ABCD1 as a Novel Diagnostic Marker for Solid... : The American Journal of Surgical Pathology (2024)

Solid pseudopapillary neoplasm (SPN) is a rare, low-grade malignant pancreatic tumor that primarily affects women aged 20 to 30, comprising 1% to 3% of all pancreatic tumors.¹ At present, SPN diagnosis primarily relies on characteristic histomorphologic features, complemented by immunohistochemical staining. The main histological manifestations consist of solid and pseudopapillary structures, often accompanied by varying degrees of hemorrhage and pseudocysts. The nuclei are typically round to oval with subtle serrations and lack distinct nucleoli. Mitosis is infrequent.² However, relying solely on morphological criteria may result in confusion with pancreatic neuroendocrine tumors (NET), pancreatoblastoma (PB), and acinar cell carcinoma (ACC).

In more than 90% of cases, a point mutation in exon 3 of the CTNNB1 gene is present,^3,4 leading to increased nuclear accumulation of β-catenin. Immunohistochemical staining for β-catenin is routinely conducted for SPN diagnosis, with most SPN samples exhibiting nuclear positivity. Nevertheless, in practice, we have observed a similar β-catenin expression pattern in certain tumors with similar morphologies, such as NET, PB, and ACC, as well as some articles indicated.^5–8 In addition, heterogeneous nuclear/cytoplasmic β-catenin expression patterns were apparent, with some regions displaying strong nuclear β-catenin positivity while others remained nuclear negative. These variations may lead to false negatives for nuclear β-catenin in puncture biopsy samples, potentially resulting in missed or misdiagnosed SPN cases. Distinguishing nuclear staining patterns in the diagnostic process is challenging in some cases with strong cytoplasmic staining intensity. Therefore, there is a need for an easily recognized, sensitive, and specific diagnostic marker to complement existing methods and enhance the diagnostic accuracy and efficacy of SPNs.

The utilization of β-catenin as a diagnostic marker for SPN is rooted in the detection of CTNNB1 mutations in over 90% of SPN patients. Currently, gene expression datasets are readily accessible for SPNs.^4,6,9,10 To further explore diagnostic markers for SPN, it is imperative to investigate the transcriptome of SPNs, adjacent normal tissues, and neuroendocrine tumors (NET). In this study, we conducted a comparative analysis of mRNA expression profiles between SPNs, normal pancreas, and NETs.

Our primary objective was to identify highly upregulated pathways and explore potential diagnostic markers for SPN using immunohistochemical staining in a substantial patient cohort that includes SPNs and other tumors prone to confusion. Significantly, our findings conclusively suggest ABCD1 as a potential robust diagnostic marker for SPN in clinical use.

MATERIALS AND METHODS

Patient Cohort

A total of 111 primary SPNs, 16 metastatic SPNs, 9 acinar cell carcinomas, 3 PBs, 108 NETs (98 NF-NETs and 10 F-NETs), 54 pancreatic ductal adenocarcinomas (PDAC), 20 pancreatic serous cystadenomas (SCA), 19 pancreatic mucinous cystadenomas (MCA), 12 pancreatic ductal intraepithelial neoplasias (PanIN), and 5 intraductal papillary mucinous neoplasms (IPMN) were collected from patients at Peking Union Medical College Hospital between 2011 and 2023.

Clinical information for all patients is presented in Table S1, Supplemental Digital Content 1, https://links.lww.com/PAS/B774. Histological re-evaluation and patient record reviews were conducted by 2 pathologists, W.W.Z. and L.Y.H. Clinical and pathological data, including age at diagnosis, sex, disease recurrence/metastasis, histology, tumor size, and staging, were extracted from medical records. This study received approval from the Institutional Review Board of Peking Union Medical College Hospital (JS-2818).

Differential Gene Expression Analysis

We used the limma package to identify differentially expressed genes (DEGs) between 2 groups using the default parameter settings. Limma is a linear model-based method for differential expression analysis, enhancing analysis stability by incorporating shrinkage estimation and borrowing information across genes. For the differential expression analysis, we adhered to the instructions outlined in the limma package documentation while maintaining default parameters. To process the provided expression data, we initially normalized the raw read count matrix through a library size factor and log transformation. Following data normalization, we utilized the ImFit, eBayes, and top table functions within the R package limma to compute P-values for each gene. The DEGs were subsequently filtered based on a false discovery rate (FDR) cutoff of 0.05.

Gene Set Enrichment Analysis

We ranked all genes based on log₂FC values, which were calculated from corresponding gene expression levels between different phenotypes, namely SPNs versus normal, NETs versus normal, and SPNs versus NETs. Subsequently, the ranked gene lists were imported into the R package cluster Profiler to conduct gene set enrichment analysis (GSEA) using the GSEA function. This analysis focused on the GO biological process gene sets; specifically GOBP_PEROXISOME_ORGANIZATION, sourced from the Molecular Signatures Database (MsigDB). The results of the analysis were then visualized using the R package Enrichplot.

Staining

Antibodies and Reagents

Anti-catalase rabbit monoclonal antibody, Cell Signaling Technology, 12980S, using concentration of 1:200 for immunofluorescence staining.

BOND Ready-To-Use Primary Antibody Beta-Catenin (17C2), Leica, PA0083, using original reagents directly for immunohistochemistry staining.

Anti-ABCD1 rabbit monoclonal antibody, Abcam, ab197013, using concentration of 1:100 for immunohistochemistry staining.

Goat anti-mouse IgG H&L (Alexa Fluor 488), Abcam, ab150113, using concentration of 1:500 for immunofluorescence staining.

Goat anti-rabbit IgG H&L (HRP), Abcam, ab6721, using concentration of 1:200 for immunohistochemistry staining.

OptiView Amplification Kit, Roche, 860-099, using original reagents directly for immunohistochemistry staining.

Immunohistochemistry Staining

IHC assays were conducted using a fully automated immunohistochemistry stainer (Roche Ventana and Leica Bond III), with corresponding reagents, following the manufacturer’s instructions, except as noted. Tissue samples were fixed in 10% neutral formalin, paraffin embedded, cut into 4-µm thick consecutive sections and dewaxed as per standard procedures. Primary antibodies were appropriately diluted with a TBS diluent, as previously described, and dispensed into user-fillable containers on the automated IHC stainer.

Immunofluorescence Staining

Fresh tissue samples were fixed in 4% paraformaldehyde, embedded in OCT, and then cut into 4-µm thick consecutive frozen sections. Subsequently, the frozen tissue sections were incubated with a mixture of 10% fetal bovine serum and 0.2% Triton X-100 for 30 minutes to block nonspecific binding sites in the tissues and permeabilize cells. Primary antibodies were applied and incubated with the sections for 2 hours, followed by incubation with fluorescently labeled secondary antibodies for 1 hour at room temperature. DAPI was employed for nuclear staining. After each liquid change, the sections were thoroughly washed 3 times with PBS for 3 minutes while ensuring they remained moist throughout the process. To prevent fluorescence photobleaching, the sections were kept in the dark during the incubation with secondary antibodies. The antibodies and reagents used in this process were consistent with previous descriptions.

Western Blotting

Frozen SPN and normal pancreas tissues were hom*ogenized using a grinder in RIPA lysis buffer containing 1% proteinase inhibitor. The resulting lysates were collected and placed on ice for 1 hour, with mixing every 15 minutes. Subsequently, the lysates were centrifuged at 14,000 rpm for 10 minutes, and the supernatant was collected. Protein concentrations were determined using a BCA Protein Assay Kit. Calibrated protein samples were diluted with 5x loading buffer, separated in 10% SDS-page gels, and transferred to nitrocellulose membranes. Membranes were blocked with 5% skim milk powder at room temperature for 1 hour and incubated with primary antibodies overnight at 4°C. After being washed 3 times for 5 minutes each with TBST, the membranes were incubated with HRP-linked secondary antibodies at room temperature for 1 hour. Immunoblotting images were acquired by Tanon visualizer according to the manufacturer’s instructions. To serve as a loading control, β-actin was used.

Histoscore Analysis

Two pathologists, W.W.Z. and L.Y.H. independently evaluated all staining results using Histoscore (Hs), and they were blinded to the clinical data. The Histoscore is a semiquantitative assessment comprising 2 components: the intensity (I) of staining, which ranges from 0 (no staining) to 3 (strong), and the percentage (P) of cytoplasmic staining. The score was calculated using the following formula: Hs= (1 × [% weak staining]) + (2 × [% moderate staining]) + (3 × [% strong staining]).¹¹

Quantification of Immunohistochemical Staining of ABCD1 Expression

Immunohistochemistry sections were digitally scanned using a whole slide image scanner (NanoZoomer S360, C13220-01, bright field, X40, single layer, automated processing). Following scanning, 5 random regions were selected from each digital image, and FIJI software was utilized for the quantitative assessment of staining intensity within the region of interest. All images were imported into FIJI software and segmented into hematoxylin and DAB images. These images were subsequently converted into 8-bit format. Using a consistent threshold value, the integrated intensity of DAB images and the area of hematoxylin images were calculated. These values represented staining intensity and cell area, respectively. The ratio of these values was considered the average intensity of the region of interest. Quantitative results were analyzed using statistical methods, as previously described.

Statistics

We conducted statistical analysis using GraphPad Prism software (Version 8.0). Statistical significance was assessed through a 2-tailed Student t test for continuous variables. For categorical variables, the proportion of each histoscore and positive rate in immunohistochemistry staining images was analyzed using Fisher exact test. The threshold for statistical significance was set at P<0.05.

Receiver operating characteristic curve (ROC) was performed using MedCalc software.

RESULTS

Peroxisomes Are Enriched in SPN

We presented a schematic summarizing the experimental design of this study in Figure 1A. We analyzed a gene expression data set comprising 13 SPNs, 6 NETs, and 5 adjacent normal pancreases.¹⁰

The results of Gene Set Enrichment Analysis (GSEA) indicated that SPNs exhibited higher expression levels of genes related to peroxisome organization compared with normal pancreatic tissues (Fig. 1B). Catalase, a crucial antioxidant enzyme located in peroxisomes, plays a role in converting reactive oxygen species, such as hydrogen peroxide, into water and oxygen, thereby mitigating the toxic effects of hydrogen peroxide. Immunofluorescence staining for catalase further confirmed a significant expansion of peroxisomes in SPN tumor tissues compared to normal tissues (Figs. 1C, D).

Furthermore, transmission electron microscopy results provided additional verification of the increase in peroxisomes in SPN specimen relative to normal pancreatic tissue (Figs. 1E, F). Collectively, these findings suggest an elevated accumulation of peroxisomes in SPNs.

ABCD1 Exhibits Elevated Expression Level in SPN

We identified exclusive markers of SPN by analyzing differentially expressed genes (DEGs) between SPNs and normal samples. Among these genes, 1895 were upregulated (P≤0.05, Log₂FC≥1). In our evaluation of peroxisome genes, ABCD1 emerged as the top-ranking peroxisome gene in SPN when compared with both normal tissues (Log₂FC=1.622) (Fig. 2A) and NET samples (Log₂FC=2.057) (Fig. 2B). To validate these findings, we initially assessed the expression of ABCD1 protein in SPN tumor tissues and normal pancreas using immunoblotting. This analysis confirmed a marked upregulation of ABCD1 in SPNs (Fig. 2C), supporting the results obtained from transcriptome analyses. These findings suggest that ABCD1 holds promising potential as a robust and positive marker for SPN at both the mRNA and protein levels. We conducted a comprehensive evaluation of ABCD1 expression in SPNs using immunohistochemistry staining. Our cohort consisted of 72 primary SPNs, 16 metastatic SPNs, and 72 paired normal pancreas samples. Representative HE and ABCD1 IHC staining images are illustrated in Figure 2D. we applied the Histoscore (Hs) method to semiquantitatively measure ABCD1 expression levels. Samples were categorized into 4 grades: strong positive (+++, 2.5≤Hs≤3), moderate positive (++, 1.5≤Hs≤2.4), weak positive (+, 0.5≤Hs≤1.4), and negative (−, 0≤Hs≤0.4). Representative examples of each grade are illustrated in Figure 2E. On the basis of the score results, ABCD1 expression was strong positive (+++) or moderate positive (++) in SPN tumor tissues, either primary or metastatic, while it was negative or weak positive (+) in normal pancreatic tissues (+) (Figs. 2F, G)

Collectively, these findings demonstrate the elevated expression of ABCD1 in SPN and suggest its potential utility as a positive marker for SPN.

ABCD1 Specifically Identifies SPN From Morphologically Similar Pancreatic Neoplasms by IHC

In the realm of clinical diagnosis, the observation of uniformly hom*ogeneous round to oval cells presents a diagnostic challenge, as it is insufficient for the accurate discrimination of SPN, NET, ACC, and PB based solely on morphological characteristics. This holds particularly true when faced with SPN or cystic pancreatic tumors. Our primary objective was to investigate the potential utility of ABCD1 as a diagnostic marker for SPN, enabling its differentiation from these morphologically similar neoplasms.

We gathered a total of 111 SPN (primary site), 108 NET (98 nonfunctional NET, 10 functional NET), 9 ACC, and 3 PB samples for immunohistochemistry staining against ABCD1. The digital images of all samples are presented in Figure S1A, Supplemental Digital Content 2, https://links.lww.com/PAS/B775.

To analyze the staining results, we utilized the Hs scoring method. Representative images depicting different degrees of ABCD1 staining for NETs, ACCs, and PBs are shown in Figure 3A. Specifically, more than 95% of SPN samples exhibited strong positive staining (+++), while no similar strong positive staining results have been observed in other morphologically similar pancreatic tumors, most of which were negative or weakly positive (Figs. 3B, C). In addition, we used FIJI software to quantitatively assess ABCD1 staining intensity between SPNs and NETs, corroborating the conclusions derived from the Hs scoring analysis (Fig. S2, Supplemental Digital Content 3, https://links.lww.com/PAS/B776).

In summary, ABCD1 expression was notably higher in SPNs compared to NET, ACC, and PB samples, suggesting its potential as a marker for distinguishing SPN from these neoplasms.

ABCD1 Also Distinguishes SPN From Other Pancreatic Tumors by IHC

To further investigate whether ABCD1 could distinguish SPN from other pancreatic tumors, we conducted IHC staining on a set of common pancreatic tumor samples, including 54 cases of pancreatic ductal adenocarcinoma (PDAC), 5 intraductal papillary mucinous neoplasms (IPMN), 12 pancreatic ductal intraepithelial neoplasias (PanIN), 19 pancreatic mucinous cystadenomas (MCA), and 20 pancreatic serous cystadenomas (SCA). The scanned digital images of all samples are presented in Figure S1B, Supplemental Digital Content 4, https://links.lww.com/PAS/B777. We also utilized the Hs scoring method to analyze the staining results. Notably, 95.5% of SPN samples exhibited strong positive staining for ABCD1 (+++). In contrast, nearly all other pancreatic tumor types showed either negative (−) or weak positive (+) ABCD1 staining, with the exception of PDAC. In PDAC, only 3 out of 54 cases displayed moderate positive expression (++), while the remainder remained negative (−) or weak positive (+) (Figs. 4A–C).

Furthermore, upon reviewing cases with weak positive ABCD1 staining in these tumors, we observed unique staining patterns in MCA, PanIN, and IPMN. Specifically, only the tumor cells on the luminal side displayed weak positive ABCD1 staining, while cells on the basal side were ABCD1 negative (Fig. S3, Supplemental Digital Content 5, https://links.lww.com/PAS/B778). This pattern differed substantially from the uniform cytoplasmic staining observed in SPN. These findings underscore the utility of ABCD1 in distinguishing SPN from other common pancreatic tumors.

ABCD1 Performs Well as a Positive Marker for SPN Diagnosis

To assess the diagnostic potential of ABCD1 as a marker for SPN, we used receiver operating characteristic (ROC) curve analysis to determine sensitivity and specificity. Specifically, we evaluated the ability of ABCD1 histoscore and quantitated intensity to distinguish SPN from NET in a cohort of 111 SPNs and 108 NETs. In addition, we assessed the accuracy of ABCD1 histoscore in discriminating SPN from common pancreatic tumors encompassing all tumor types in our study.

For SPN and NET differentiation based on ABCD1 histoscore, the area under the curve (AUC) was 0.999 (95% CI: 0.982-1.000). The optimal diagnostic performance was achieved when the criterion value was set above histoscore 1 (weak positive, +), resulting in a sensitivity of 99.1% (95% CI: 95.1%-100%), specificity of 100% (95% CI: 96.6%-100%), and a Youden index of 0.9910. Importantly, even when utilizing X (above histoscore 0) as the criterion value to differentiate SPNs from NETs, the diagnostic efficacy remained high, with a sensitivity of 100% and specificity of 87.04% (Fig. 5A).

In terms of ABCD1 quantitated intensity, the AUC was 0.994 (95% CI: 0.970-1), with the highest diagnostic accuracy achieved at a criterion standard of intensity value of 0.109. This yielded a sensitivity of 97.98% and specificity of 94.51%, accompanied by a Youden index of 0.9249 (Fig. S4, Supplemental Digital Content 6, https://links.lww.com/PAS/B779).

For distinguishing SPN from all pancreatic tumors mentioned above (NET, ACC, PB, PDAC, SCA, MCA, IPMN, and PanIN), the AUC was 0.999 (95% CI: 0.987-1.000). Optimal diagnostic performance was obtained when the criterion value was set above histoscore 1 (weak positive, +), resulting in a sensitivity of 99.10% (95% CI: 95.1%-100%), specifically of 97.83% (95 CI: 95.0%-99.3%), and a Youden index of 0.9693. When using ABCD1-positive staining (above histoscore 0) as the criterion value to differentiate SPN from other pancreatic tumors, the diagnostic efficacy remained high, with a sensitivity of 100% and specificity of 81.3% (Fig. 5B).

In conclusion, our findings highlight the potential of ABCD1 as a valuable diagnostic marker for distinguishing SPN.

DISCUSSION

In our study, we compared the transcriptomes of SPNs, NETs, and normal pancreatic tissues, revealing significant upregulation of the peroxisome organization pathway in SPNs. We validated this enrichment by conducting immunofluorescence staining for catalase and transmission electron microscopy. To assess ABCD1’s potential as a diagnostic marker for SPN, we conducted IHC staining in a comprehensive cohort, including primary and metastatic SPN, NET (NF-NET and F-NET), ACC, PB, PDAC, PanIN, SCA, MCA, and IPMN samples. Our results demonstrated pronounced ABCD1 protein enrichment in SPNs (both primary and metastatic sites), while other tumors exhibited either no or low ABCD1 expression levels. Diagnostic efficacy analysis further affirmed ABCD1’s value as a positive marker for SPN diagnosis.

Although nuclear β-catenin is commonly used for SPN diagnosis, CD10, cyclin D1, vimentin, progesterone receptor, glutamine synthase, TFE3, SOX11, LEF1, FUS, WIF-1, CD56, CD138, CD200, CD99, and E-cadherin have also been explored as diagnostic markers.^6,10,12–17 However, their sensitivity and specificity when used alone vary. Notably, some SPN cases exhibit negative staining for these markers. In addition, using nuclear β-catenin alone for SPN diagnosis presents challenges, such as variable nuclear staining intensity within SPN samples and its presence in other tumors with similar morphology. These underscores the need for new markers that are specific, sensitive, and easy to work with.

As a rare pancreatic tumor, distinguishing SPN from other pancreatic tumor types is challenging. Our findings indicated increased ABCD1 expression at both primary and metastatic SPN sites compared to normal pancreas. In contrast, ABCD1 was negative or weakly positive in NET, ACC, and PB, as well as most other pancreatic tumors. Tumor samples with over moderate intensity diffused ABCD1 staining were highly indicative of SPN, especially in cases with partially nuclear-negative β-catenin or unclear nuclear expression due to deep cytoplasmic staining. ABCD1 IHC staining on 16 metastatic SPN samples consistently showed strong positive staining, suggesting its utility in identifying SPN-origin metastasis in patients with a history of SPN. While multicenter validation is warranted to further explore its feasibility in clinical diagnosis.

To assess ABCD1 as a supplementary diagnostic marker for SPN alongside β-catenin, we analyzed ABCD1 expression in nuclear β-catenin-negative tumor cells. Although none of our 111 SPN cases was found completely negative for nuclear β-catenin staining, there were some cases with negative or obscure nuclear β-catenin staining in some scattered areas, where the corresponding ABCD1 staining is consistently positive (Fig. S5A, Supplemental Digital Content 7, https://links.lww.com/PAS/B780). This suggests that pathologists may use ABCD1 alongside β-catenin, particularly in puncture biopsy samples, where nuclear β-catenin may be negative in some areas. Further research on fine needle aspiration and cytological specimens is needed. In addition, we observed negative ABCD1 staining in some β-catenin-positive NET samples (Fig. S5B, Supplemental Digital Content 7, https://links.lww.com/PAS/B780), highlighting ABCD1’s potential in distinguishing SPN from morphologically similar pancreatic tumors to avoid false positive diagnoses. Future studies should explore the combined performance of ABCD1 and β-catenin in SPN diagnosis.

In our cohort, 95.5% of SPN cases were strongly positive for ABCD1 (+++), with none showing negative staining. Although ABCD1 was also expressed positively in some morphologically similar tumors, it lacked the strong positive pattern seen in SPN. This straightforward interpretation of ABCD1 staining primarily based on intensity and area comparison makes it suitable for automated quantification using software in the future.

In conclusion, our work demonstrates the specific enrichment of peroxisomes and high ABCD1 expression in SPN compared with other pancreatic tumors. Our exploration of ABCD1 as a diagnostic marker will aid pathologists in specifically, sensitively, and easily distinguishing SPN in clinical practice.

ACKNOWLEDGMENTS

The authors thank Liu Lulu, Zhong Qing, Hao Jianyu, and Cai Zian of the Cell Biology Core Facility of National Science and Technology Key Infrastructure on Translational Medicine, Li Mengjie in the department of General Surgery in Peking Union Medical College Hospital for assistance with the instruments and sample collection in this research.

REFERENCES

Supplemental Digital Content

• PAS_2024_02_16_WU_AJSP-D-23-00342R2_SDC1.docx; [Word] (20 KB)
• PAS_2024_02_16_WU_AJSP-D-23-00342R2_SDC2.tif; [Other] (9.59 MB)
• PAS_2024_02_16_WU_AJSP-D-23-00342R2_SDC3.docx; [Word] (97 KB)
• PAS_2024_02_16_WU_AJSP-D-23-00342R2_SDC4.tif; [Other] (5.74 MB)
• PAS_2024_02_16_WU_AJSP-D-23-00342R2_SDC5.docx; [Word] (395 KB)
• PAS_2024_02_16_WU_AJSP-D-23-00342R2_SDC6.docx; [Word] (146 KB)
• PAS_2024_02_16_WU_AJSP-D-23-00342R2_SDC7.docx; [Word] (1.25 MB)