, leading to disease initiation in very young mice. Furthermore, recombination tends to occur in multiple cell types and ubiquitously throughout the pancreas, leading to multiple initiation events, while extrapancreatic expression of oncogenes can lead to “off-target” pathologies. Some researchers have developed inducible models to manipulate gene expression more stochastically in the adult pancreas; however, induction of pancreatitis may be required to drive PDAC formation in these systems, raising the question of how many initiating events occur throughout a systemically inflamed pancreas where multiple cells express oncogenic KRAS. Moreover, in most models, the induction of mutations occurs simultaneously, rather than in a sequential manner, although dual recombinase systems have been developed to partially address this limitation (3), and recent evidence suggests that tumors in the KPC model continue to accumulate subclonal somatic alterations as they progress (4). With increased genetic complexity, lower breeding efficiency and higher cost are also considerations. Improvements in CRISPR/Cas9 and embryonic stem cell–based models may help circumvent some of these drawbacks in the future.
Transplant models have traditionally exhibited a much shorter latency than GEMMs and are comparatively easy and cheap to run, but the utility of the model is influenced by the source material. There are numerous “off-the-shelf” cell lines available; however, these do not reflect the heterogeneity or disease progression of patient tumors, and the resulting tumors lack significant stromal interactions. Organoids grown directly from patient tumors have addressed some of these concerns and may better reflect tumor heterogeneity and response to therapy (5), as well as recapitulating some aspects of progression from precursor lesions to invasive, metastatic adenocarcinoma (6). In the past decade, there has been an expansion in studies using patient-derived xenografts (PDX), and these have been championed as bench-to-bedside systems in which to test personalized therapies in each patient's own tumor. There are advantages of PDX models over simple cell line transplants, namely that the transplanted tissue better represents the original tumor from the patient and can retain the genetic landscape and heterogeneity, as well as the stromal characteristics of the donor tumor. However, there are still clear limitations, the most obvious being that the recipient mice lack an intact immune system. Furthermore, unlike tumor evolution in patients, these models lack any consideration of the progression from normal tissue through precursor lesions to invasive tumor, and the interactions that take place during these processes. Syngeneic models, in which both the grafted tissue and the host are species- and strain-matched, allow the tumor to develop in the context of an intact immune system and can partially recapitulate the mutations found in human tumors; however, tumor heterogeneity is not well represented in these models.
In this issue, Miyabayashi and colleagues introduce another model for consideration (1). They present a xenotransplant model of PDAC in which single-cell suspensions of patient-derived organoids are injected directly into the ducts of the murine pancreas via modified retrograde intraductal injection rather than the traditional orthograft into the pancreatic interstitium. This technique, which they term intraductally grafted organoid (IGO), led to enhanced engraftment rate, but, more significantly, the resulting lesions exhibited more diverse histology and could be separated into two main subtypes: a slow-progressing subtype with a glandular phenotype that was restricted to the intraductal region and lacked a desmoplastic response, and a fast-growing more invasive and less glandular subtype with abundant stromal deposition (Fig. 1). Interestingly, the speed of progression in this intraductal model was consistent with the tumor stage of the parental tissue, and although proliferative rates in vitro did not predict the speed of progression of IGOs in vivo, organoids that generated slow progressors exhibited less dense and more cystic morphology compared with organoids formed by fast progressors. In comparison, tumors arising in the orthotopically grafted organoid (OGO) model did not follow this bimodal distribution of survival and were more uniform in phenotype, with rapid development of extraductal neoplasms, abundant interactions between cancer and stromal cells, and prominent collagen deposition (Fig. 1). The authors suggest that this dichotomy in behavior is a result of the cues received by the tumor cells, with IGO transplants exposed to signals primarily from the pancreatic ducts (where human PDAC progression is thought to evolve), whereas OGO transplants receive signals from a variety of cells and tissues including stellate cells, vasculature, and extracellular matrix. This is not unreasonable, as previous studies have shown that alterations within the tumor microenvironment can influence transcriptional subtype (7); however, it also serves as a reminder that because immunodeficient mice are the hosts for this model, these tumors still evolve within an incomplete immune microenvironment.
Over the past few years, studies have identified transcriptional subtypes of PDAC that can predict prognosis. Although there is variation between studies, these subtypes can be distilled down into a poorly prognostic, poorly differentiated squamous/basal subtype, and a better differentiated classical/progenitor subtype (8). However, understanding of how the different subtypes arise is still lacking, and accurately representing them in vivo has proved a challenge. Here, Miyabayashi and colleagues find that the site of growth can influence the transcriptional subtype of lesions. Genes involved in epithelial–mesenchymal transition were enriched in OGO versus IGO tumor cells whereas TGFβ ligands were upregulated in both the cancer cells and the stromal cells of OGO tumors. Squamous/basal signatures were also enriched in OGO tumors, whereas classical/progenitor signatures were enriched in IGOs. The dichotomy was also clear when comparing slow-progressing intraductal lesions with fast-progressing invasive tumors, with the former bearing transcriptional signatures reminiscent of the classical/progenitor subtype and the latter enriched for squamous/basal signatures (Fig. 1). Thus, this novel IGO model may better represent human PDAC because both subtypes are represented at proportions more reflective of human tumors.
Interestingly, invasive squamous/basal subtype tumors were also enriched for KRAS signaling and exhibited increased KRAS copy number (Fig. 1). In fact, hyperactivation of KRAS in organoids that generated slow-progressing intraductal tumors was sufficient to drive enrichment of squamous/basal signatures in vitro and in vivo and promote an invasive phenotype and desmoplastic reaction in vivo, in keeping with previous reports of a correlation between KRAS gene dosage and aggressive undifferentiated phenotype (9). Although hyperactivation of the RAS pathway was not required to drive formation of squamous/basal subtype tumors, it may be sufficient to drive switching toward a squamous/basal specification, although it should be noted that prior studies have demonstrated that other genetic lesions can also support this specification. Miyabayashi and colleagues also find that even single clones can simultaneously display both squamous/basal-like and progenitor/classical phenotypes and show that in slow-progressing intraductal tumors, where small invasive lesions develop, squamous/basal markers are expressed and RAS signaling is enhanced. They suggest that the subtype of a transplanted tumor is not predetermined by genetics, but rather is influenced by the surrounding microenvironment with switching resulting from cell plasticity (1). This is not inconsistent with recent analyses of mixed-subtype tumors from autopsy samples, which suggested that the squamous/basal subtype represented a subclonal expansion from classical/progenitor tumors, although in that study squamous/basal expansion was associated with mutations in chromatin modifiers and MYC amplification (10).
It is becoming clear that tumor heterogeneity is not only a result of the mutations within the tumor cells, but also due to the environment in which those cells find themselves. This heterogeneity needs to be considered when selecting or designing models and when evaluating therapies. This novel model increases the options available and may be useful for evaluating therapies, particularly those targeted toward specific subtypes.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.