Unique Stroma of Higher-Grade Pancreatic Intraepithelial Neoplasia: Fibroblastic Cluster Formation and CD36 Downregulation

Abstract

Introduction

Identifying pathological markers to differentiate between high- and low-grade pancreatic intraepithelial neoplasia (PanIN) remains an unmet clinical need. Recent studies suggest that the tumor microenvironment undergoes dynamic changes during PanIN progression. To establish objective pathological markers, this study investigated histological, immunohistochemical, and molecular alterations in PanIN-associated stroma across different grades.

Methods

Fibroblastic cluster (FC) was defined and characterized using immunohistochemistry (IHC) for SMA, CD31, and CD34, and their frequency was assessed according to PanIN grade in 202 PanIN lesions and 58 normal ducts. To identify molecular differences among PanIN types, multiple gene expression analysis (MGEA) was performed for six normal ducts and for three PanIN1, three PanIN2, and six PanIN3 lesions. IHC and fluorescent dual IHC for CD31 and CD36 were performed to quantify CD36-positive stromal vessels. RNA scope for PPARγ and IHC for thrombospondin-1 (TSP-1) and forkhead box O1 (FOXO-1) were performed to explore the molecular basis of CD36 downregulation.

Results

FCs were identified in 86.6% of PanIN3 stroma and were significantly more frequent than in lower-grade PanINs. MGEA revealed downregulation of CD36 in PanIN2–3 compared with PanIN1, and CD36-positive vessels were significantly fewer in higher-grade PanIN stroma. PPARγ signals were present in low-grade but absent in high-grade PanIN stroma. TSP-1 expression paralleled CD36 expression. These findings suggest that the possible relationship between the repression of CD36 in the stromal vasculature of high-grade PanIN and reduced PPARγ and TSP-1 signaling.

Conclusions

Higher-grade PanINs harbor a unique stroma characterized by FC formation and an increased proportion of CD36-negative vessels.

Introduction

The prognosis of pancreatic ductal carcinoma (PDAC) is poor, with a 5-year survival rate of approximately 12% [1]. Factors contributing to the poor prognosis include the formation of an immunosuppressive tumor microenvironment (TME) specific to PDAC, which is characterized by abundant stroma near the tumor and few immune-competent cells within the stroma, and the poor therapeutic effect of immune checkpoint inhibitors [2]. Given this background, early detection and resection of PDAC precursor lesions are of great clinical importance. The recognized precursor lesions of PDAC include intraductal papillary mucinous neoplasms (IPMNs) and pancreatic intraepithelial neoplasia (PanIN), from which most PDACs are thought to arise [3, 4].

IPMN is characterized by proliferation of mucus-producing epithelium within a dilated pancreatic duct, typically in a papillary configuration. Some, although not sensitive enough, clinical guidelines to detect high-grade IPMN have been established [5, 6]. In contrast, PanIN is a flat to micropapillary epithelial proliferative lesion of the pancreatic ducts, and no clinical method currently exists to distinguishing high-grade PanIN from low-grade PanIN [4]. Moreover, the pathological distinction between high- and low-grade PanIN relies largely on subjective assessment, and no sufficiently objective diagnostic criteria have been established. PanIN was historically classified into a three-tiered system – PanIN1 (mild dysplasia), PanIN2 (moderate dysplasia), and PanIN3 (severe dysplasia or carcinoma in situ) – based on the degree of cytological and architectural atypia and is considered to progress stepwise toward PDAC (the dysplasia-carcinoma sequence) [7, 8]. However, studies of autopsy cases and brain-dead donors have demonstrated that many individuals without known pancreatic disease harbor PanIN lesions, even from early in life, suggesting the majority of PanINs are clinically indolent [8, 9]. Thus, elucidating histomolecular events that drive progression to high-grade PanIN is essential for establishing objective criteria to distinguish high-grade or treatment-necessary PanINs from indolent lesions and to prevent malignant progression of PanINs [10]. To date, no epithelial genetic alterations have been identified that reliably discriminate low-grade from high-grade PanIN, although genetic abnormalities in KRAS, TP53, CDKN2A, and DPC4 occur at various frequencies as PanIN progresses [11, 12].

Recently, we reported that focal pancreatic parenchymal atrophy frequently accompanies PanIN lesions, particularly those of higher grade [13, 14]. In addition, in routine pathological practice, we have observed that clusters of several stromal spindle-shaped fibroblasts with plump nuclei – referred to as fibroblastic clusters (FCs) in this paper – often appear just beneath the basement membrane (BM) of PanINs, especially in PanINs of higher grade, suggesting grade-dependent epithelial-stromal interactions. Supporting this concept, studies using transgenic mouse models have shown that inflammatory and fibrotic stromal changes driven by a high-fat diet promote PanIN progression [15, 16], and a human tissue study has suggested that the transition from tumor-restraint to tumor-promoting stroma is associated with PanIN progression [17]. Collectively, these findings indicate that PanIN progression is not governed solely by epithelial genetic alterations but is also strongly influenced by the surrounding TME.

Therefore, this study aimed to characterize stromal differences across PanIN grades – including the frequency of FC formation – to identify objective markers that may aid in the pathological distinction between high- and low-grade PanIN and to provide insights into epithelial-stromal crosstalk involved in PanIN progression. Although the two-tiered PanIN grading system has been adopted since the 2019 WHO classification [3], the previous three-tiered system was used in this study because a substantial proportion of our PanIN cases had been resected before 2019. In addition, the authors hypothesized that the frequency of FCs may increase stepwise in a grade-dependent manner.

Methods

Case Selection and PanIN Extraction

This study included 46 pancreatectomy cases, each containing 1–15 PanIN lesions, for a total of 202 PanIN lesions. To obtain PanIN-grade-specific histomolecular data, target PanIN lesions were extracted from pancreatectomy specimens resected between January 2016 and April 2022 from patients who had not received preoperative chemotherapy. Lesions meeting any of the following seven exclusion criteria were omitted: (1) PanINs involving the main pancreatic duct (MPD); (2) PanINs showing histological evidence of duct-obstructive change, including cystic alteration; (3) PanINs with inflammatory change unrelated to the PanIN itself; (4) PanINs located within 2 mm of invasive carcinoma; (5) PanINs adjacent to and sharing stroma with another PanINs; (6) PanINs for which cancerization of duct or intraductal spreading of PDAC could not be excluded; and (7) PanINs for which authors could not reach consensus on grading. Because the stromal components of the MPD differ from those of small branch ducts – such as the presence of abundant elastic fibers and accessory glands – PanINs involving the MPD were excluded. When selecting high-grade PanINs from PDAC cases, lesions showing p53 overexpression and/or SMAD4 loss by immunohistochemistry (IHC) were excluded to avoid misclassification of cancerization of ducts (pseudo-PanIN), according to a previous study [12].

The extraction process and clinicopathological characteristics of the 46 cases containing 202 PanIN lesions (76 PanIN1s, 59 PanIN2s, and 67 PanIN3s) and 58 normal ducts are summarized in online supplementary Figure S1 (for all online suppl. material, see https://doi.org/10.1159/000553573) and online supplementary Tables S1, S2. Histological review was performed independently by three authors (Y.A., Y.F., and J.N.), and PanIN lesions were graded as PanIN1, PanIN2, or PanIN3 on hematoxylin and eosin-stained sections according to previous consensus studies of PanIN [11, 18].

IHC for PanIN Identification and FC Evaluation

IHC for p53 and SMAD4 was performed to discriminate possible cancerization of duct from PanIN lesions when suspected, based on a previous study [12]. IHC for smooth muscle actin (SMA), CD34, and CD31 was performed to identify stromal fibroblasts. In this study, we defined a 50 μm (width) × 25 μm (depth) area as one unit of duct stroma. We defined those fibroblasts as activated fibroblasts in which the short-axis width of the nucleus was more than twice that of background quiescent fibroblasts and showed SMA-positive, CD34-negative, and CD31-negative vessels. A FC was defined as an aggregate of ≥3 activated fibroblasts per one unit stromal area immediately beneath the ductal BM. The authors selected ≥3 as cutoff value for FC in this study according to our pilot receiver operating characteristic study using 20 PanIN2s and 20 PanIN3s (Doc. S1). To assess the reproducibility of FC identification, kappa values were calculated for all target ducts (76 PanIN1s, 59 PanIN2s, and 67 PanIN3s, and 58 normal ducts). Two authors (Y.A. and Y.F.) independently evaluated each lesion as FC positive or negative in a blinded manner. For cases in which the initial assessments differed, the final FC status was determined through joint review and discussion.

Typically, one to three FC foci were observed per PanIN duct, and a duct was considered FC positive when at least one FC focus was present. Representative histology of FC, information of antibodies, and IHC protocols are provided in online supplementary Figure S2 and Tables S3,S4 and in Document S2.

Multiple Gene Expression Analysis: Exploratory-Phase Study

Among the 46 cases, mRNA samples for MGEA were obtained from six normal ducts and from three PanIN1, three PanIN2, and six PanIN3 lesions. These 18 cases were selected based on the availability of sufficient number of ducts of the target grade to ensure adequate RNA yield. MGEA was performed using the nCounter^® PanCancer Immune Profiling Panel (Bruker Spatial Biology Inc., Billerica, MA, USA; nCounter^® PanCancer Immune Profiling Panel | NanoString), which quantifies 770 genes from different immune cell types, common checkpoint inhibitors, circulating tumor antigens, and genes that cover both adaptive and innate immune responses. Details of RNA isolation, hybridization, and data processing and selection of target gene of interest are provided in online supplementary Documents S3 and S4. Relative expression values were calculated by dividing the mean values of all samples in the study groups by the mean values of six normal pancreatic ducts. We compared the mRNA expression of 770 immune-related genes according to duct type using hierarchical clustering and volcano plots. A volcano plot was drawn between each of four study groups (normal pancreatic duct, PanIN1, PanIN2, and PanIN3) and between the two groups, namely, the PanIN1 group and PanIN2 + PanIN3 group (higher-grade group). Bioinformatic and statistical analyses were performed using nSolver Analysis Software, ver.4.0.62, and the PanCancer Pathways Advanced Analysis module. The p values were calculated using the Student’s t test.

Immunohistochemical Evaluation of CD36 Expression in Stromal Vessels: Validation-Phase Study

Based on MGEA findings indicating differential expression of several genes, including CD 36, across PanIN grades, CD36 expression was further evaluated by IHC. Since CD36 expression was restricted to vascular endothelial cells, except for occasional plasma cells and rare adipocytes in the PanIN stroma, therefore, serial sections were stained for CD31 to determine the ratio of CD36-positive to CD31-positive vessels. For each of the 202 PanIN lesions and 58 normal ducts, stromal vessels were evaluated within three distance-defined regions from the BM:

• closest (<50 μm from BM)

• semi-closest (≥50 μm and <100 μm from BM)

• distant (≥100 μm and <150 μm)

using three rectangular areas of 50 × 100 μm each (shown in online suppl. Fig. S3). The evaluation of CD31- and CD36-positive vessels were performed in a blinded manner to PanIN grade. Fluorescent-labelled dual IHC was performed to confirm CD36/CD31 double positivity and CD36-negative/CD31-positive vessels. Information on the antibodies and fluorescent dyes used is summarized in online supplementary Table S2. BX-63 (Olympus, Tokyo, Japan) was used to observe the IHC slides, and ApoTome.2 (ZEISS, Oberkochen, Germany) was used to observe the fluorescent double IHC slides.

RNAscope Assay for PPARγ

To assess the relationship between PPARγ expression and CD36 expression in PanIN vessels, RNAscope analysis for PPARγ was performed using RNAscope™ 2.5 HD Reagent Kit-BROWN (#322300; Advanced Cell Diagnostics Inc., Newark, CA, USA), following the manufacturer’s instructions. A probe targeting PPARγ splicing variant 1 (NM_138712.3) was used. Six micrometer-thick FFPE sections of 36 normal ducts, 44 PanIN1, 15PanIN2, and 31 PanIN3 lesions were prepared for RNAscope. Positive (#313901, Hs PPIB) and (#310043, Hs-Dab) control probes were used to ensure interpretability.

IHC of Forkhead Box O1 and Thrombospondin-1

To examine the relationship of the expression of FOXO-1 and TSP-1 and that of CD36 in PanIN vessels, IHC for these markers was performed. Antibody information and IHC protocols are provided in online supplementary Table S1 and Document S2.

Statistics

Comparisons of categorical data among each PanIN and normal duct group were performed using Fisher’s exact test and those of serial data were performed using Mann-Whitney’s U test using EZR (Juichi Medical University Saitama Medical Centre, Saitama, Japan). Statistical significance was set at p < 0.05. Since multiple lesions (ranging from 1 to 15 PanINs per case) were extracted from 46 pancreatic specimens, the observations of CD36-positive vessel counts were not fully independent. Therefore, to assess the potential influence of individual cases on CD36 quantification, we additionally performed a median test comparing four case groups: (1) 4 cases containing ≥4 normal ducts, (2) 6 cases containing ≥5 PanIN1 ducts, (3) 6 cases containing ≥4 PanIN2 ducts, and (4) 7 cases containing ≥5 PanIN3 ducts.

Results

Frequent FC Formation in Higher-Grade PanIN

FCs were identified in 86.6%, 37.2%, 10.5%, and 0% of PanIN3, PanIN2, PanIN1, and normal ducts, respectively (Fig. 1). The frequency of FC formation was significantly higher in PanIN3 lesions than in normal ducts, PanIN1, and PanIN2 (p < 0.001 for all comparisons). FCs were also significantly more frequent in PanIN2 than in PanIN1 and normal ducts (p < 0.05 and 0.001, respectively). A representative histological image of FCs is shown in Figure 2. Regarding interobserver reproducibility in FC determination, the kappa values were 1.00, 1.00, 0.813, and 0.640 for normal ducts, PanIN1, PanIN2, and PanIN3 lesions, respectively, indicating perfect, perfect, almost perfect, and substantial agreement.

Fig. 1.

Frequency of fibroblastic cluster (FC) formation by pancreatic intraepithelial neoplasia (PanIN) grade. a Normal duct. b PanIN1. c PanIN2. d PanIN3.

Fig. 2.

Representative histology of fibroblastic cluster (FC) by pancreatic intraepithelial neoplasia (PanIN) grade. a PanIN1 without FC. b PanIN2 with FC (circled). c PanIN3 with FC (ciecled). d PanIN3, larger view of c. The arrows show fibroblasts (component of FC) harboring enlarged nuclei. e PanIN3, same part of d. The fibroblasts (component of FC, arrows) were positive for SMA with IHC.

Downregulation of CD36 in Higher-Grade PanIN Revealed by MGEA

The results of multiple gene expression analysis (MGEA) are presented as a volcano plot (Fig. 3). In the higher-grade group (PanIN2 + PanIN3), 22 genes were significantly upregulated and six genes were significantly downregulated compared with PanIN1 (p < 0.05). Notably, three of the six downregulated genes – PLA2G6, CFD, and CD36 – were associated with fatty-acid metabolism. In contrast, upregulated genes in the higher-grade group included tumor promoting factors (e.g., CEACAM6) and inflammatory mediators (e.g., TNFRSF11A and PTGS2).

Fig. 3.

Results of multiple gene expression analysis (volcano plot). Comparison between higher-grade group (PanIN2 + PanIN3) and PanIN1 group showed 22 upregulated and 6 downregulated genes in the former group with statistical significance (p < 0.05). Genes in the right and left halves of the graph were upregulated and downregulated, respectively, in the higher-grade group compared with the PanIN1 group. X axis, log2 (fold change); Y axis, −log10 (p value).

Decrease in CD36-Positive Vessels in Higher-Grade PanIN, Particularly at Basement Membrane-Closest Site

The number of CD36-positive vessels was significantly fewer in PanIN3 than in normal ducts and PanIN1 across all three stromal distance categories (p < 0.001 for all). PanIN2 lesions also showed significantly fewer CD36-positive vessels than normal ducts and PanIN1 (p < 0.05, for PanIN2 vs. normal at the distant site; p < 0.001 for all other comparisons). At the site closest to the BM, CD36-positive vessels were significantly fewer in PanIN3 than in PanIN2 (p < 0.001) (Fig. 4). CD31-positive vessels were significantly more abundant in PanIN3 than in normal ducts at all distance categories, and the ratio of CD36-positive to CD31-positive vessels was nearly zero in PanIN3 at the site categories of closest and semi-closest (online suppl. Fig. S4). The median test revealed no significant case-related effects on CD36 quantification across any PanIN or normal duct group and across all distance categories (NS for all). Representative IHC and fluorescent dual IHC images are shown in Figures 5 and 6.

Fig. 4.

CD36 expression for vessels of normal and pancreatic intraepithelial neoplasia (PanIN) stroma. The density of CD36-expressed vessels in stroma closest (a), semi-closest (b), and distant (c) to the basement membrane (BM) of PanIN. X axis: N, normal duct; P1, PanIN1; P2, PanIN2; P3, PanIN3; Y axis: number of CD36-positive vessels among CD31-positive vessels per 50 × 100 μm².

Fig. 5.

Representative immunohistochemistry (IHC) for CD31 and CD36 for vessels of pancreatic intraepithelial neoplasia (PanIN) stroma. a, d PanIN1. b, e PanIN2. c, f PanIN3. a–c Arrows indicate vessels positive for CD31. d–f Filled arrowheads indicate vessels double positive for CD31 and CD36 and open arrowheads indicate vessels positive for CD31, whereas negative for CD36. Note that many CD31-positive/CD36-negative vessels at the PaNIN2 and PanIN3 stroma, especially at the closer site to basement membrane (BM). a–c IHC for CD31. d–f IHC for CD36. d–f Same part of a, b, and c, respectively.

Fig. 6.

Representative view of dual fluorescent IHC for CD31 and CD36 by pancreatic intraepithelial neoplasia (PanIN) grade. a, d PanIN1. b, e PanIN2. c, f PanIN3. Arrows indicate the vessels positive for both CD31 (red signal) and CD36 (green signal), and arrowheads indicate vessels positive for CD31 whereas negative for CD36. Note the increase of CD31-positive/CD36-negative vessels at PanIN2 and PanIN3 stroma (d, e, and f are larger view of a, b, and c, respectively).

RNAscope Analysis for PPARγ

RNAscope for PPARγ demonstrated positive signals in adipocytes of non-tumorous areas. Within PanIN lesions, signals were predominantly detected in epithelial cells, with occasional stromal signals close to the ductal BM (Fig. 7). Epithelial PPARγ expression was substantially higher in PanIN3 than in PanIN2 or PanIN1. Stromal PPARγ signals were scattered in PanIN1-2 stroma but were absent in PanIN3 stroma.

Fig. 7.

RNA scope for PPARγ.a PanIN1. No PPARγ signal is seen for tumor epithelia. b PanIN2. Several PPARγ signals are seen for tumor epithelia of PanIN2 (Arrows). c PanIN3. Abundant PPARγ signals are seen for tumor epithelia. d PanIN1. PPARγ signals are seen in the stroma (arrows). e Adipose tissue. PPARγ signals are also seen for normal adipocytes (arrows).

IHC for FOXO-1 and TSP-1

IHC showed FOXO-1 positivity in stromal vessels of all normal ducts and PanIN1-3 lesions, irrespective of CD36 expression. TSP-1 was expressed in adipocytes and stromal vessels of all normal ducts and PanIN1-3 lesions, except for CD36-negative vessels located immediately adjacent to PanIN3 epithelia (Fig. 8).

Fig. 8.

Representative IHC for CD36, TSP-1, and FOXO-1. a–d PanIN1. The vessel located just beneath the PanIN epithelia shows CD36-positive (b, arrows), TSP-1-positive (c, arrows), and FOXO-1-positive (d, arrows) vessels. e–h PanIN3. The vessel of the PanIN3 stroma shows CD36-negative (f, arrows), TSP-1-negative (g, arrows), and FOXO-1-positive (h, arrows) vessels.

Discussion

This study identified histological, immunohistochemical, and molecular differences in the PanIN stroma according to PanIN grade, suggesting grade-dependent epithelial-stromal crosstalk. Histologically, we demonstrated the following: (1) FCs, as defined above, were present in 10.5%, 37.2%, and 86.6% of PanIN1, PanIN2, and PanIN3 lesions, respectively, but were absent in normal pancreatic ducts; (2) vessel phenotypes changed with increasing PanIN grade, characterized by fewer CD36-positive vessels and more CD31-positive/CD36-negative vessels, with almost no CD36-positive vessels in the BM-closest and semiclosest sites (0–100 μm) of PanIN3; regarding molecular and immunohistochemical alterations, (3) MGEA revealed downregulation of fatty-acid metabolism-related genes, including CD36, in higher-grade PanIN compared with PanIN1; and (4) RNAscope demonstrated reduced PPARγ expression in PanIN3 stroma compared with PanIN1 and PanIN2, while IHC showed downregulation of TSP-1 – but not FOXO-1 – in vascular endothelial cells of PanIN3 stroma.

In routine pathological practice, distinguishing high-grade from low-grade PanIN relies largely on subjective assessment of hematoxylin and eosin-stained sections, and poor interobserver reproducibility has long been recognized [10]. To date, no sufficiently objective diagnostic method has been established. The stromal alterations identified in this study – particularly FC formation and the increase in CD31-positive/CD36-negative vessels near the BM – may provide objective markers to differentiate high-grade from low-grade PanIN.

Epithelial-mesenchymal crosstalk is increasingly recognized as a driver of precursor lesions in several carcinomas, highlighting the role of fibroblast phenotypic changes in lesion progression [19, 20]. However, only a few studies have characterized the PanIN-specific TME. Tian et al. reported overexpression of serum proteins, fibrinogens, and von Willebrand factors in the PanIN TME using mass spectrometry-based proteomics, suggesting that a fibrin-rich matrix may promote neoplastic cell migration and angiogenesis [21]. Carpenter et al. [9] demonstrated a unique PanIN TME using multiplex fluorescent IHC, showing more heterogeneous fibroblast phenotypes and a distinctive inflammatory infiltrate.

To our knowledge, this study is the first to demonstrate that histologically defined FC formation is associated with higher PanIN grade. It is also the first to show that increased CD36-repressed vessels correlate with higher-grade PanIN. In routine practice, we have observed that PanIN – particularly high-grade lesions – often exhibit a distinctive stroma composed of small aggregates of fibroblasts immediately beneath the ductal BM, accompanied by background edematous change and enlarged nuclei compared with surrounding spindle-shaped stromal cells. Although our definitions of activated fibroblasts (short-axis nuclear width more than twice that of background quiescent fibroblasts) and FCs (clusters of ≥3 fibroblasts within an area of 50 µm × 25 µm immediately beneath the BM) are somewhat arbitrary, the significant association between FC formation and PanIN grade suggests that FC formation represents an essential event in PanIN progression.

Regarding interobserver reproducibility in FC identification, the kappa values were 1.00, 1.00, 0.813, and 0.640 for normal ducts, PanIN1, PanIN2, and PanIN3 lesions, respectively, indicating perfect, perfect, almost perfect, and substantial agreement. Examination of the cases in which the initial judgments of the two authors were discordant revealed that the width of enlarged fibroblastic nuclei was occasionally difficult to assess (online suppl. Fig. S5). The nuclei of activated fibroblasts observed within FCs are clearly, though only slightly, larger than those of background quiescent fibroblasts and are phenotypically distinct from the activated fibroblasts/myofibroblasts seen in the desmoplastic stroma of invasive carcinoma.

CD36 downregulation in tumor vessels has been reported in several invasive carcinomas, including gastric and ovarian cancers [22]. In precursor lesions, Caruso et al. recently reported more frequent CD36 loss in ductal carcinoma in situ (DCIS) lesions associated with subsequent invasive ductal carcinoma than in DCIS without subsequent invasive ductal carcinoma, suggesting that CD36 repression is an early proactive event in breast tumorigenesis [23]. CD36 is a multiligand scavenger receptor expressed on the stromal cell populations, including endothelial cells, adipocytes, mononuclear phagocytes, and platelets. As a high-affinity receptor for TSP-1, endothelial CD36 transduces potent anti-angiogenic signals that suppress microvascular expansion [22, 24, 25]. Under physiological conditions, this CD36/TSP-1 axis functions as a negative regulator of angiogenesis. In TMEs, however, growth factor-driven repression of CD36 and TSP-1 attenuates these anti-angiogenic cues, thereby enabling a proangiogenic, tumor-promoting state [22, 24]. In addition, Caruso et al. [23] have reported that, in the stroma of breast DCIS, CD36-lost pericytes play as a cellular source of activated myofibroblasts, demonstrating phenotypic transition from pericytes (loss PDGFRB and RGS5) to myofibroblasts (increased SMA filaments). The histogenesis/cellular source of FCs are beyond the scope of this study since we have not performed lineage-related analyses or we have not examined of spatial overlap between FCs and CD36-negative vessels; however, it is demanded to investigate the relationship of these two cellular components, when considering the co-localization of CD36-repressed vessels and FCs near BM in higher-grade-PanIN in this study. Nowadays, numerous studies have shown that endothelial-to-mesenchymal transition contributes to fibroblast populations in malignant, vascular, inflammatory, and fibrotic disorders [26, 27]. Our findings that higher-grade PanIN more often harbored CD36-repressed vessels in their wall as well as previous study with breast DCIS suggest that CD36 repression may represent a proactive stromal event in the early stages of tumorigenesis in both PanIN and breast DCIS.

To explore the molecular basis of CD36 downregulation in higher-grade PanIN, we examined PPARγ, TSP-1, and FOXO-1 expression across PanIN grades since CD36 expression is regulated by transcription factors such as PPARγ, a ligand-dependent activator, and FOXO-1, an inhibitor [22, 23]. Our RNAscope analysis showed loss of stromal PPARγ signal in PanIN3, whereas it was retained in PanIN1 and 2, suggesting that PPARγ deficiency may relate to CD36 repression in PanIN3 vessels. IHC demonstrated parallel expression of CD36 and TSP-1 in stromal vessels, suggesting that loss of TSP-1 signaling may relate to CD36 repression. In contrast, FOXO-1 expression did not correlate with CD36 status as FOXO-1 was expressed in all vessels regardless of CD36 expression. Further experiments using functional analysis are demanded.

RNAscope analysis also revealed PPARγ mRNA expression in PanIN epithelia, with substantially higher expression in PanIN3 than in PanIN2 or PanIN1. Although the role of PPARγ in tumorigenesis remains controversial [27], protumorigenic functions have been suggested in several cancers [28, 29]. While elucidating the epithelial role of PPARγ is beyond the scope of this study, further investigation may clarify additional molecular events in PanIN progression and potentially provide supplementary markers to distinguish low- from high-grade PanIN.

It is noteworthy that three of the six genes mostly downregulated in PanIN2/PanIN3 compared with PanIN1 in the MGEA – PLAG2G6, CFD, and CD36 – are involved in fatty-acid metabolism, suggesting that lipid metabolic dysregulation may contribute to PanIN progression. Several recent studies have demonstrated that high-fat diets promote PanIN proliferation and progression to PDAC in mice harboring oncogenic Kras [15, 16, 25, 30].

In this study, 37.2% of PanIN2 lesions harbored FCs. In addition, CD36-positive vessels in PanIN2 stroma were significantly fewer than those in PanIN1 stroma at all stromal sites, whereas were significantly more abundant than those in PanIN3 stroma at the site closest to the BM. These findings suggest PanIN2 lesions may comprise two stromal phenotypes: PanIN2 with stroma resembling that of PanIN3 and PanIN2 without such stromal features. Although the current two-tiered PanIN grading system provisionally categorizes PanIN2 as low-grade PanIN, our results indicate that PanIN2 may be better stratified by stromal type. Further studies are warranted to clarify the clinicopathological differences between PanINs with and without FCs.

This study has several limitations. First, although we identified multiple histomolecular alterations associated with PanIN grade and suggested grade-dependent epithelial-stromal crosstalk, clear-cut differences between PanIN2 and PanIN3 were not detected. Consequently, biologically or clinically meaningful distinctions specifically separating PanIN2 from PanIN3 could not be established. When PanIN2 and PanIN3 lesions were analyzed separately, the MGEA did not yield differentially expressed stromal-specific genes.

FCs were significantly more frequent in PanIN3, yet they were also present in 10.5% of PanIN1 and 37.2% of PanIN2 lesions. Similarly, although CD36-repressed vessels were more abundant in PanIN3, scattered CD36-negative vessels were also observed in PanIN1 and PanIN2. Second, MGEA analysis could not isolate pure stromal RNA. Because PanIN ducts are thin and the required RNA quantity was substantial, RNA was extracted from both epithelium and stroma. Therefore, we relied on literature and the Human Protein Atlas (https://www.proteinatlas.org) to infer cellular sources when interpreting genes of interest. However, we acknowledge that epithelial RNA contamination in our MGEA analysis may have substantially influenced the results, potentially obscuring important stromal-specific genes. In conclusion, this study demonstrated that higher-grade PanIN harbors a unique stroma characterized by frequent FC formation and CD36 repression.

Acknowledgments

The authors thank medical technologist Katsumi Miyahara, PhD, Department of the Laboratory of Morphology and Image Analysis, Biomedical Research Core Facilities, Juntendo University, for his assistance in capturing fluorescent images. We would like to thank Editage (www.editage.com) for the English language editing.

Statement of Ethics

This study was approved by the Institutional Review Board of Juntendo University (approval codes: #M18-0045). This study is a retrospective analysis using left-over tissue obtained after the completion of routine pathological diagnosis, the Institutional Review Board of Juntendo University waived the requirement for written informed consent from all participants.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This study was supported by the following four funding sources: (a) Grant-in-Aid from the Japan Society for the Promotion of Sciences (JSPS) KAKENHI (Grant No. #24K10113 to Y.F. and J.N.) for support of the RNAscope analysis; (b) Grant-in-Aid from the Japan Society for the Promotion of Sciences (JSPS) KAKENHI (Grant No. #23K14582 to J.N.) for support of MGEA study; (c) Tokyo Metropolitan Hospital Clinical Research Project Fund (Grant No. R050403006 to J.N.) for support of the MGEA study; (d) Pancreas Research Foundation of Japan (to J.N.) for support of the IHC study.

Author Contributions

Yuuji Aoki conducted the study. Yuki Fukumura designed and supervised the study. Yuuji Aoki, Jun Nakahodo, Yifare Maimaitiaili, and Mulading Maimaitituerxun contributed sample collection and pathological data collection. Yuuji Aoki, Yuki Fukumura, and Jun Nakahodo analyzed the data and Yoshihiro Mezawa assisted data analyses. Yuuji Aoki and Yuki Fukumura wrote the paper and Takashi Yao provided advice for this study.

Funding Statement

This study was supported by the following four funding sources: (a) Grant-in-Aid from the Japan Society for the Promotion of Sciences (JSPS) KAKENHI (Grant No. #24K10113 to Y.F. and J.N.) for support of the RNAscope analysis; (b) Grant-in-Aid from the Japan Society for the Promotion of Sciences (JSPS) KAKENHI (Grant No. #23K14582 to J.N.) for support of MGEA study; (c) Tokyo Metropolitan Hospital Clinical Research Project Fund (Grant No. R050403006 to J.N.) for support of the MGEA study; (d) Pancreas Research Foundation of Japan (to J.N.) for support of the IHC study.

Data Availability Statement

All data generated or analyzed during this study are included in this article. Further inquiries can be directed to the corresponding author.