Disruption of the pancreatic vasculature in zebrafish affects islet architecture and function

Specialized to sense and rapidly respond to circulating metabolites, pancreatic islets are crucial in regulating glucose homeostasis. Attesting to their importance, these endocrine cell clusters are densely vascularized (Bonner-Weir, 1988), drawing up to 20 times more blood compared with the surrounding exocrine tissue (Lifson et al., 1985). Thus, understandably, when their vascular connections are severed during isolation for transplantation purposes, the islets experience ischemia, resulting in poor engraftment and survival (Pepper et al., 2013). Transplanted islets seldom achieve the high degree of vascularization observed in native pancreatic islets (Mattsson et al., 2002). Thus, elucidating the interactions between pancreatic endocrine cells and blood vessels, and exploring the consequences of their modulation, could facilitate improved designs for transplantation protocols.

During mammalian development, pancreatic progenitor cell differentiation and self-renewal, as well as expression of endocrine markers, depend on signals from the endothelium (Lammert et al., 2001; Yoshitomi and Zaret, 2004; Edsbagge et al., 2005; Magenheim et al., 2011). The vasculature also regulates β-cell function and proliferation through the establishment of a complex basement membrane (Nikolova et al., 2006; Parnaud et al., 2009). More recently, insulin secretion in β-cells was shown to be targeted towards the vasculature, owing to a basolateral vascular face in β-cells (Gan et al., 2017, 2018). Recent attempts at uncovering the role of the vasculature in mature islets using temporal inactivation of VEGFA (Reinert et al., 2013) or transgenic overexpression of a soluble isoform of VEGF receptor 1 (sFLT1) (D'Hoker et al., 2013) report that a marked decrease of pancreatic vasculature resulted in no major defects. These models achieved a reduction in islet vascular density down to 30% of wild-type levels, leading to mild hypoxia. However, it is unclear whether compensatory effects of β-cell function (Chen et al., 2017) from the spared, i.e. vascularized, islets might explain the observations made in these studies. Hence, the consequences of complete vascular depletion in an islet remain to be determined.

Owing to its many advantages (Ober et al., 2003; Beis and Stainier, 2006), the zebrafish is an ideal model to study pancreas development and function, including hepatopancreatic duct development (Manfroid et al., 2012; Villasenor and Stainier, 2017), β-cell development (Ninov et al., 2012; Matsuda et al., 2013), α-to-β-cell transdifferentiation (Ye et al., 2015; Lu et al., 2016), β-cell regeneration (Andersson et al., 2012; Delaspre et al., 2015), islet innervation (Yang et al., 2018) and vascularization (Hen et al., 2015; Toselli et al., 2019), as well as endocrine cell motility (Freudenblum et al., 2018). The developing zebrafish has a single functional primary islet (Biemar et al., 2001; Field et al., 2003), thereby facilitating the observation of all the organism's β-cells and their interactions with blood vessels. In this study, combining live imaging with two potent tools to disrupt the pancreatic vasculature – a dominant negative form of Vegfaa (dnVegfaa) and sFlt1 – we identified distinct roles for the pancreatic vasculature in regulating islet architecture and function.

In the developing zebrafish pancreas, pdx1-positive progenitors appear at around the 10-somite stage [13.25 h post fertilization (hpf)] and preproinsulin (ins)-positive cells around the 12-somite stage (14 hpf), followed by their migration to the midline to form the endocrine islet (24 hpf) (Biemar et al., 2001; Field et al., 2003). At this stage, the embryonic islet is sandwiched between the dorsal aorta (DA) and the posterior cardinal vein (PCV). Subsequently, between 36 and 72 hpf, endothelial cells from the subintestinal vein (SIV) and the anterior PCV (aPCV) invade the islet to form the pancreatic vasculature (Hen et al., 2015). To interrupt vascularization of the islet, we expressed dnVegfaa under the control of the ins promoter. dnVegfaa, with its inability to bind VEGF receptors, although retaining its property to dimerize with multiple Vegf family members, possesses potent anti-angiogenic activity (Rossi et al., 2016).

We did not observe any defects in early embryonic islet formation or vascular development in 28 hpf embryos upon dnVegfaa expression (Fig. 1A,B). However, by 48 hpf, the endocrine islet was completely dispersed (Fig. 1C,D), and it remained disrupted at later stages, despite the fact that β-cells maintained contact with other endocrine cells (Fig. S1A,B), including α-cells (Fig. S1C,D). No significant change in β-cell number was observed (Fig. S1E). In vivo time-lapse imaging revealed that, although the β-cells in Tg(ins:Tomato) embryos migrated together towards the aPCV, the β-cells in Tg(ins:Tomato-P2A-dnvegfaa) embryos began separating between 38 and 42 hpf (Fig. 1E,F, Movies 1 and 2). We also observed defects in some intersegmental vessels (ISVs) at 34 hpf (Fig. 1F). Interestingly, the dnVegfaa-expressing β-cells were still closely associated with endothelial cells, and only the ISVs situated directly above the dnVegfaa-expressing β-cells exhibited defects (Fig. 1F-F′‴). A possible explanation for this islet dispersion is that interference with Vegf signaling affects the expression of endothelial cell-derived factors (Johansson et al., 2006) and extracellular matrix deposition in the islet (Nikolova et al., 2007), resulting in altered migratory signals to the β-cells (Peiris et al., 2014; Hogan and Hull, 2017). Together, these observations indicate a rather short-ranged anti-angiogenic effect of dnVegfaa, resulting in the local disruption of the vasculature, itself leading to islet dispersion.

To gain mechanistic insight into how dnVegfaa-mediated vascular disruption causes endocrine cell dispersal, we wanted to determine the likelihood that an isolated β-cell expressing dnVegfaa would disperse away from the islet and whether Vegf signaling pathway mutants also exhibited these phenotypes. First, the Tg(ins:Tomato-2A-dnVegfaa) construct was injected into Tg(ins:GFP) embryos to overexpress dnVegfaa in a subset of β-cells. The islet remained intact in these mosaic animals when dnVegfaa was expressed in 1-20% of the β-cells (Fig. 2A,B). When 20-50% of the β-cells expressed dnVegfaa, a few β-cells, not necessarily the dnVegfaa-expressing ones, were observed away from the islet cluster (Fig. 2A). Severe dispersion of β-cells was only observed when more than half of the β-cells expressed dnVegfaa (Fig. 2B). Together, these observations indicate an indirect and dose-dependent role for dnVegfaa in the dispersal of β-cells.

Disruption of islet architecture occurred even when dnVegfaa expression was induced postembryonically (Fig. S1F-H), suggesting that dnVegfaa was not merely interfering with early endocrine cell migration, but rather affected an active mechanism regulating islet architecture. To gain further insights into this mechanism, we examined the expression of several reporter lines in the developing islet at 48 hpf. We detected vegfaa promoter activity in β-cells, as well as in nearby non-endocrine cells, and vegfab promoter activity in endocrine cells (Fig. S2A-C). Promoter activity of receptor genes implicated in Vegf and Pdgf signaling, such as kdrl, pdgfrb and flt1, was also observed in the vessels within the islet (Fig. S2D,E). Next, we analyzed zebrafish mutants for Vegf ligand and receptor genes including vegfaa, vegfab, flt1, kdr and kdrl, and compared their β-cell architecture and surrounding vasculature with those in Tg(ins:dnvegfaa) embryos. In 48 hpf wild-type embryos, the β-cells clustered tightly together, with blood vessels running along the periphery of the islet and three to four intra-islet vessels (Fig. 2C, Movie 3). In Tg(ins:dnvegfaa) embryos, the β-cells were dispersed, and the surrounding and intra-islet vasculature was also disrupted. Notably, the β-cells were still in contact with blood vessels (Fig. 2D and Fig. S3A,B). However, unlike wild-type β-cells, which have multiple contact points with the vasculature, dnVegfaa-expressing β-cells contacted the vasculature at only one point (Fig. 2C,D). In all the mutants examined, we did not detect major defects in β-cell clustering or peripheral/intra-islet vascularization (Fig. 2E-I). Notably, the Tg(ins:sflt1) line, which was generated as an additional dominant negative approach to interfere with the vasculature near the β-cells, contained islets with an intact β-cell architecture but no intra-islet vasculature (Fig. 2J, Movie 4). These observations, combined with the effects of dnVegfaa on the β-cell vasculature, suggest that during the establishment and maintenance of islet vascularization there is redundancy amongst the Vegf receptor genes in zebrafish (Rossi et al., 2016; Toselli et al., 2019); in mouse however, Kdr mutants exhibit severe vascular defects (Shalaby et al., 1995). In addition, because the expression levels of the Vegf receptor genes in β-cells are minimal (Table S1) (Tarifeño-Saldivia et al., 2017; Salem et al., 2019), we postulate that dnVegfaa and sFlt1 do not affect β-cells directly.

The mechanistic differences between dnVegfaa and sFlt1 action are not yet completely understood. sFlt1 can attenuate Vegfa- and Vegfb-mediated signaling without affecting Vegfc- or Vegfd-dependent signaling (Kendall and Thomas, 1993; Koch and Claesson-Welsh, 2012). In contrast, dnVegfaa can potentially dimerize with all Vegf ligands, sterically hinder Vegfr2 activation by other ligands and thereby block all Vegfr2-mediated signaling, including Vegfc- and Vegfd-dependent processes (Rossi et al., 2016). In addition, steric inhibition of Vegfr2 by dnVegfaa might also block the angiogenic action of other ligands such as gremlin (Mitola et al., 2010). Thus, it is possible that residual Vegf receptor signaling in Tg(ins:sFlt1) animals is responsible for the differences in the vascular phenotypes between the models. Further experiments expressing dnVegfaa in other vascular beds and studying endothelial cell behavior at single-cell resolution are needed to improve our mechanistic understanding of these reagents.

Functional β-cells are crucial in regulating glucose levels at early embryonic stages in zebrafish (Jurczyk et al., 2011). Activity of the key gluconeogenic gene, pck1, results in increasing glucose levels in larvae, which peaks at ∼120 hpf (Gut et al., 2013). To determine the effects of dispersed or non-vascularized islets on β-cells, we analyzed the pancreatic vasculature and β-cell function in 120 hpf animals. The pancreas in wild-type animals is densely vascularized with an SIV-derived vessel running along the pancreatic duct, as well as multiple blood vessels within the islet, as previously reported (Hen et al., 2015) (Fig. 3A-A″). Using live imaging in Tg(gata1:DsRed); Tg(fli:GFP) animals, active blood circulation was observed within these islets at 120 hpf (Movie 5). Similar to observations at 48 hpf, dispersion of the β-cells was still observed in 120 hpf Tg(ins:Tomato-2A-dnvegfaa) animals. These β-cells remained in close proximity to blood vessels (Fig. 3B-B″ and Fig. S3A,B), and blood circulation was also observed in these pancreata (Movie 6). In contrast, the pancreas in 120 hpf Tg(ins:sflt1) animals was entirely devoid of blood vessels (Fig. 3C-C″), resulting in a lack of β-cell-vascular interactions. This observation was also supported by the absence of any detectable blood circulation in the islets of Tg(ins:sflt1) animals (Movie 7).

Although glucose-stimulated insulin secretion assays have been performed in vitro to assess β-cell function in models of vascular disruption (D'Hoker et al., 2013; Reinert et al., 2013), these experiments could be limited by confounding hypoxic effects in the isolated islets, even from control animals. In contrast, our models offer the advantage of being able to visualize responses to glucose in vivo. To this aim, we used calcium imaging in β-cells to determine their responsiveness to glucose. β-Cells in 120 hpf Tg(ins:GCaMP5) animals were imaged with or without exogenous glucose and scored as being responsive, depending on visual changes in fluorescence intensity (Fig. S4A-G). In Tg(ins:DsRed) animals, 47±12% (mean±s.e.m) of the β-cells responded to endogenous glucose and 62±6% to exogenous glucose (Fig. 3D). In Tg(ins:dnvegfaa) animals, 40±17% of the β-cells responded to endogenous glucose and 44±18% of them to exogenous glucose. In contrast, in Tg(ins:sflt1) animals, only 2 (+4 or −2)% of the β-cells responded to endogenous glucose, likely owing to the lack of vasculature and blood flow within the islet. Addition of exogenous glucose, which leads to a higher concentration throughout the animal (including by its non-vascularized β-cells), resulted in a much higher fraction of glucose-responsive β-cells (29±19%), suggesting that they retain their ability to respond to glucose. Furthermore, free glucose levels, which serve as a measure of functional β-cell mass, were threefold higher in Tg(ins:sflt1) animals, but only 20% higher in Tg(ins:dnvegfaa) animals (Fig. 3E). pdx1 mRNA levels in 120 hpf Tg(ins:sflt1) animals were similar to those in Tg(ins:dsRed) animals (Fig. S4H), suggesting that differences in β-cell differentiation are unlikely, whereas higher ins mRNA expression levels (Fig. S4I) could be a compensatory response to elevated glucose levels. Together, these observations indicate that although β-cells in Tg(ins:sflt1) animals are capable of responding to glucose, the lack of pancreatic vasculature severely limits their function.

Using a combination of imaging and blood glucose readouts, we next aimed to examine whether the effects of β-cell-specific expression of dnVegfaa or sFlt1 persisted into adulthood, and how they affected glucose homeostasis. Wild-type and Tg(ins:DsRed) zebrafish contain a primary islet in the anterior head of their pancreas, along with several smaller secondary islets located along their pancreatic ducts (Fig. 4A) (Chen et al., 2007). The primary islet is densely vascularized, and large numbers of β-cells are clustered at the core, with almost every β-cell in direct contact with blood vessels (Fig. 4B and Fig. S5A).

β-Cells in Tg(ins:dnVegfaa) animals were dispersed throughout the pancreas with no clear indication of a primary islet (Fig. 4C), although a greater density of β-cells was observed in the head of the pancreas (Fig. 4C). A small fraction of the dispersed β-cells remained in contact with endothelial cells (Fig. 4D). In 72 hpf animals, we observed a few β-cells in the blood circulation (Movie 8). To assess whether these circulating β-cells were functional, and to quantify pancreatic ins expression, we isolated RNA from several sections of adult fish and measured ins mRNA levels (Fig. S5E). Compared with wild-type samples, lower levels of ins mRNA were detected in sections containing the anterior pancreas (containing the primary islet) and the posterior pancreas (containing secondary islets) of Tg(ins:dnvegfaa) animals (Fig. S5F), supporting our imaging observations of a decrease in total β-cell volume (Fig. S5G) and volume of β-cells in secondary islets (Fig. S5H). No significant increase in ins mRNA levels was detected in the head or tail sections of Tg(ins:dnvegfaa) animals compared with wild type (Fig. S5F), making it unlikely that the circulating β-cells were functional. The lack of large secondary islets (Fig. S5H) indicates that the β-cells are unable to form clusters in these animals, in line with observations at larval stages (Fig. 3B).

Although Tg(ins:sflt1) animals contained a prominent primary islet and several secondary islets along the pancreatic ducts, the morphology of these islets appeared to be drastically altered (Fig. 4E). Strikingly, the β-cells in the primary islet appeared as a spherical shell, giving rise to a ‘hollow’ core (Fig. 4F), and amounted to a lower overall β-cell volume (Fig. S5G). No blood vessels were detected in the core of the islet, or in close proximity to the β-cells, although DAPI-positive cells were observed in the core (Fig. S5D). β-Cell clusters in secondary islets in these animals also exhibited an elongated sheet-like morphology, enveloping the pancreatic ducts, with β-cell volume in most islets comparable with that seen in wild types (Fig. S5C,H). This arrangement was in stark contrast to the spherical secondary islets observed in Tg(ins:Tomato) animals (Fig. S5A). The lack of vasculature and the single-layered arrangement indicate an adaptation of the β-cells to a possibly hypoxic environment within the islet owing to sFlt1 expression. Our model likely achieves a higher degree of vascular depletion compared with the mouse sFLT1 expression model (D'Hoker et al., 2013), possibly explaining the morphological differences observed.

As a measure of functional β-cell mass, we determined fasting blood glucose levels and found that, compared with Tg(ins:Tomato) controls, Tg(ins:sflt1) animals did not exhibit any major differences, whereas Tg(ins:dnvegfaa) animals displayed mildly lower levels (Fig. 4G). However, a glucose tolerance test in these animals revealed an impaired ability to clear glucose in both Tg(ins:dnvegfaa) and Tg(ins:sflt1) animals compared with Tg(ins:Tomato) controls (Fig. 4H). The dispersed islet architecture in Tg(ins:dnvegfaa) animals likely affects the interactions between β-cells, and could thereby alter their coupling, which is crucial to mount a synchronized response to glucose (Ravier et al., 2005). Likewise, the lack of islet vasculature in Tg(ins:sflt1) animals could be responsible for the reduced ability of their β-cells to sense circulating glucose, as well as for a reduced ability for secreted insulin to reach the blood circulation (D'Hoker et al., 2013; Reinert et al., 2013).

In conclusion, our study combines live imaging with potent tools to disrupt the vasculature; it reveals two distinct consequences of abrogating interactions between pancreatic islets and the vasculature. Despite variation in cell layout, pancreatic endocrine cells across species cluster together as micro-organs (Steiner et al., 2010), facilitating cell-cell interactions. It will be important to investigate the molecular mechanisms that lead to endocrine cell clustering and the role of the resulting interactions on their function. In addition, it will be interesting to further investigate the morphological adaptations observed in islets with depleted vasculature. Altogether, these findings should help design improved stem cell-derived cell transplantation strategies with reduced ischemia, immunogenicity and inflammation, thereby ensuring cell and islet engraftment and long-term function.

Transgenic lines used in the study were: Tg(ins:GFP)^zf5 (Huang et al., 2001), Tg(gata1:DsRed)^sd2 (Traver et al., 2003), Tg(fli1a:GFP)^y1 [referred to as Tg(fli:GFP); Lawson and Weinstein, 2002], Tg(ins:DsRed)^m1018 (Anderson et al., 2009), TgBAC(neurod:EGFP)^nl1 (Obholzer et al., 2008), TgBAC(vegfaa:EGFP)^pd260 (Karra et al., 2018), Tg(pdgfrb:GFP)^ncv22Tg (Ando et al., 2016), Tg(kdrl:tagBFP)^mu293 (Matsuoka et al., 2016), Tg(-0.8flt1:RFP)^hu5333 (Bussmann et al., 2010) and Tg(UAS-E1b:NfsB-mCherry)^c264 (Davison et al., 2007).

Mutant lines used in the study were: vegfaa^bns1 (Rossi et al., 2016), vegfab^bns92 (Rossi et al., 2016), kdrl^hu5088 (Hogan et al., 2009), flt1^bns29 (Matsuoka et al., 2016) and kdr^bns32.

Zebrafish husbandry was performed under standard conditions in accordance with institutional (Max Planck Gesellschaft) and national ethical and animal welfare guidelines approved by the ethics committee for animal experiments at the Regierungspräsidium Darmstadt, Germany.

We used 1.1 kb of the zebrafish ins promoter to drive β-cell-specific expression of proteins in the following transgenic lines: Tg(ins:TagRFP-T-P2A-sflt1)^bns286 [referred to as Tg(ins:sflt1)], Tg(ins:TagRFP-T-P2A-dnvegfaa)^bns284 containing the T46A mutant isoform of Vegfaa-121 [Tg(ins:dnvegfaa); Rossi et al., 2016], Tg(ins:memTdTomato-P2A-dnvegfaa)^bns282 [Tg(ins:Tomato-P2A-dnvegfaa)], Tg(ins:memTdTomato-P2A)^bns211 [Tg(ins:Tomato)], Tg(ins:LoxP-Gfp-LoxP-memTom-P2A-dnvegfaa)^bns287 and Tg(ins:GCaMP5)^bns306. All transgenic lines were generated using I-SceI meganuclease-mediated transgenesis (Thermes et al., 2002). Multiple founders were identified and analyzed to ensure consistent expression of reporters and observed phenotypes.

To generate the TgBAC(pdx1:CreERT)^bns307 line, a BAC containing 200 kb of the pdx1 locus was used and recombineering steps were performed as described previously (Helker et al., 2019) with the inclusion of a CreERT2-Kanamycin cassette. The TgBAC(vegfab:gal4ff)^bns273 line was generated by following the standard BAC recombineering protocol described previously (Bussmann and Schulte-Merker, 2011). Briefly, the BAC clone DKEY-234H13 (Source Bioscience) was engineered to replace the vegfab start codon with gal4ff. The engineered construct was injected into one-cell stage embryos together with Tol2 transposase mRNA.

To generate mutants for kdr, the CRISPR design platform (http://crispr.mit.edu) was used to design sgRNAs binding to exon 6, and constructs were assembled and mutagenesis was performed as described previously (Jao et al., 2013; Varshney et al., 2015). An 8 bp deletion was obtained.

To image blood flow in 120 hpf zebrafish and circulating cells in 72 hpf zebrafish, the animals were mounted laterally in 0.8% low-melting agarose with 0.02% Tricaine. A Zeiss Spinning Disk confocal microscope was used to acquire 5 s of images at 40 frames per second from multiple planes in the islet.

For calcium imaging, Tg(ins:GcaMP5) larvae were treated from 72 to 120 hpf with 1.5 mM of Pck1 inhibitor 3-mercaptopicolinic acid (3-MPA) in Hepes-buffered egg water (Jurczyk et al., 2011). At 120 hpf, the larvae were mounted in 0.8% low-melting agarose with 0.02% Tricaine. An LSM880 (Zeiss) confocal microscope was used to image multiple planes in the islet for 10 min without addition of glucose, and for 15 min post glucose addition. Images were processed on Imaris (Bitplane) and maximum projection images were generated.

Immunostaining was performed as described previously (Mullapudi et al., 2018). Antibody dilutions used are as follows: guinea pig anti-insulin polyclonal (1:100, A0564, Thermo Fisher Scientific), mouse anti-Glucagon (1:300, G2654, Sigma-Aldrich), rabbit anti-tagRFP (1:200, AB233, Evrogen), chicken anti-GFP (1:300, GFP-1010, Aves), goat anti-guinea pig AlexaFluor568 (1:300, A-11075, Thermo Fisher Scientific), goat anti-chicken AlexaFluor488 (1:500, A-11039, Thermo Fisher Scientific). For CLARITY-based tissue clearing, digestive tracts of adult zebrafish were fixed and cleared according to the recommended protocol (Chung and Deisseroth, 2013) using the X-CLARITY Tissue Clearing System (Biozym GmbH). β-Cell volume was measured using the ‘Surface’ tool in Imaris (Bitplane). For total β-cell volume quantification, the sum of the volumes encompassed by all surfaces was computed. For quantification of secondary islet volume, all islets except the primary islet were included for analysis. The volumes of the individual islets were computed and plotted.

Glucose measurements in larvae and adults were performed as described previously (Eames et al., 2010; Jurczyk et al., 2011; Matsuda et al., 2018). The glucose tolerance test was performed by injecting a 0.1 mg/µl glucose (Sigma-Aldrich) solution intraperitoneally at 1 mg/g fish weight (Eames et al., 2010). The injected animals were incubated at 28.5°C for 90 min, and then euthanized to collect blood and perform glucose measurements using a FreeStyle Freedom Lite Glucose Meter (Abbott).

Total RNA from sections of adult zebrafish was extracted using TRIZOL (Ambion) and then cleaned with RNA Clean & Concentrator-5 (Zymo Research). Then 500 ng RNA was used to synthesize cDNA using the Maxima First-Strand cDNA kit (Thermo Fisher Scientific). SYBR Green (Thermo Fisher Scientific) was used to quantify ins expression levels on a CFX connect Real-time System (Bio Rad) and normalized to rpl13 expression across samples. Primers used for ins were 5′-TTTAAATGCAAAGTCAGCCACCTCAG-3′ and 5′-TTTAAATGCAAAGTCAGCCACCTCAG-3′; and for pdx1 primers were 5′-GGACCAGCCAAATCTTACCG-3′ and 5′-CCTCGGCCTCGACCATATA-3′.

We thank Radhan Ramadass for help with microscopy; Rashmi Priya, Sébastien Gauvrit, Rubén Marin Juez, Anabela Bensimon Brito and Zacharias Kontarakis for comments and suggestions; and Martin Laszczyk and the rest of the animal caretaking team for zebrafish care.