A nanobody-based tracer targeting DPP6 for non-invasive imaging of human pancreatic endocrine cells

There are presently no reliable ways to quantify endocrine cell mass (ECM) in vivo, which prevents an accurate understanding of the progressive beta cell loss in diabetes or following islet transplantation. To address this unmet need, we coupled RNA sequencing of human pancreatic islets to a systems biology approach to identify new biomarkers of the endocrine pancreas. Dipeptidyl-Peptidase 6 (DPP6) was identified as a target whose mRNA expression is at least 25-fold higher in human pancreatic islets as compared to surrounding tissues and is not changed by proinflammatory cytokines. At the protein level, DPP6 localizes only in beta and alpha cells within the pancreas. We next generated a high-affinity camelid single-domain antibody (nanobody) targeting human DPP6. The nanobody was radiolabelled and in vivo SPECT/CT imaging and biodistribution studies were performed in immunodeficient mice that were either transplanted with DPP6-expressing Kelly neuroblastoma cells or insulin-producing human EndoC-βH1 cells. The human DPP6-expressing cells were clearly visualized in both models. In conclusion, we have identified a novel beta and alpha cell biomarker and developed a tracer for in vivo imaging of human insulin secreting cells. This provides a useful tool to non-invasively follow up intramuscularly implanted insulin secreting cells.

sensitivity (in the pico/nanomolar-range) 10 , a (sub)millimeter spatial resolution and proven performances in translational models with a growing number of tracers 11 .
To identify and develop novel tracers for ECM, we used a systems biology approach to mine the human pancreatic islet transcriptome for suitable islet biomarkers 12 . This approach then based on array analysis has allowed us to identify a beta cell specific biomarker, namely FXYD2γa 13 . We have now identified, based on RNA sequencing, a novel ECM biomarker that is expressed on the cell surface of pancreatic endocrine cells, namely dipeptidyl peptidase 6 (DPP6). We next developed a nanobody-based tracer targeting DPP6. Nanobodies are the variable domain derivatives of homodimeric heavy chain-only antibodies occurring naturally in camelidae. These small (13)(14) polypeptides display unique features in respect to monodispersity, immunogenicity, stability, and versatility 14 ; they are amenable for a wide range of radiolabeling technologies [15][16][17] and have already been used for imaging purposes by SPECT or PET in both animal models of cancer 16,17 , immunity 18,19 or atherosclerosis 20 and in clinic 21 . We now show that they can also be used to successfully image human insulin secreting cells implanted into the muscle of immunodeficient mice, without any pre-manipulation or loading of the transplanted cells.

Results
Discovery of DPP6 as an ECM-enriched gene transcript. We used a RNA sequencing-based system biology approach to identify ECM and beta cell targets 12,13 (Fig. 1). The identification of DPP6 was based on RNAsequenced human pancreatic islets, treated and untreated with IL-1β and IFN-γ, and on a comparison with 16 normal human tissues (ref. 12 , Illumina Body Map 2.0:GSE30611) (Fig. 2). The DPP6 was preferentially expressed in human pancreatic islets, with a mean expression of 31 ± 8 reads per kilobase of transcript per million mapped reads (RPKM) (n = 5), several-fold higher than in other tissues, except brain ( Fig. 2A). Expression of DPP6 mRNA was not modified by proinflammatory cytokines in human pancreatic islets ( Fig. 2A) or by the saturated free fatty acid palmitate 22 . Furthermore, exposure of 5 human islet preparations for 24h to 28 mM glucose, as compared to 6.1 mM glucose (human preparations and experimental conditions as described in ref. 23 ) did not significantly changed DPP6 expression: (qPCR corrected per actin ×10 3 ), human islets at 6.1 mM glucose: 7 ± 3; human islets at 28 mM glucose: 8 ± 2 (mean ± SEM; n = 5). We have also checked expression of DPP6 in laser captured human islets obtained from type 2 diabetic patients and respective controls, as studied by microarray analysis (data from 24 ). This is a more pathophysiological relevant condition, where human islets are chronically exposed to metabolic stress. The data obtained (mean ± SEM; n = 10) in respective Controls and T2D are, 539 ± 46 and 445 ± 32 (n = 10) again did not show a significant difference between groups. As a whole, the above information indicates that neither inflammation-nor metabolic-induced stress significantly modifies DPP6 expression in human islets. The step-by-step approach used to identify new endocrine cell biomarkers. Schematic overview of the approach taken to mine for new endocrine cell biomarkers in the transcriptome of human islet preparations (n = 5) analysed by RNA sequencing 12 under both control condition and following treatment with proinflammatory cytokines (IL-1β + IFN-γ). Enriched pancreatic islet specific transcripts were identified by comparing transcriptomes of human pancreatic islets against 16 different normal human tissues. IPA: ingenuity pathway analysis, http://www.ingenuity.com/products/ipa. DPP6-001 (ENST00000404039.5) was the main DPP6 splice variant present in human pancreatic islets (Fig. 2B), and this splice variant showed a higher expression as compared to other tissues evaluated, again except for brain (Fig. 2C). Analysis of available databases of purified human endocrine cells 25,26 indicated that DPP6 is expressed both in beta cells (84 ± 26 transcripts per million (TPM)) and alpha cells (61 ± 20 TPM) 25 , but not in acinar or ductal cells 26 . Expression of DPP6. The RNA-sequencing data was confirmed by qPCR in 14 different human tissues as compared to human pancreatic islets and in insulin-producing EndoC-βH1 cells treated or not with cytokines ( Fig. 3A), as well as human exocrine material and two exocrine cell lines Capan-2 and PANC. In the pancreatic islets DPP6 expression was nearly 20-fold higher (p ≤ 0.05) than in the exocrine material (which contained around 1-2% beta cells), while both Capan-2 and PANC cells had at least 4000 times lower DPP6 expression. There was again a higher expression of DPP6 in human pancreatic islets as compared to the surrounding extra-pancreatic tissues, and DPP6 expression was not modified by proinflammatory cytokines (Fig. 3A). The DPP6 expression was even higher in insulin-producing EndoC-βH1 cells (Fig. 3A), a finding confirmed by histology ( Fig. 3C-E), which also indicated a cell surface localization of the biomarker (Fig. 3F). In these cells, the expression of DPP6 mRNA increased by 1.5-fold (p ≤ 0.05) after exposure to proinflammatory cytokines IL-1β and IFN-γ (Fig. 3A), but this increase was not observed at the protein level (Fig. 3B). The lack of cytokine-induced changes in human islets (mRNA level) or in EndoC-βH1 cells (protein level) suggests that the inflammation prevailing in T1D will not modify the expression of this biomarker.
Detection of DPP6 by immunohistochemistry of human pancreas. The histological localization of DPP6 on normal human pancreata was investigated using a commercial mAb targeting the protein (Fig. 4). DPP6 co-localized with insulin ( Fig. 4D,J) and glucagon positive cells ( Fig. 4I) but not with somatostatin positive cells (Fig. 4E). DPP6 immunoreactivity was present in 90 ± 3% of insulin cells and 74 ± 10% of glucagon cells (n = 3). There were only rare cells positive for DPP6 in the exocrine pancreas, with no clear evidence for staining in acinar or duct cells ( Supplementary Fig. 1). Analysis of pancreata obtained from patients with long-term type 1 diabetes Expression pattern of DPP6 splice variants in human pancreatic islets exposed or not to IL-1β + IFN-γ (cyto); human pancreatic islets express mainly the DPP6-001 variant, (n = 5); (C) Expression of the isoform DPP6-001 in human pancreatic islets exposed or not to IL-1β + IFN-γ (cyto), (n = 5) as compared to other human tissues; the highest extra-pancreatic expression is seen in brain, colon and thyroid.
(n = 3) showed nearly complete loss of insulin positive cells (Fig. 4N), while some DPP6 (Fig. 4L) and glucagon positive cells (Fig. 4M) remained. Pancreata from T1D patients had a reduced DPP6 positive area when compared to the healthy controls (p ≤ 0.05) (Fig. 4K). Interestingly, DPP6 expression was also clearly present in the insulin positive cells from two separate human insulin producing tumors (insulinomas) (Supplementary Fig. 2).
Generation and in vitro characterization of nanobodies against DPP6 and selection of the lead nanobody 4hD29. Anti-DPP6 nanobodies were generated by immunization of a dromedary with a recombinant protein, derived from the extracellular domain of the HsDPP6 transmembrane protein. Biopannings were subsequently performed of a phage-displayed immune nanobody library on the recombinant protein, and selection of binders by phage-ELISA. Thirteen nanobodies with different protein sequences were identified as binding to HsDPP6 recombinant protein and divided into 9 families based on the homology of their complementarity determining region (CDR) 3. Among these nanobodies ( Supplementary Fig. 3J) eight were successfully produced and purified as hexahistidine-tagged proteins from E. coli fermentation cultures and the nanobodies' affinities were analysed using Surface Plasmon Resonance (SPR) (Supplementary Fig. 3A-H). The best scoring nanobody, designated 4hD29, had an affinity (K D value) towards DPP6 of 1.2 nmol/l ( Supplementary Fig. 3H,I). Subsequently, the 8 nanobodies were screened for their recognition of the transmembrane DPP6 on cells using flow cytometry. Nanobody 4hD29 (Fig. 5A) and 5 other nanobodies ( Supplementary Fig. 4), but not the control non-specific nanobody, showed clear binding to HsDPP6 transiently-transfected CHO cells (Fig. 5A). None of these nanobodies bound to wild type CHO cells (Fig. 5B), except for nanobody 3hD36 that also non-specifically stained untransfected cells (data not shown). Despite a 93% homology between human and murine DPP6 in their extracellular region, 4hD29 did not recognized murine DPP6 in transiently-transfected CHO cells (data not shown). Next, we evaluated the nanobody candidates using the human Kelly neuroblastoma cell line, which expresses DPP6 27 . 4hD29 and 6 other nanobodies, but not the control non-specific nanobody, bound to Kelly cells (p ≤ 0.01; n = 4 for 4hD29) (Fig. 5C,H, Supplementary Fig. 5).   Based on these flow cytometry experiments and the SPR data, nanobody 4hD29 was selected as the lead compound and its binding was further characterized on EndoC-βH1 cells where 4hD29 binding could be determined (p ≤ 0.01; n = 5) (Fig. 5D,I). Finally, 4hD29 bound significantly (p ≤ 0.05; n = 4) to human endocrine tissue (Fig. 5E,J) as compared to exocrine tissue (Fig. 5F,J), where no 4hD29 binding could be observed. Endo-and exocrine tissue was differentially gated as shown in Fig. 5G.

Biodistribution and in vivo imaging of mice xenografted with human cells expressing DPP6.
We next investigated whether the lead nanobody 4hD29 can be used as an in vivo imaging diagnostic tool to monitor DPP6-expressing cells. For this purpose, two humanized mouse models were used, namely immunodeficient mice with s.c. Kelly tumours grown for about 2-weeks or implanted intramuscularly with EndoC-βH1 cells in rubber rings (empty rings were used as negative controls; see Methods). Implanted EndoC-βH1 cells grew slowly, and only generated palpable tumours after 9-10 weeks, while the mice had a normal growth ( Supplementary  Fig. 6A,D). At this stage, the mice became hypoglycemic, which was partially counteracted by addition of 20% glucose in the drinking water ( Supplementary Fig. 6B,E) and high levels of human C-peptide were detectable in serum samples ( Supplementary Fig. 6C,F). Using these models, microSPECT/CT imaging and ex vivo biodistribution analyses were done respectively 60 and 80 minutes after i.v. administration of 99m Tc-labelled nanobody 4hD29 or negative control nanobody BcII10 (Figs 6 and 7).
Based on the biodistribution analyses (Fig. 6A,B) and the images (Fig. 7) in both mouse models, 99m Tc-4hD29 displayed a 2-fold higher uptake over 99m Tc-BcII10 in the Kelly neuroblastoma tumour (1.2 ± 0.10 %IA/g) vs. (0.5 ± 0.1 %IA/g) (Fig. 6C) and in the EndoC-βH1 tumour (1.0 ± 0.1 %IA/g) vs. (0.5 ± 0.04 %IA/g) (Fig. 6G). Both radiolabelled 4hD29 and control nanobodies showed a high uptake in the kidneys, bladder and a fast wash-out from blood and other non-targeted tissues. Tumour 4hD29 radioactivity levels were specific, since tumour-to-blood and tumour-to-muscle ratios were respectively 2.9 ± 0.2 and 9.9 ± 3.2 in neuroblastoma-bearing mice (Fig. 6D,E) and 2.5 ± 0.4 and 9.9 ± 2.2 in EndoC-βH1-bearing mice (Fig. 6H,I), values that were respectively 4.3-, 3.1-, 4.6-and 3.3-fold higher than those observed for the control nanobody. Tumour-to-liver ratios were 2.6 ± 0.6 for Kelly tumours (Fig. 6F) and 1.1 ± 0.3 for the EndoC-βH1 grafts (Fig. 6J). Importantly the high tumour-to-background levels of the lead anti-DPP6 nanobody in the dissection analyses were translated into a clear visualization of tumours in fused SPECT/CT images ( Fig. 7 and Supplementary video 1). Indeed, in vivo administration of 99m Tc-4hD29 (but not 99m Tc-BcII10) allowed visualization of both the Kelly tumour (Fig. 7A) and the EndoC-βH1 transplant (Fig. 7B and Supplementary videos 1 and 2) with high contrast. Image quantification in the latter mice (Fig. 7C) showed a 6.8-fold higher signal of 99m Tc-4hD29 in the EndoC-βH1 transplant as compared to the empty vehicle ring and a 3.7-fold higher signal than that observed in the EndoC-βH1 transplant following injection of the radiolabelled control nanobody. Similar results were obtained with 111 In-labelled nanobodies in the Kelly-tumour mouse model ( Supplementary Fig. 7), where ex vivo quantification showed that 111 In-4hD29 specifically accumulated in the Kelly tumour while unbound tracer was quickly removed from non-targeted tissues through renal clearance. This resulted in tumour-to-blood and tumour-to-muscle ratios as high as 16.7 and 29.1, respectively, 2h after i.v. administration.

Discussion
We presently identified DPP6 as a new biomarker of the endocrine pancreas, expressed preferentially in human beta and alpha cells and also in insulinomas. To validate its use for in vivo imaging, we generated a high-affinity camelid single-domain antibody (nanobody) targeting human DPP6. Nanobodies are suitable for clinical in vivo imaging with nuclear modalities, both in animal models [15][16][17] and clinically 21 . The lead nanobody, 4hD29, was radiolabelled and in vivo SPECT/CT imaging and biodistribution studies were performed in immunodeficient mice transplanted with DPP6-expressing Kelly neuroblastoma cells or insulin-producing human EndoC-βH1 cells. To our knowledge, this is the first description of the use of nanobodies for in vivo imaging of human insulin secreting cells. Human DPP6-expressing cells were clearly visualized in both models, with very low background.
The available knowledge of beta and alpha cell mass is based on autopsy data and it is presently not possible to follow the survival of transplanted human pancreatic islets or endogenous ECM changes over time in diabetic patients 3 . Development of both beta cell specific 5 and ECM tracers 4 are of importance considering the potential plasticity of mature pancreatic cells 8,[28][29][30][31] and the possibility that alpha-cells may provide a reservoir for beta-cells 8 . Furthermore, beta and alpha cells are the most frequent cell types in human pancreatic islets, and novel ECM imaging modalities would be very useful for the follow up of transplanted pancreatic islets 32 . In this context, it is important to identify targets that are not affected by inflammatory or metabolic stress, which is the case of DPP6, allowing the estimation of beta cell mass unaffected by the stressful conditions to which islet cells are exposed in diabetes or following transplantation.
Previous strategies to screen for beta cell or islet cell biomarkers have often relied on antibody arrays 13,33-35 . Our present approach to search for novel biomarkers based on RNA sequencing and analysis of the islet transcriptome, as compared to other tissues, provides a more sensitive and specific approach, avoiding the unspecific binding issue associated with antibody-based screenings 36 .
The DPP6 gene encodes for a single pass transmembrane peptidase without peptidase activity. At the protein level, DPP6 co-ensembles with Kv4-potassium channels that take part in exocytosis. DPP6 functions as a regulator of the channels' recovery rate and voltage dependency 37 and promotes the cell surface localization of KCND2 38 , besides other functions that remain to be identified 39 . The different splicing variants of DPP6 have different lengths in their intracellular part, with a relatively constant extracellular part 38 . The splicing variant DPP6-001 is the DPP6 variant with the highest expression in pancreatic islets (present data). All the main DPP6 variants are expressed in the central nervous system, except DPP6-001, which is preferentially expressed in pancreas 37 . Considering the expression pattern of DPP6 and the unchanged nature of the extracellular part of the full gene, we have chosen the protein encoded by the full gene as the best target, instead of only the DPP6-001 variant.
Once a relevant islet biomarker is identified, the next step is to develop ligands and tracers for in vivo imaging. For this purpose, we have presently generated anti-DPP6 nanobodies. The camelid nanobody tracer backbone was chosen for its inert properties, fast clearance rate and short blood retention time 17 . Antibody based approaches are less optimal in this context, as they have high non-specific retention due to their slow clearance rate 40 . On the other hand, nanobodies, due to their small size (15kDa), are cleared as small hydrophilic proteins and are excreted via the renal route 41 .
After dromedary immunization, biopanning of a phage displayed nanobody library and detailed in vitro characterization, the lead anti-DPP6 nanobody 4hD29 was selected for in vivo evaluation. 4hD29 is specific for the extracellular part of human DPP6, without binding to endogenous murine DPP6. Due to the species-specific nature of the nanobody, we performed all in vivo validation steps in humanized mouse models, i.e. mice implanted either with insulin-producing EndoC-βH1 cells in the quadriceps or femoris muscle, or xenografted in the subcutaneous tissue with the neuronal Kelly cancer line that spontaneously expresses DPP6. It will be important to confirm these findings in future experiments using implanted human islets. The tracers displayed optimal in vivo specificity, with high EndoC-βH1 transplant-to-muscle and high Kelly tumour-to-blood ratios, and excellent in vivo visualization of these DPP6-expressing cells implanted either in the muscle or in the subcutaneous tissue (Fig. 7, Supplementary Videos 1 and 2). Future experiments must test whether 4hD29 allows visualization of human islets implanted in rodents in the intraportal system, spleen, kidney capsule, peritoneum, muscle or in the subcutaneous region. Intramuscular grafting of pancreatic islets is under development 42 , while 4hD29-mediated visualization of islets implanted under the kidney capsule, and to a certain extent also of liver islet transplants, is deemed less optimal due to the potential nonspecific accumulation of nanobody-tracers at this site(s).
In conclusion, we have presently identified DPP6 as a new potential islet biomarker and generated a novel nanobody tracer that targets it. The tracer displays high specificity for DPP6 and its in vivo properties were validated in mice xenografted with either human insulin secreting cells or a neuronal derived tumour that spontaneously expresses DPP6.

Ethical statements. Pancreases not suitable for clinical purposes were obtained via the Endocrinology and
Metabolism of Organ and Cellular Transplantation unit at Cisanello University Hospital at the University of Pisa, Italy and from the department of Pathology, UZ-Brussel, Vrije Universiteit Brussel (VUB), Brussels, Belgium with informed written consent and processed with the approval of the local Medical and Health Research Ethics Committees of the Pisa University, Italy and of the VUB, Belgium. The donors were anonymised, and all experiments and methods using human pancreatic islets were approved by and performed in accordance with the guidelines and regulations made by the regional Medical and Health Research Ethics committees of the Pisa University, Italy and of the VUB, Belgium. Human islet isolation and culture was performed as previously described 43 . No organs/tissues were procured from prisoners. The human pancreatic islet or pancreas preparations used in the present study are described in Supplementary Table 1. Historical material from two insulinomas was provided by the Pathology Department of the VUB. Information on the patients and tumours are provided in the legend for Supplementary Figure 2. Male NMRI-Foxn1 nu /Foxn1 nu , female SCID CB-17/Icr-Prkdc scid /Rj mice (Janvier labs), all 6 weeks old, and C57Bl6 mice (both from Charles River Laboratories, Saint-Germain-sur-l′Arbresle, France) were used and housed according to the rules of the Belgian regulations for animal care and in accordance with the Animal Act 1986; 2013, and all experiments and methods involving these animals were performed with approval by the Ethical Committee for Animal Welfare (CEBEA) of the Medical Faculty of the ULB and ULB-Centre for Microscopy and Molecular Imaging (CMMI) (Campus Biopole Charleroi) (ethical permits CMMI-2013-03;2014 and 485N), Belgium.

Mouse models hosting human cells.
Male NMRI-Foxn1 nu /Foxn1 nu and female SCID CB-17/ Icr-Prkdc scid /Rj mice were transplanted with human insulin-producing EndoC-βH1 cells as described 44,47 . Male NMRI-Foxn1 nu /Foxn1 nu mice were used for methodological development/validation, but all follow up experiments were performed in female SCID CB-17 mice. At the day of inoculation, 4-6 × 10 6 EndoC-βH1 cells were seeded on a rubber toric joint (EFJM), supported in Matrigel HC (Corning) supplemented with MmVEGF-164 (1 ng/ml) (BioLegend). The cell-containing or the empty vehicle rubber rings were then inserted under the epimysium in the biceps or quadriceps femoris muscle. The mice were anesthetized with 3 % isoflurane. They received short-term analgesic (Buprenorphine 0.1 mg/kg) and long-term analgesic (3 mg/ml of Acetaminophen-supplemented water for 10 consecutive days) respectively before and after the surgery. Random glycaemia was measured weekly with an ACCU-CHEK Nano glucometer (ROCHE). Once the tumour became palpable, the mice received 20 % glucose-supplemented drinking water to counter the progressive hypoglycemia induced by the EndoC-βH1 cells. Human C-peptide was measured in plasma with a human Ultrasensitive C-peptide ELISA (Mercodia). Kelly tumour-bearing mice were generated by s.c. injections of 5 × 10 6 cells mixed 1:1 with Matrigel (Corning) in the hind leg of female Crl:NU(NCr)-Foxn1 nu mice.
Generation, production and selection of Nanobodies. Nanobodies were generated as detailed elsewhere 20,40 . A recombinant protein produced in NS0 cells encoding the extracellular domain amino acids 118-865 of human DPP6 (HsDPP6, R&D Systems, Abingdon, UK) was used to immunize a dromedary, for enrichment by biopanning of a phage-displayed immune nanobody library and for screening binders by ELISA. Specific nanobody binders were sequenced and unique sequences were identified using Geneious software (Biomatters). Hexahistidine-tagged nanobodies, cloned in the pMES4 plasmid, were produced and purified from E. coli cells as previously described 40 . As a non-targeting negative control nanobody BcII10, directed against a bacterial enzyme 48 , was used. Affinity measurements were performed on a Biacore T200 instrument 40 . Briefly, recombinant