Pax8 Detection in Well- Differentiated Pancreatic Endocrine Tumors: How Reliable is it? To the Editor: One of the most challenging areas in pancreatic endocrine tumor (PET) biology is the identification of specific biomarkers that can morpho- logically and histochemically distin- guish well-differentiated PETs from other types of neuroendocrine tumors (NETs). Novel markers could im- prove diagnosis accuracy and could consequently lead to a more appro- priate and tailored treatment and therefore to a better prognosis. This is highly relevant from the clinical point of view as at least 10% of patients with NETs have a wide- spread disease of unknown primary origin, and available antineoplastic agents are fundamentally effective only on tumors of pancreatic ori- gin. 2,3 Unfortunately, current mar- kers including synaptophysin and chromogranin A are unable to distin- guish whether a particular NET originated initially from the endocrine pancreas or from other sites such as the gastrointestinal tract or the bron- chopulmonary system. In this regard, a recent study published in the Amer- ican Journal of Surgical Pathology has identified the transcription factor Pax8 as a potentially new biomarker capable of performing such a task. 7 Although previously not detected in the pancreas, 4 Long and colleagues reported strong Pax8 immunostaining in normal human islets and have suggested that this transcription fac- tor could be a specific marker of mature pancreatic endocrine cells and could even be implicated in terminal differentiation of islet cells during development. This premise was corroborated by a second Amer- ican Journal of Surgical Pathology publication confirming the expression of Pax8 in healthy human islets. 6 The discovery of Pax8 expression in the endocrine pancreas led Long and colleagues to assess whether Pax8 was expressed in PETs and in other well-differentiated NETs and to eval- uate its potential value as a specific biomarker. They showed that a sig- nificant percentage of primary well- differentiated PETs displayed strong Pax8 staining, whereas ileal and pulmonary carcinoid tumors were negative. Furthermore, liver meta- stases of PET origin were often Pax8 positive. The investigators concluded that Pax8 could be a useful marker to determine the primary origin site of metastatic well-differentiated PETs lodged within the liver. 7 Interestingly, an independent study substantiated these findings and proposed a flow- chart in which Pax8 was included in a panel of immunohistochemical stain- ings to determine the site of origin of metastatic well-differentiated NET in the liver. 11 Unfortunately, we would like to communicate to the American Journal of Surgical Pathology readership that the conclusion of these studies is potentially inaccurate and that Pax8 proteins are not detectable in either healthy islets or PETs. 8 Indeed, sub- sequent to the publication by Long et al 7 we embarked on a study to determine whether Pax8 could also be expressed during pancreatic islet de- velopment, thereby pinpointing to a yet unknown function of this tran- scription factor in islet physiology. To this end, we analyzed Pax8 expression during mouse pancreatic development and in adult islets by immunohisto- chemistry using the same Pax8 poly- clonal antibody (ProteinTech Group Inc., Chicago, IL; Cat. No. 10336-1- AP) that Long et al 7 used in their study. It is noteworthy that this polyclonal antibody was also used for most, if not all, clinical studies that demonstrated Pax8 staining in PETs. 5,6,11 Consistent with these pub- lished reports we were successful in detecting Pax8 immunostaining in mature mouse and human islets and in the developing mouse pancreas. However, in an attempt to comple- ment our immunohistochemical data with transcript expression profile, we unexpectedly found that Pax8 mRNA levels in both human and mouse islets were low to undetectable. These find- ings led us to question the specificity of the Pax8 polyclonal antibody. In- deed, as the 9 Pax members share a high degree of homology at the N-terminal paired domain, the region included in the peptide use for orig- inating the antibody, a potential cross-reactivity of the antibody to another Pax member could be envis- aged. This premise was indeed con- firmed by demonstrating that the polyclonal Pax8 antibody also recog- nized the islet-enriched Pax6 protein both by Western blotting and by immunohistochemistry. 8 To circum- vent this caveat, a novel Pax8 mono- clonal antibody from Abcam (No. ab53490) generated against the more variable C-terminal region of Pax8 was used to reevaluate the expression pattern of Pax8 during development and in islets. This antibody recognizes the 2 predominant isoforms of Pax8 expressed in most tissues and tu- mors. 10 Consistent with our QT-RT- PCR data, neither the developing pancreas nor mature islets exhibited staining with this novel Pax8 mono- clonal antibody as compared with the polyclonal antibody. 8 These results led us to reassess whether Pax8 was truly present in PETs. To this end, serial sections of PETs were immunostained with Pax8 antibodies (polyclonal and monoclo- nal) and with anti-Pax6 sera (Fig. 1). 8 Immunohistochemical analysis re- vealed positive staining both in non- neoplastic islets and in tumor areas of pancreas sections when the Pax8 polyclonal antibody was used, where- as no staining was revealed with the Pax8 monoclonal antibody (Figs. 1A, B). An immunostaining pattern sim- ilar to that of the Pax8 polyclonal antibody was discerned with the Pax6 antibody, substantiating cross-reac- tivity of the commonly used Pax8 polyclonal antibody with Pax6 and possibly with other members of the Pax family (Fig. 1C). 8 Thus, in con- trast to the study by Long et al, 7 we LETTER TO THE EDITOR 1906 | www.ajsp.com Am J Surg Pathol  Volume 35, Number 12, December 2011