One of the current criteria for clinical islet allotransplantation is ABO blood group compatibility between donor and recipient.1 The distribution of ABH antigens in human pancreata was first examined >40 y ago, revealing expression in acinar and vascular cells but not in islets.2 This underlined the importance of ABO matching in whole pancreas transplantation, but the situation in islet transplantation is more complicated. Although islets in situ appear to be ABH-negative,2 islets prepared for transplant are invariably contaminated with exocrine tissue including acinar cells. Furthermore, it is not known whether the islet isolation and transplantation process itself influence ABH expression. It is therefore somewhat surprising that ABH expression in islet preparations before and after transplant has not been characterized until now. In this issue of Transplantation, Verhoeff et al3 present a detailed immunohistochemical and flow cytometric survey of ABH antigen levels in human pancreatic tissue, isolated human clinical-grade islets, and islets derived from human embryonic stem cells (hESCs). After confirming that ABH antigens are present on acinar cells but not on endocrine and ductal cells in the pancreas, the authors demonstrate that islet preparations consist primarily of ABH-negative endocrine cells but are significantly contaminated with ABH-positive acinar cells. When islets were transplanted into immunodeficient rats, the endocrine component was ABH-negative at 24 wk, whereas remnant acinar tissue stained positive. Pancreatic endocrine progenitors derived from hESCs were ABH-negative and remained so after in vivo maturation following transplantation into rats, although this was not unexpected given that the hESC donor was blood group O. The human-to-immunodeficient rat islet xenograft model used in the latter part of the study does have some limitations. First, the kidney subcapsular site does not accurately reflect the clinical scenario, in which islets are transplanted intraportally.1 Subcapsular delivery co-localizes the islets with acinar tissue fragments and potentially exposes the islets to “bystander” injury and inflammation when anti-ABH antibodies attack the acinar cells. In contrast, intraportal delivery disperses islets and acinar fragments widely throughout the liver microvasculature, which probably minimizes co-localization and thus reduces the likelihood of bystander effects. Furthermore, subcapsular but not intraportal delivery shields the graft from exposure to anti-ABH IgM antibodies. Although these differences are not relevant in the immunodeficient rat model, they may well be relevant in the clinical setting. Second, the immunodeficient rat model does not allow investigation of the importance of the in vivo anti-ABH immune response to islet grafts. I have worked with one of the authors (L.J.W.) to develop a mouse model of anti-A antibody-mediated rejection of transgenic blood group A-expressing cardiac grafts.4 However, these A-transgenic donor mice are unsuitable for examining anti-A–mediated islet graft rejection because expression of the antigen is endothelial cell-restricted and absent on islets. Humanized mouse models may be more appropriate, as suggested by the authors. The take-home message of this interesting paper is that ABO matching is indeed likely to be important in islet allotransplantation—as is the case for HLA matching5—but may become less so if islet purification efficiency improves sufficiently or if hESC-derived islets are used. On the former point, the authors speculate that the restriction of surface expression of ABH antigens to acinar tissue (starkly illustrated by Figure 6A in the paper) provides an opportunity to enrich the purity of A, B, or AB islets by some form of negative selection. Although practicality at scale may be an issue, this is an intriguing proposition that is worthy of further consideration.