Pancreatic Neoplasms Are Frighteningly Common: A Role for Immune Surveillance?

What separates benign tissue from cancer? In this issue of Cancer Discovery, work from the Pasca di Magliano lab blurs the definition of normal (1). Carpenter and colleagues established a unique collaboration with their local organ donation center to analyze healthy adult pancreata from 30 donors. The donor pool included a wide distribution of ages, sexes, racial and ethnic backgrounds, as well as smoking and alcohol use patterns, which collectively represent the diversity of adults living in Michigan. The tissue was well preserved, and this is the only effort of its kind to have obtained a healthy pancreas in an undigested, nonnecrotic state. The care and attention that went into each sample cannot be overstated.

An atlas of the healthy human pancreas would have been commendable on its own, but what the authors found is truly surprising. A staggering 60% of healthy adults had a pancreas riddled with pancreatic intraepithelial neoplasia (PanIN), the precursors to pancreatic ductal adenocarcinoma (PDAC). This was true even in donors younger than 40 years old. Given that PDAC is exceedingly rare, something must prevent these lesions from progressing in the vast majority of individuals.

PanINs are the direct precursors to PDAC. In mice and humans, acinar-to-ductal metaplasia (ADM) is generated during pancreatitis as a defense measure and may spontaneously resolve; however, PanINs indicate dysplastic disease progression (2). In mice, acinar cell expression of mutant Kras is sufficient to cause PDAC by 12 to 18 months (2). Carpenter and colleagues find, consistent with prior studies, PanINs expressing oncogenic KRAS mutations, connecting these early lesions with late-stage disease. They then compare PanINs from healthy people with tumor-adjacent PanINs from patients with PDAC. Both types of PanINs organize a unique microenvironment of fibroblasts, myeloid cells, and T cells. PanIN differences are surprisingly slight—tumor-adjacent PanINs express a basal-like signature not present in healthy PanINs, but by most criteria, the healthy and tumor-adjacent profiles were similar. Fibroblasts from healthy and tumor-adjacent samples were distinct from the FAP⁺ fibroblasts found in PDAC, indicating that fibroblast changes in humans happen late, at the transition to cancer. Myeloid cell changes, on the other hand, appear to happen earlier in humans than would have been predicted from mouse models (2). Healthy and tumor-adjacent pancreas already contain myeloid cells transcriptionally similar to tumor-associated macrophages, although granulocyte frequencies increase in the setting of cancer. Changes to the microenvironment precede cancer formation and give a unique insight into cancer progression in humans.

So why don't we all have pancreatic cancer? One possibility is that the transition from PanIN to PDAC occurs very slowly, relying on other genetic alterations such as loss of TP53, which may take longer than the average human life span to develop. This hypothesis implies different rates of mutation for KRAS versus other genes, or perhaps a selective pressure for activated KRAS, as has been proposed in mouse models as a protective adaptation for sustaining ADM and thereby limiting tissue damage from pancreatic enzymes during repeat pancreatitis (3). Although epithelial cell–intrinsic pathways are certainly plausible, an intriguing possibility involves immune recognition and establishment of an equilibrium state between T cells and precursor lesions.

Immune surveillance has been theorized for more than a century; however, whether the immune system actively prevents cancer formation or outgrowth in humans has been difficult to demonstrate (4). It was not until 2010, when the registration trial for ipilimumab (anti–CTLA-4) showed monotherapy efficacy in metastatic melanoma, that we realized a significant fraction of patients with cancer had endogenously generated antitumor T cells held in check by CTLA-4. The fraction of immunotherapy-responsive patients expanded with PD-1 blockade and again with LAG3 blockade. Whenever immunotherapies succeed in reinvigorating antitumor T cells, we can claim that antitumor T cells preexisted. This circular logic fails to explain how these antitumor T cells arose in the first place, nor does it explain whether antitumor T cells are present in patients who do not respond to our current armamentarium of checkpoint blockade immunotherapies.

Enticingly, images from Carpenter and colleagues show PanIN lesions surrounded by a rim of CD4⁺ and, to a lesser extent, CD8⁺ T cells. These T cells do not appear to be classically defined regulatory T cells (Treg), as they lack the lineage-defining transcription factor FOXP3. All else about these T cells remains a mystery. Could they be recognizing PanIN-specific antigens? Most proteins expressed by PanINs are shared by normal acinar or ductal cells for which central tolerance in the thymus should delete any autoreactive T cells, but a few possibilities for PanIN-specific antigens emerge. Claudin18, proposed here as a marker for PanINs and being pursued clinically as a target for pancreatic cancer therapy, could provide a source of antigens for T-cell recognition. Mutant KRAS itself could generate T cells recognizing the specific KRAS-mutant peptide. Other potential antigens include posttranslationally modified peptides, such as phosphopeptides from hyperactivation of the RAS signaling cascade.

Assuming that antigenic targets are present, the next question involves the state of the dendritic cell bearing those potential antigens. Dendritic cells are constantly surveying tissues. In homeostasis, immature dendritic cells trafficking to lymph nodes at low frequencies induce immunologic tolerance by presenting self-antigens to naive T cells in the absence of costimulatory ligands, thereby inducing anergy or forming Tregs, particularly if TGFβ is also present (5). During inflammation, dendritic cells in tissues become activated, upregulate costimulatory ligands CD80/CD86, and migrate in larger numbers to the draining lymph node to engage naive T cells and convert them to effector CD4⁺ and CD8⁺ cells (5). Effector T cells proliferate, acquire the ability to produce cytokines and cytotoxic granules, and return via the blood to the site of inflammation. In the pancreas, dendritic cells are predominantly skewed toward tolerance; however, bouts of pancreatitis or more minor insults from alcohol or smoking could induce activated dendritic cells that incite effector T-cell priming against PanIN antigens.

Single-cell sequencing from Carpenter and colleagues provides further clues as to the state of CD4⁺ and CD8⁺ T cells in the healthy pancreas. As a caveat, no actual PanIN cells could be detected in the single-cell datasets; nevertheless, T cells survive single-cell processing more efficiently than epithelial cells, and likely some of the T cells captured had been surrounding ADM and PanIN lesions. T cells exist in multiple subsets, and certain genes can distinguish whether they have seen antigens recently (CCL5) or not (CCR7; refs. 6–8). From the gene expression patterns reported by Carpenter and colleagues, approximately two of three CD8 T cells and one of three CD4⁺ T cells are likely specific for antigens expressed in the pancreas. Of these, CD8⁺ T cells appear to be cytolytic and exhausted in equal ratios; CD4⁺ lineages are more difficult to discern from these datasets. Whether the antigen-experienced CD4⁺ T cells fall neatly into Th1, Th2, Th17, or Treg categories is as unclear as their exact spatial location, leaving open the question of whether the immune landscape surrounding PanINs is immune stimulatory or tolerogenic.

Two immune-related hypotheses emerge that could explain the paradox of high-frequency neoplastic lesions with low-frequency transition to cancer (Fig. 1). First, CD4⁺ and CD8⁺ effector T cells, primed by activated dendritic cells during routine insults to the pancreas, could curtail the outgrowth of PanINs. Because overt cytolysis in the pancreas risks release of digestive enzymes and further tissue destruction, this immune surveillance could produce an equilibrium state in most individuals, allowing the continued existence of PanINs kept in check by T cell–produced cytokines such as IFNγ (9). If immune surveillance is highly effective, cancer would arise only in the rare times when effector T cells fail. This hypothesis has several limitations. Pancreatic cancer risk is not dramatically increased by AIDS or other immunosuppressive conditions, although immunosuppression rarely reaches tissue-resident T cells and perhaps other infections or malignancies more rapidly afflict these patient groups. If a pool of antitumor effector T cells commonly exists in the pancreas, it is not capable of being productively reinvigorated by any immunotherapies tested thus far for the treatment of established PDAC. This could indicate either multiple layers of immune regulation in established cancer or abnormally poor T-cell responses in those rare people who progress to PDAC.

Alternatively, dendritic cells presenting PanIN-derived peptides in steady state may induce primarily anergic and regulatory T cells. CD4 Tregs, whether expressing FOXP3 or merely secreting immunosuppressive cytokines such as IL10, may be responsible for preventing pancreatitis throughout life and thereby suppress innate immune responses that can promote the final stages of oncogenesis. In a mouse model of Treg depletion from this same group of authors, loss of Tregs during inflammation-induced pancreatitis promoted a more rapid outgrowth of pancreatic cancer (10). Whether Tregs similarly guard against pancreatitis and, consequently, pancreatic cancer in humans is worth investigating.

Carpenter and colleagues provide one strikingly unequivocal conclusion that will spur research for years to come—the average healthy adult harbors pancreatic neoplasms starting from a young age. The immune system is not ignorant of these pancreatic cancer precursor lesions, although whether immune equilibrium is maintained by effector versus tolerogenic T cells remains to be determined.