Autoimmune pancreatitis (AIP) is characterized by tumefactive lesions and the clinical and radiologic findings often mimic pancreatic carcinoma.1,2 The overlap with pancreatic cancer and consequent pancreatic resections offered pathologists a window into the morphologic and immunologic aspects of the disease. These studies revealed an inhomogenous pancreatic disorder with 2 distinct diseases: (1) type 1, an IgG4-related disease, and (2) type 2, also referred to as idiopathic duct-centric pancreatitis.3–6 The latter disorder, unrelated to IgG4-related disease, lacks elevations of serum and tissue IgG4.3,4
Serum and tissue IgG4, although imperfect, remains the preferred biomarker for the diagnosis of IgG4-related disease and type 1 AIP, and in conjunction with other clinical and radiologic features, facilitates a presumptive or definitive diagnosis, often without the need for a biopsy.1,3 The diagnosis of type 2 AIP presents a far greater challenge, primarily because of the lack of a serum or tissue-based marker. Per the International Consensus Diagnostic Criteria for AIP, the diagnosis of type 2 disease requires either supportive histology and/or concurrent inflammatory bowel disease.7 However, the incidence of inflammatory bowel disease is as low as 10%.8–11 Although the diagnosis relies on morphologic evaluation, the cardinal histologic feature, granulocytic epithelial lesion, may not be identified on biopsy and, furthermore, duct-associated neutrophils are present, albeit uncommonly, in peritumoral pancreatic tissue.11 The inability to definitively distinguish type 2 AIP from pancreatic carcinoma underlies the frequent diagnostic uncertainly associated with these patients, a scenario that occasionally leads to an inadvertent pancreatic resection. There is thus a significant need for serum and/or a tissue-based biomarker to assist in the diagnosis of type 2 AIP.
One of the authors noted the expression of programmed death-ligand 1(PD-L1) and indoleamine 2,3-dioxygenase 1 (IDO1) in pancreatic ductal epithelium of patients with type 2 AIP raising the possibility that these immunomodulatory proteins play a mechanistic role and could have diagnostic value. Herein, we examine the expression of PD-L1 and IDO1 as a diagnostic tool to distinguish type 2 AIP from other forms of pancreatitis and pancreatic ductal adenocarcinoma (PDAC).
MATERIALS AND METHODS
Resected Examples of Type 2 AIP
We evaluated all non-neoplastic pancreatic resections between 1990 and 2018 and identified 35 patients with type 2 AIP. The diagnosis of type 2 AIP was based on a constellation of clinical and histologic features and required, at a minimum, periductal lymphoplasmacytic inflammation and intraductal neutrophils. In addition, the majority of patients showed one or more clinical features associated with type 2 AIP, including the presence of a mass-lesion, pancreatic and bile duct strictures and/or inflammatory bowel disease. These cases also lacked pathologic changes of type 1 AIP, alcohol-related pancreatic injury and groove pancreatitis. The resections had fewer than 50 IgG4 positive plasma cells per high-power field and showed an IgG4 to IgG ratio of <40%.
We recorded demographic, clinical and histologic data. We graded the intensity of periductal lymphoplasmacytic infiltrates as: 1 (mild, illustrated in Fig. 1A) and 3 (severe, illustrated in Fig. 1C); intermediate lesions were graded as 2. Intraductal neutrophils were graded as 1 (ductal neutrophils without abscesses), 2 (ductal abscesses), or 3 (ductal ulceration). Interlobular fibrosis was graded as: 0 (absent), 1 (mild to moderate, defined as delicate fibrous septa, typically less than thickness of 2 islets) or 2 (severe, septa greater than thickness of 2 islets). Pancreatic atrophy was graded as absent, mild (<50% loss of acinar tissue) and severe (>50% loss of acinar tissue).
Type 1 Autoimmune Pancreatitis
The diagnosis of type 1 AIP (n=14) was based on cross-sectional imaging, elevated serum IgG4, histologic evaluation of the pancreatectomy specimen and tissue IgG4 and IgG stains.7
Pancreatic Ductal Adenocarcinoma
PDAC was selected as a disease control because of its overlapping clinical and radiologic features with AIP. Furthermore, peritumoral pancreas may show granulocytic epithelial lesions. We utilized tissue microarrays comprised of 237 patients with PDAC. Each patient was represented by 4 tissue cores, each measuring 2 mm, from 3 paraffin blocks. Although the tissue microarrays were designed to represent tumor, peritumoral areas were also included in the tissue microarray.
To comprehensively assess the peritumoral pancreas we also evaluated whole sections from 41 additional patients with PDAC.
Other Forms of Chronic Pancreatitis
We also evaluated 10 pancreata from patients with chronic pancreatitis—not otherwise specified, the latter defined as chronic alcoholic pancreatitis or pancreatitis of uncertain etiology. To simulate an overlap with type 2 AIP we required a moderate periductal lymphoplasmacytic infiltrate. We also evaluated 14 patients with groove pancreatitis; these cases showed pancreatic fibrosis and inflammation localized to the “groove” between the bile duct and pancreas.
Validation of PD-L1 Stain in Type 2 AIP
We evaluated 9 patients who presented to this institution between 2011 and 2018 who underwent an endoscopic ultrasound (EUS)-guided pancreatic core biopsy for a clinical suspicion of AIP. In addition, we also included EUS guided biopsies from 4 patients with type 2 AIP from Japan and the United Kingdom. EUS-guided biopsies from 7 patients with type 1 AIP and 17 patients with PDAC were studied. We reviewed clinical and radiologic data and specifically interrogated electronic databases for the use of immunosuppressive drugs and response of strictures/mass lesion to these agents.
PD-L1 and IDO1 in Ulcerative Colitis
Given the close association between type 2 AIP and inflammatory bowel disease, we also evaluated IDO1 (n=25) and PD-L1 (n=20) in mucosal biopsies from patients with active ulcerative colitis.
Immunostaining was performed on the Leica Bond-Max automatic immunostainer (Leica, Bannockburn, IL) (Table 1). Immunohistochemistry for PD-L1 and IDO1 was performed on whole sections while that for PD-1 and CD8 on tissue microarrays. A PD-L1 stain was judged positive when membranous reactivity was noted in ductal epithelium, regardless of the diameter of the duct; cytoplasmic only staining was not considered in this study. IDO1 reactivity was recorded as present when ductal cells showed cytoplasmic staining.
To assess the validity of our PD-L1 platform we evaluated a cohort of colonic adenocarcinomas. PD-L1 reactivity in colon carcinoma is closely linked to microsatellite instability status.12,13 Twenty-eight percent (11/39) of mismatch repair deficient colon carcinomas and 6% (8/145) of mismatch repair proficient colon cancers were positive for PD-L1. The results are similar to those reported in prior studies.12,13 A quantitative analysis for PD-1 and CD8 was performed using digital image analysis software (Visiopharm, Hoersholm, Denmark).
All statistical tests were performed using SPSS software (version 21). Categorical variables were assessed by χ2 or Fisher exact test, as appropriate. Continuous variables were compared using the Student t test. P-values <0.05 were considered significant.
Type 2 Autoimmune Pancreatitis
Demographic and Clinical Information (Resected Samples)
The mean age of this cohort was 50 years with no sex predilection (F:M=1.2:1). Twenty-percent (7/35) of patients reported a history of inflammatory bowel disease. Virtually all patients reported abdominal pain (32/35), while 17% (6/35) presented with obstructive jaundice.
On cross-sectional imaging 21 (60%) patients showed a mass lesion. On imaging a common bile duct stricture was identified in 7 (20%) of patients while pancreatic duct strictures were noted in 11 (31%) patients.
Histologic Features (Resected Samples)
Most pancreata showed moderate to severe periductal inflammation (80%). All cases were associated with granulocytic epithelial lesions (Fig. 1C) and/or ulceration of ductal epithelium (Table 2); neutrophils within acini were also noted in 23% of cases. Although the pancreatic ducts were surrounded by significant inflammation, epithelial atypia was absent. The pancreata showed significant atrophy, prominence of endocrine elements, and interlobular and intralobular fibrosis. Histologic stigmata of alcoholic pancreatitis, inspissated proteinaceous material in ducts, were not identified and calcifications were noted in 3 cases. Obliterative phlebitis was not identified.
Pancreatic ductal PD-L1 (Figs. 1B, D) reactivity was noted in 69% (24/35) of cases. The reactivity was agnostic to duct diameter: both large and small caliber ducts were positive for PD-L1. Notably, in an analysis of a subset cases of cases (n=20), 47% of PD-L1-positive ducts were not associated with granulocytic epithelial lesions. Ductal IDO1 (Fig. 2) reactivity was noted in 58% (15/26) of patients with type 2 AIP. There was no correlation between PD-L1 and IDO1 reactivity (Pearson correlation, P=1.0).
The mean number of CD8 and PD-1-positive cells were 1590 (range: 533 to 5118, SD=240) and 97 per mm2 (range: 0 to 1397, SD=293), respectively. PD-1-positive cells did not correlate with ductal PD-L1 reactivity (P=0.85). Neither ductal PD-L1 (P=0.74) nor ductal IDO1 (P=0.11) reactivity correlated with CD8-positive cells.
Correlation of Immune Markers With Clinical and Histologic Parameters
Ductal PD-L1 reactivity correlated with higher grades of fibrosis (P=0.045) although not with intraductal neutrophilic infiltrates (P=0.18) or pancreatic atrophy (P=0.14). Ductal IDO1 did not correlate with intraductal neutrophils (P= 0.58), fibrosis (P=0.48) or atrophy (P=0.044).
Type 1 AIP, Chronic Pancreatitis, and Groove Pancreatitis
PD-L1 and IDO1 ductal reactivity was absent in all 14 patients with type 1 AIP. Among the 24 other cases of chronic pancreatitis, a single patient with groove pancreatitis showed ductal PD-L1 reactivity; all cases were negative for IDO1.
Pancreatic Ductal Adenocarcinoma
The tumors from 6 (2.5%) patients were positive for PD-L1; peritumoral non-neoplastic pancreatic parenchyma containing ducts and acini, present in 210 cases, was negative. The reactivity ranged from 30% to 100% (mean: 56%) of tumor cells. Twenty-eight percent of PDACs were positive for IDO1, reactivity ranged from 1% to 90% (mean 19%) of tumor cells. IDO1 reactivity was confined to neoplastic epithelium, the adjacent non-neoplastic acinar and ductal cells were negative. Non-neoplastic duodenal (n=12) and gastric epithelium (n=15), lacked PD-L1 and IDO1 reactivity.
In addition, among cases in which we examined whole sections (n=41), a single PD-L1-positive tumor was noted. Peritumoral pancreatic ducts were negative for PD-L1. Notably, the 3 cases that showed intratumoral neutrophils, simulating granulocytic epithelial lesions of type 2 AIP, were negative for PD-L1.
Validation of PD-L1 Stain in Type 2 AIP on EUS-guided Biopsies
To test the validity of our results on biopsies, we evaluated 37 EUS-guided pancreatic biopsies (Table 3). Eighteen patients were eventually diagnosed as type 1 or type 2 AIP, based on a combination of clinical, radiologic, and histologic features and a response to steroids.
The mean age of the 11 patients with the eventual diagnosis of type 2 AIP was 36 years (range: 12 to 57) with a near equal number of male and female patients (Table 3). On imaging 7 patients showed mass-like lesions, while the other 4 showed diffuse parenchymal abnormalities. Serum IgG4 was elevated in 2 cases (<1x upper limit of normal) and normal in the other patients.
Intact tissue cores were present in 5 cases; the other 6 biopsies were comprised of fragmented tissue admixed with blood. Histologically, 4 patients showed granulocytic epithelial lesions, supporting a diagnosis of type 2 AIP. The other 7 cases either lacked neutrophils or the neutrophils were admixed with blood (Figs. 3A, B, 4A). Although no malignant cells were seen, the possibility of an unsampled malignancy could not be excluded.
Four biopsies showed ductal membranous reactivity for PD-L1; in 4 cases the reactivity was confined to detached strips of epithelium (Figs. 3C, 4C). Notably, 2 biopsies negative for PD-L1 consisted of tiny fragments of tissue and lacked pancreatic ductal elements. In one case pancreatic ducts were negative for PD-L1 (case 11). Acinar, endocrine, and contaminating duodenal and gastric mucosa was negative for PD-L1. All cases showed fewer than 10 IgG4-positive plasma cells per high-power field. All patients showed a favorable response to steroids. Two patients had pancreas-based recurrence, both responded to additional immunosuppressive therapy. Of note, none of these patients showed evidence of extrapancreatic involvement, a finding characteristic of type 1 AIP and IgG4-related disease.
Notably, both mimics of type 2 AIP, granulomatosis, and polyangiitis (formerly Wegener) (Fig. 3D) and alcohol-related acute and chronic pancreatitis, showed granulocytic epithelial lesions, and were negative for PD-L1.
The 7 patients with type 1 AIP and 17 patients with PDAC were negative for ductal PD-L1 staining.
Sensitivity and Specificity
Collectively, the sensitivity and specificity of PD-L1 as a marker of type 2 AIP was 70% and 99%, respectively.
IDO1 and PD-L1 in Ulcerative Colitis
IDO1 positive staining was noted in the colonic epithelium in 19 of 25 cases (76%) (Supplementary Fig. 1, Supplemental Digital Content 1, http://links.lww.com/PAS/A792). However, colonic epithelium in all 20 biopsies was negative for PD-L1.
We identified pancreatic ductal PD-L1 expression in patients with type 2 AIP, a finding nearly absent in mimics of this disease. Although type 2 AIP also showed ductal IDO1 expression, one-third of PDACs also expressed this protein. Pancreatic ductal PD-L1 reactivity did not correlate with activity (ductal neutrophils), although an association with pancreatic fibrosis was noted. Pancreatic ductal PD-L1 reactivity was noted in 8 of 11 EUS-guided biopsies from patients with type 2 AIP and absent in inflammatory and neoplastic mimics, suggesting that the assay may serve as an ancillary test for the diagnosis of this disease. Theoretically, the PD-1/PD-L1 axis would predict a correlation between PD-L1, PD-1, and CD8, however, no such association was identified. Nevertheless, ductal PD-L1 and IDO1 expression could represent a mechanism to suppress the deleterious effects of the aberrant immune response in type 2 AIP.
The resected examples of type 2 AIP showed clinical evidence of this disease in the form of mass-forming lesion, obstructive jaundice, and bile duct and pancreatic duct strictures, mimicking pancreatic adenocarcinoma.4 The definitive diagnosis of type 2 AIP and its distinction from PDAC requires either characteristic histologic features or clinical evidence of inflammatory bowel disease.4,7 The incidence of inflammatory bowel disease in type 2 AIP ranges from 10% to 48%.4,8–11 Granulocytic epithelial lesions, the often-quoted pathognomonic type 2 AIP lesion, is neither specific nor entirely sensitive for this entity. As seen in this study, ductal abscesses may not be visualized in small and fragmented pancreatic biopsies. It is notable that biopsies from type 2 AIP tend to fragment (Figs. 3, 4) more often than those from type 1 AIP. In this series, granulocytic epithelial lesions were noted in only 4 of 11 (36%) cases. This is not surprising since granulocytic epithelial lesions are noted only focally. Notably, PD-L1 reactivity in type 2 AIP was noted in the absence of granulocytic epithelial lesions.
The distinction of contaminating duodenal and gastric epithelium from pancreatic ductal cells on EUS-guided biopsies may be challenging. Neutrophils admixed with blood, as seen in 2 cases in this series, may not be readily distinguished from tissue-based neutrophils. Furthermore, intratumoral and peritumoral neutrophils may be identified in PDAC and a proportion of PDAC are very well-differentiated. On patients with an EUS-guided biopsy we could not exclude the possibility of an unsampled malignancy. Finally, given the lack of a positive marker the diagnosis of type 2 AIP requires the exclusion of a wide range of pancreatic diseases including: (1) pancreatic neoplasms, (2) type 1 AIP, and (3) groove pancreatitis.
Access to a biomarker such as PD-L1, now routinely performed in most immunohistochemistry laboratories, has the potential to alter the current diagnostic algorithm, and facilitate a definitive diagnosis of type 2 AIP. The distinction of type 1 from type 2 disease is clinically relevant; the incidence and pattern of relapse of type 2 disease differs from type 1 disease.8 Type 2 disease is confined to the pancreas; relapse, although uncommon, is restricted to the pancreas. Conversely, the great majority of patients with type 1 AIP relapse, often at extrapancreatic locations. Ductal PD-L1 or IDO1 reactivity in patients with type 2 AIP, a finding not seen in type 1 disease, could assist in the distinction of the 2 diseases. These observations also highlight differences in the biological underpinnings of the 2 diseases, both currently placed in the category of AIP. Epithelial IDO1 reactivity (but not PD-L1) in ulcerative colitis, provides an additional link between type 2 AIP and inflammatory bowel disease.
The paucity of PD-L1 reactivity in PDAC and its absence in peritumoral tissue would support its use as an ancillary test in distinguishing type 2 AIP from this malignant neoplasm. Although 3% of PDACs are positive for PD-L1, the reactivity is confined to neoplastic epithelium. The incidence of PD-L1 reactivity in the current study is similar to that of the largest cohort (n=373) of PDACs (3.2%).14 PD-L1 reactivity in both studies was restricted to cytoplasmic membranes.14
A recent study, the first attempt to identify a biomarker for type 2 AIP, noted IL-8-positive lymphocytes and epithelial cells in type 2 AIP.15 Unlike PD-L1, IL-8 is not widely available in clinical laboratories. The authors suggest that IL-8 is a highly sensitive and moderately specific marker of this disease.15 Peritumoral pancreas from patients with PDACs, however, also exhibited IL-8 expression in the epithelium (3/12 cases, 25%) and inflammatory cells (10/12 cases, 83%); the overlap diminishes its efficacy as a biomarker. Cells co-expressing CD3 and IL-8 were restricted to type 2 AIP, although this assay is technically demanding.
There are several limitations that require mention, foremost among these is the lack of universal ductal PD-L1 reactivity in type 2 AIP; the sensitivity in resected cases was 70% and that for needle biopsies 73%, although the latter could be potentially higher if pancreatic ducts are included in the biopsy. It should be noted that PD-L1 reactivity is also noted in small caliber ducts and fragmented ducts. Although the specificity of the assay is high (99%), it would be prudent to emphasize the higher incidence of PDAC relative to type 2 AIP. PDACs also express PD-L1, albeit only in a significant minority of patients. That said, patients with type 2 AIP are generally younger with a low clinical suspicion for malignancy. Given the challenge in targeting PDAC on imaging, the lack of staining in peritumoral pancreas and contaminating gastric and duodenal epithelium, a near constant feature of EUS-guided biopsies, makes PD-L1 a promising marker for type 2 AIP. The relatively high incidence of IDO1 reactivity in PDAC precludes its use as a discriminatory assay in this setting.
In conclusion, we report ductal PD-L1 and IDO1 expression in type 2 AIP and IDO1 in ulcerative colitis, potentially implicating these immune modulatory proteins in the pathogenesis of these diseases. A PD-L1 stain could distinguish type 2 AIP from PDAC as well as discriminate between this disease and other forms of chronic pancreatitis. However, as with all biomarker studies, it would be prudent to validate these results in an independent cohort of EUS-guided biopsies. Finally, the results of the assay should be viewed, not in isolation, but in the larger clinical and histologic context.
1. Stone JH, Zen Y, Deshpande V. IgG4-related disease
. N Engl J Med. 2012;366:539–551.
2. Finkelberg DL, Sahani D, Deshpande V, et al. Autoimmune pancreatitis
. N Engl J Med. 2006;355:2670–2676.
3. Nagpal SJS, Sharma A, Chari ST. Autoimmune pancreatitis
. Am J Gastroenterol. 2018;113:1301–1309.
4. Hart PA, Levy MJ, Smyrk TC, et al. Clinical profiles and outcomes in idiopathic duct-centric chronic pancreatitis (type 2 autoimmune pancreatitis
): the Mayo Clinic experience. Gut. 2016;65:1702–1709.
5. Ectors N, Maillet B, Aerts R, et al. Non-alcoholic duct destructive chronic pancreatitis. Gut. 1997;41:263–268.
6. Notohara K, Burgart LJ, Yadav D, et al. Idiopathic chronic pancreatitis with periductal lymphoplasmacytic infiltration: clinicopathologic features of 35 cases. Am J Surg Pathol. 2003;27:1119–1127.
7. Shimosegawa T, Chari ST, Frulloni L, et al. International consensus diagnostic criteria for autoimmune pancreatitis
: guidelines of the International Association of Pancreatology. Pancreas. 2011;40:352–358.
8. Sah RP, Chari ST, Pannala R, et al. Differences in clinical profile and relapse rate of type 1 versus type 2 autoimmune pancreatitis
. Gastroenterology. 2010;139:140–148; quiz e112–e143.
9. Kamisawa T, Chari ST, Giday SA, et al. Clinical profile of autoimmune pancreatitis
and its histological subtypes: an international multicenter survey. Pancreas. 2011;40:809–814.
10. Detlefsen S, Zamboni G, Frulloni L, et al. Clinical features and relapse rates after surgery in type 1 autoimmune pancreatitis
differ from type 2: a study of 114 surgically treated European patients. Pancreatology. 2012;12:276–283.
11. Zamboni G, Luttges J, Capelli P, et al. Histopathological features of diagnostic and clinical relevance in autoimmune pancreatitis
: a study on 53 resection specimens and 9 biopsy specimens. Virchows Arch. 2004;445:552–563.
12. Llosa NJ, Cruise M, Tam A, et al. The vigorous immune microenvironment of microsatellite instable colon cancer is balanced by multiple counter-inhibitory checkpoints. Cancer Discov. 2015;5:43–51.
13. Lee LH, Cavalcanti MS, Segal NH, et al. Patterns and prognostic relevance of PD-1 and PD-L1
expression in colorectal carcinoma. Mod Pathol. 2016;29:1433–1442.
14. Liang X, Sun J, Wu H, et al. PD-L1
in pancreatic ductal adenocarcinoma: a retrospective analysis of 373 Chinese patients using an in vitro diagnostic assay. Diagn Pathol. 2018;13:5.
15. Ku Y, Hong SM, Fujikura K, et al. IL-8 expression in granulocytic epithelial lesions of idiopathic duct-centric pancreatitis (type 2 autoimmune pancreatitis
). Am J Surg Pathol. 2017;41:1129–1138.
autoimmune pancreatitis; IgG4-related disease; immunomodulatory proteins; PD-L1; IDO1
Supplemental Digital Content
Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.