Introduction

Neuroendocrine neoplasia of the pancreas (pan) and especially the small intestine (si) is rare [1, 2], frequently has a long asymptomatic course [3, 4], and remains undiagnosed until radiological examinations are done due to diffuse abdominal symptomatology [5].

As shown recently, 43 (23.8%) of 181 patients with NEN located in various sites of the gastroenteropancreatic tract were diagnosed at presentation with liver metastasis. In five (1.8%) additional patients, the primary could not be located. Even functional imaging failed to identify the primary in these patients [1].

The hint to the origin of the “unknown primary” may occasionally be based on biopsies of the liver applying specific immunohistochemical markers [6]. Insulin gene enhancer protein Islet-1 (ISL-1), thyroid transcription factor 1 (TTF-1), serotonin, and homeobox protein CDX-2 (CDX2) are transcription factors that can be expressed in pan and siNENs [7, 8].

To our knowledge, no systematic studies on differences regarding the expression of those markers in liver metastases of either entity have so far been published.

Methods

Fifty patients with siNENs of the ileum (20 females, 30 males; mean age 61, range 37–81) and 37 patients with panNENs (21 females, 16 males; mean age 54, range 25–81) and synchronous liver metastasis were retrospectively analyzed. The metastases were evaluated histologically, and immunohistochemical staining of chromogranin A, synaptophysin, Ki-67, ISL-1, TTF-1, CDX2, and serotonin was performed.

Tumor tissue was routinely formalin-fixed and paraffin-embedded. Hematoxylin and eosin staining was applied to 3-μm sections of each block. Immunostainings against proliferation marker Ki-67 antigen (Molecular Immunology Borstel 1 [MIB-1] mouse monoclonal, Novocastra, Newcastle, UK, dilution 1:20), CDX2 (1H9 mouse monoclonal, abcam, Cambridge, UK, undiluted), ISL-1 (1H9 mouse monoclonal, abcam, Cambridge, UK, dilution 1:400), TTF-1 (SP141 rabbit monoclonal, Ventana, Tucson, Arizona, USA, undiluted), and serotonin (5HT-H209 mouse monoclonal, DakoCytomation, Denmark, dilution 1:100) were performed using an automatic immunostainer (Ventana Medical Systems Inc., BenchMark® or BenchMark® ULTRA, Tucson, Arizona, USA). For antigen retrieval, slides for Ki-67, CDX2, ISL-1, and TTF-1 staining were boiled with a commercially available puffer (Ventana Medical Systems Inc., Cell Conditioning 1, Tucson, Arizona, USA) for 256, 256, 64, and 180 min, respectively. A commercially available amplification kit (Ventana Medical Systems Inc., Amplification Kit, Tucson, Arizona, USA) was used for Ki-67 staining. Protease 1 pretreatment (for 8 min) was applied for serotonin staining.

The histologies of all cases were reviewed by one experienced pathologist (O.K.), classified according to the WHO classification outlined in 2017 [9, 10] and staged according to the European Neuroendocrine Tumor Society guidelines of 2006 and Union international contre le cancer 2010 [9,10,11]. For grading, the Ki-67 labeling index with MIB-1 antibody was used and the Ki-67 index was assessed in 500 tumor cells in areas in which the highest nuclear labeling was observed using an eye grid ocular. Expression of ISL-1, TTF-1, CDX2, and serotonin was evaluated semiquantitatively as follows: no tumor staining (−), weak: staining in ≤ 10% (+), moderate: staining in > 10 and < 100% (++), and strong: 100% of cells (+++).

Results

All 87 liver metastases were of neuroendocrine origin confirmed by both synaptophysin and chromogranin A expression. The primary tumor was detected with further evaluations of the patients (functional imaging, endoscopic ultrasound with biopsy), and adequate treatment was performed as feasible. The immunohistochemic profile of the primary was identical to the liver metastasis in all cases. The details of the staining patterns are summarized in Table 1.

Table 1 Staining pattern in siNENs and panNENs
Full size table

CDX2

The stains for CDX2 were at least moderately positive in 92% of the siNENs and negative in 92% of the panNENs.

ISL-1

ISL-1 was negative in all siNENs, whereas 75% of the pan tumors showed at least a moderate stain.

Serotonin

Serotonin showed strong positive stains in 90% of the siNENs, whereas it was moderately positive in only one panNEN (2.7%).

TTF-1

Immunostaining for TTF-1 was negative in most siNENs (96%) and panNENs (92%).

Combined Staining

Positive staining of CDX2 and serotonin yielded an almost 100% specificity for the siNENs. Photomicrographs of the typical staining pattern of liver metastasis of an ileal NEN compared to a pancreatic NEN are shown in Fig. 1.

Fig. 1
figure 1

The top row shows the typical staining pattern of a neuroendocrine liver metastasis of pancreatic provenience with strong positivity of ISL-1, whereas the bottom row illustrates the staining pattern of a liver metastasis of an ileal NEN, yielding positivity for serotonin and CDX2

Full size image

Discussion

Patients with NENs frequently share a long way of suffering due to uncertainty. Unspecific abdominal symptoms lead to radiological examinations of liver lesions, and biopsy may evidence NENs. The journey then continues, as the primary is still missing. As shown recently [1], at least 2% of all NENs are diagnosed with an unknown primary. In this study, the expression of various immunohistochemical markers in the liver metastases of subjects with NEN was examined and the frequency of the positivity of the set of markers was analyzed. Retrospectively, the staining quality of the markers was correlated with the sites of the NEN, by definition being either the si or the pan. The results of the analysis applying a specific set of neuroendocrine markers on liver biopsies should help to gain information concerning the possible sites of the NEN in patients with unknown primary but liver metastasis.

Serotonin has been proposed to be a potent marker for siNENs in biopsies [6]. In the current series, however, serotonin was positive in 90% of the siNENs and only moderately positive in one panNEN, thus being consistent with the literature [12] and resulting in high levels of sensitivity and specificity. While CDX2 has been reported to be positive in almost all siNENs [13,14,15], we can only confirm that it was at least moderately positive in 90% in our series. CDX2 was negative in over 90% of the panNENs, thus showing higher specificity than previous findings [13, 16] and suggesting that CDX2 is a key marker (in combination with serotonin) to differentiate siNENs from panNENs.

ISL-1 was seen to be an ideal marker to exclude siNENs, as no positive stain was found in our specimens. In line with other authors’ findings [17,18,19], however, panNENs can neither be excluded nor confirmed with this single marker, as more than 20% of the liver metastases of panNENs in our series were immunohistochemically negative for ISL-1.

TTF-1 should not be considered a valuable marker to distinguish between siNENs and panNENs, as the vast majority of liver metastases of either entity were negative for TTF-1 labeling. Still, it remains highly useful to discriminate gastroenteropancreatic NENs from those of the lung [20].

In the current study, the combination of these markers was applied to differentiate siNENs with ileal primaries from panNENs. As demonstrated, the combination of positivity for CDX2 and serotonin in liver metastasis showed an almost 100% level of specificity for NENs of si origin. Therefore, (at least) this combination of markers should be used when considering si provenance in patients with liver metastasis and an unknown primary. Furthermore, functional imaging should be applied in any patient with a metastatic NEN. Therefore, immunohistochemical markers must always be seen in connection with clinical and diagnostic findings to serve the patient best. A single marker cannot replace the diagnostic algorithm in complex neuroendocrine tumors.