|Home | About | Journals | Submit | Contact Us | Français|
Development of the pancreas and cerebellum require Pancreas-specific transcription factor 1a (Ptf1a), which encodes a subunit of the transcription factor complex PTF1. Ptf1a is required in succession for specification of the pancreas, proper allocation of pancreatic progenitors to endocrine and exocrine fates, and the production of digestive enzymes from the exocrine acini. In several neuronal structures, including the cerebellum, hindbrain, retina and spinal cord, Ptf1a is transiently expressed and promotes inhibitory neuron fates at the expense of excitatory fates. Transcription of Ptf1a in mouse is maintained in part by PTF1 acting on an upstream autoregulatory enhancer. However, the transcription factors and enhancers that initially activate Ptf1a expression in the pancreas and in certain structures of the nervous system have not yet been identified. Here we describe a zebrafish autoregulatory element, conserved among teleosts, with activity similar to that described in mouse. In addition, we performed a comprehensive survey of all non-coding sequences in a 67 kilobase interval encompassing zebrafish ptf1a, and identified several neuronal enhancers, and an enhancer active in the ventral pancreas prior to activation of the autoregulatory enhancer. To test the requirement for autoregulatory control during pancreatic development, we restored ptf1a function through BAC transgenesis in ptf1a morphants, either with an intact BAC or one lacking the autoregulatory enhancer. We find that ptf1a autoregulation is required for development of the exocrine pancreas and full rescue of the ptf1a morphant phenotype. Similarly, we demonstrate that a ptf1a locus lacking the early enhancer region is also capable of rescue, but only supports formation of a hypoplastic exocrine pancreas. Through our dissection of the complex regulatory control of ptf1a, we identified separate cis–regulatory elements that underlie different aspects of its expression and function, and further demonstrated the requirement of maintained ptf1a expression for normal pancreatic morphogenesis. We also identified a novel enhancer that mediates initiation of ptf1a expression in the pancreas, through which the signals that specify the ventral pancreas are expected to exert their action.
The vertebrate pancreas forms from evaginations of the endoderm, the dorsal and ventral pancreatic buds, that later fuse to produce a single organ composed of hormone–secreting endocrine cells, digestive enzyme–secreting exocrine cells, and duct cells. In the zebrafish, the dorsal bud forms by 24 hours post fertilization (hpf)  and exclusively produces endocrine cells of the principal islet. The ventral bud arises at 34hpf and grows towards the dorsal bud–derived islet, with which it completely merges by 52hpf . The pancreas specific transcription factor la (Ptf1a) has a critical role in pancreatic development. Loss of Ptf1a in mice results in loss of the ventral pancreatic bud and early arrest in development of the dorsal bud [3, 4]. In zebrafish, depletion of ptf1a through antisense morpholinos results in the absence of a ventral bud, but the dorsal bud is unaffected [5, 6]. Despite the critical requirement for Ptf1a in early pancreatic development, the signals that initiate its expression in the ventral pancreas are poorly characterized. vHNF1 is an attractive candidate for the direct activation of Ptf1a, since loss of vHNF1 in mouse results in loss of endodermal Ptf1a expression  and similarly in zebrafish results in the loss of pancreatic marker expression . Identification of the Ptf1a regulatory elements that mediate the activation of its pancreatic expression is therefore an important step to determine the involvement of specific transcription factors in ventral pancreas specification.
In addition to its function in pancreas specification, Ptf1a has a dosage sensitive role in the allocation of pancreatic progenitors to the exocrine and endocrine compartments in both mouse [9, 10] and zebrafish . A cross-repressive loop between the pro–exocrine Ptf1a and the pro–endocrine Nkx6 operates to segregate endocrine and exocrine cells in the pancreas . Both Ptf1a and Nkx6.1 are known to activate their own expression [12, 13] and the autoregulatory function of both genes in the context of competing cell fate determinants has been proposed to either confer stability to the bipotential progenitor state  or to maintain stability of the commitment to a given lineage after the resolution of the bipotential fate choice.
Ptf1a is also important in the development of several populations of inhibitory neurons. Specifically, in the cerebellar primordium Ptf1a is expressed in the ventricular zone, a germinative zone that generates progenitors of the inhibitory Purkinje cells and interneurons . Loss of Ptf1a leads to a loss of Purkinje cells and interneurons during embryonic development, which secondarily results in postnatal cerebellar agenesis in mice  and humans . Within more caudal rhombomeres of the mouse hindbrain, Ptf1a expressing domains contribute inhibitory neurons to the cochlear nucleus  and climbing fiber neurons of the precerebellar system . Similarly, in zebrafish the ptf1a expressing domain contributes to the precerebellar system . Within the dorsal spinal cord, Ptf1a is required for the generation of two populations of inhibitory interneurons . Also, in the retina Ptf1a is required in both mouse  and zebrafish  for the generation of inhibitory interneurons that modulate the signals generated by the photoreceptor cells.
Despite the critical requirement for Ptf1a in the development of pancreas and cerebellum, the regulatory elements that initiate expression in these organs have remained elusive. A region of ~30 kb encompassing mouse Ptf1a was examined for cis–regulatory activity . This study identified sequences downstream of the gene that direct transcription to the hindbrain, spinal cord and retina, but are not active in cerebellum. The distal portion of the upstream sequence (−15.6 to −13.4kb) harbors two conserved predicted PTF1 binding sites and directs transcription to the embryonic pancreas. The transcriptional activity in the pancreas is absent in Ptf1a mutants, and PTF1 complex binds directly to the sequence in vitro, consistent with autoregulatory function. The upstream enhancer is also active in the dorsal spinal cord, retina, hindbrain, and cerebellum , where its activity is similarly autoregulatory. In the pancreas, the autoregulatory control has been hypothesized to effect an irreversible commitment to the acinar cell program, although its role has not been directly tested.
We have carried out a comprehensive survey of the genomic interval spanning zebrafish ptf1a for cis–regulatory activity, focusing specifically on elements regulating transcription in the developing pancreas. We identified an autoregulatory enhancer analogous to that in the mouse, and showed that the predicted PTF1 binding sites are both necessary and sufficient for its activity. We also identified a sequence capable of initiating early expression in the pancreas, in a region with high homology between teleosts and mammals. We describe additional sequences active in regions of the nervous system that normally express ptf1a, including the cerebellum. Finally, we tested specifically for the requirement for ptf1a autoregulation in pancreas development, and find that it is necessary for early expansion of the exocrine pancreas.
All fish were maintained under standard conditions, in accordance with regulations of the Institutional Animal Care and Use Committee of the University of Pennsylvania. Transgenic lines ptf1a:egfpjh and ins:mCherryjh2  have been previously described.
The +6ptf1a, +11ptf1a and +24ptf1a fragments were produced by NcoI digest of the BAC CH73-141D4 and cloned into a Tol2 vector with a minimal mouse cFos promoter and an egfp reporter followed by the SV40 polyadenylation signal. The remaining non-coding sequences were amplified with the primers listed in Table 1.
Amplicons produced with primers containing att sites were cloned in the pDONR221 vector (Invitrogen) through Gateway BP recombination (Invitrogen). Amplicons without att sites were cloned in the Gateway entry vector pCR8-TOPO (Invitrogen). The synthetic autoregulatory enhancer was generated by annealing the complementary oligos Synth-S and Synth-AS (Table 1). The annealed double stranded product was then cloned in the pCR8-TOPO vector. All non-coding sequences were cloned into the pGW-cfosEGFP [27, 28] vector through Gateway LR recombination (Invitrogen).
BAC recombineering was performed basically as described ; primer sequences are listed in Table 1. To delete the region containing dlgap1a, FRT3 sites flanking the gene were introduced by recombination. To insert the site upstream of dlgap1a, the primers 35hom-FRT3-tetR and 35hom-tetF were used to generate a targeting cassette. Then the CH211-142H2 BAC, modified by replacement of the ptf1a coding sequence with mCherry, was grown in the EL250 bacterial strain to OD600=0.6 at 32°C, followed by manual mixing in a 42°C waterbath for 15 minutes to induce recombination proteins. Electrocompetent cells were prepared and transformed with the first targeting cassette following standard protocols with an Eppendorf Electroporator 2510. After a one hour recovery, recombinant clones were selected with the tetracycline. The targeting cassette to introduce a second FRT3 site close to the 3′ end of dlgap1a was generated using the primers 35hom-FRT3-KanF and 35hom-KanR, and transformed as above; recombinants were selected with kanamycin. After the sequential recombination steps that introduced FRT3 sites flanking the dlgap1a region, recombination of the two sites and excision of the 93kb sequence was achieved by FLPe induction with arabinose treatment (0.1% for 1hr at mid-log phase). Recombinant clones at each step were verified through PCR or sequencing.
Additional BAC modifications were made similarly, through sequential transformations and selections, followed by FLPe-mediated recombination. To generate a ptf1a coding sequence resistant to translation blocking MO, ptf1a cDNA was first amplified from the IMAGE clone using primers EcoRI-KzZPtf1a-F and XhoI-zPtf1a-R, and then flanked by 300bp homology arms. An FRT-flanked selection cassette was amplified using primers PacIFRTKanF and PacIFRTKanR and added downstream of the ptf1a cDNA. The entire coding sequence targeting cassette was then amplified using the primers 300promF and 3′UTR-R, before transformation and antibiotic selection. To make the Δ-5-3 BAC, the primer pairs 5hom-F3-KanF/5hom-KanR and 3hom-SpecF/3hom-F3-SpecR were used to generate targeting cassettes to flank the autoregulatory enhancer with FRT sites. Similarly, to make the Δ+1+6 BAC, the primer pairs 1homF3-KanF/1hom-KanR and 6hom-SpecF/6hom-F3-SpecR were used to amplify the fragments used to flank the early enhancer with FRT sites. For efficient transgenesis, a cassette containing the inverted Tol2 arms was targeted to the chloramphenicol resistance gene of the BAC backbone, as previously described . Zebrafish embryo injections. Transgenesis with Tol2–based plasmid constructs was performed as previously described [27, 28]. Unless otherwise noted, consistent reporter gene expression was observed in at least three independent lines for each construct. For BAC transgenesis each embryo was injected with 100pg BAC DNA and 40pg Tol2 transposase RNA. To block ptf1a translation, embryos were injected with 0.5pmoles of ptf1a–MO:CCAACACAGTGTCCATTTTTTGTGC (Gene Tools) as previously described .
Embryos were visualized by epifluorescence on a Zeiss V12 Stereomicroscope, and imaged with AxioVision 4.5 software. For confocal imaging, embryos were anesthetized in Tricaine and mounted in MatTek glass bottom culture dishes in 1% low melt SeaPlaque agarose. Images were acquired on an Olympus IX 81 microscope equipped with a Yokogawa CSU 10 scan head with a Hamamatsu EMCCD camera (model C9100-13, Bridgewater, NJ). Diode lasers for excitation (488 nm for eGFP and 561 nm for mCherry) were housed in a Spectral Applied Research launch (Richmond Hill, Ontario). Hardware was controlled by Slidebook version 5.0 (Intelligent Imaging Innovations), and confocal image stacks were processed with ImageJ (http://rsbweb.nih.gov/ij).
Given the hypothesized importance of Ptf1a autoregulation in mouse pancreas  we sought to determine if Ptf1a autoregulation is a shared feature in teleosts. Precalculated multiple sequence alignments generated by the PHAST package did not identify sequences homologous to the characterized mouse autoregulatory enhancer in any of the teleost species with available genome sequences (zebrafish, medaka, stickleback, fugu and tetraodon). We hypothesized that Ptf1a autoregulation is conserved in teleosts, but that sequence homology is confined to the enhancer’s transcription factor binding sites. Therefore, we searched for the motifs CANNTGN6TTCCC (one helical turn spacer) and CANNTGN16TTCCC (two helical turns spacer), which are based on the PTF1 binding sites in the Rbpjl promoter  and the Ptf1a autoregulatory enhancer . We identified 3 matches to these motifs in the region between 5kb and 3kb upstream of the zebrafish ptf1a transcriptional start site (Fig. 1). There were no other matches to the PTF1 consensus binding sequence in the 50kb on either side of the zebrafish ptf1a gene, suggesting that this is the only element mediating autoregulation of ptf1a expression in zebrafish. Further, we identified 2–4 matches to the consensus PTF1 binding sequence within ~5kb upstream of the other teleost ptf1a genes (Fig. 1A), strongly supporting a function for these sites in ptf1a autoregulation.
We tested the ability of sequences containing the predicted zebrafish PTF1 binding sites to regulate expression of a fluorescent reporter gene, in conjunction with a heterologous minimal promoter, in transgenic fish (Fig. 1C, D). A sequence containing all three predicted sites had activity in the pancreas and the hindbrain, consistent with function as an autoregulatory enhancer. We further determined that a shorter fragment containing the two most proximal predicted PTF1 binding sites was able to similarly recapitulate endogenous ptf1a expression. Site-directed mutagenesis of these two sites abolished enhancer activity (Suppl. Fig. 1). We also generated a 90 bp synthetic enhancer containing the three PTF1 sites. The synthetic construct regulated an expression pattern similar to the intact enhancer, indicating that the three PTF1 binding sites are sufficient for in vivo activity of the enhancer (Fig. 1E, F). The synthetic enhancer exhibited activity in the retina, which although it is an endogenous site of expression, is not observed with the full autoregulatory enhancer. This observation suggests that additional sequences in the autoregulatory enhancer function to suppress its activity in the retina.
We next sought to identify all additional cis–regulatory elements associated with ptf1a, in particular those that initiate expression in the pancreas and cerebellum prior to activation of the autoregulatory enhancer. We examined the non-coding sequence between the upstream dlgap1a gene (17kb upstream) up to 48kb downstream of ptf1a for regulatory activity. We divided the sequence of this interval into 7 consecutive fragments (Fig. 2A). These were tested as described above for their ability to direct reporter gene expression to sites of endogenous ptf1a expression in transgenic zebrafish. We identified several enhancers that regulated expression of the transgene in the nervous system, including the hindbrain, retina and spinal cord (Fig. 2). Three enhancers, +6ptf1a, +11ptf1a and +24ptf1a, organized consecutively downstream of ptf1a, regulated expression in different regions of the hindbrain (Fig 2 C–E, H–J). Two of these elements, +6ptf1a and +11ptf1a, also exhibited robust activity in the anterior hindbrain, including the area corresponding to the cerebellar primordium (Fig. 2 H, I). Three regulatory elements were active in the spinal cord, specifically +6ptf1a, +11ptf1a and +24ptf1a (Fig 2E and data not shown). Finally, three regulatory elements (+6ptf1a, +11ptf1a, and +36ptf1a) were active in the retina (Fig. 2C, D, F). We further compared the expression patterns regulated by several neuronal enhancers to that of the intact locus in double transgenic animals (Suppl. Fig. 2). We find that the hindbrain expression of +11ptf1a:EGFP and the spinal cord expression of +24ptf1a:EGFP extend to more dorsal domains than the expression from the intact locus. These results suggest that genomic context and interactions with surrounding regulatory elements are necessary to faithfully produce the correct expression pattern.
The farthest upstream sequence tested, −17ptf1a, directed expression ubiquitously in neural tissue (Fig. 2B, G), an activity that might be related to the upstream dlgap1a gene, which encodes for a synaptic protein . Additionally, the most downstream sequence tested, +36ptf1a, was active in the notochord (Fig. 2F), expression likely related to the downstream gene si:ch211-207d6.2, whose murine homologue Sickle tail is expressed in the notochord and required for proper vertebrae development .
To test whether the interval described above contains all necessary regulatory sequences, we utilized a bacterial artificial chromosome (BAC) transgene that faithfully recapitulates endogenous ptf1a expression . The neighboring gene upstream of ptf1a differs between teleosts and mammals. We expect that regulatory elements conserved among vertebrates will remain syntenic to the gene throughout evolution and therefore we predicted that conserved regulatory elements would lie within the region bounded by the nearest upstream gene, dlgap1a. We deleted 93kb of upstream sequence, including the dlgap1a gene (Fig. 3A). We showed that the truncated BAC (with 17kb of direct upstream sequence and 50kb of downstream sequence) has activity identical to the original BAC (Fig. 3B–D) and therefore contains all necessary regulatory elements for proper ptf1a expression.
Within the tested interval, we identified a single region, +1ptf1a, which regulated expression in the early pancreas. The cells in which +1ptf1a is active are in close proximity to but do not overlap with the dorsal bud derived β-cells, marked by the expression of the insulin:mCherry transgene (Fig. 3E–J). This localization is consistent with the position of the ventral bud derived pancreatic progenitors that surround the earlier appearing cluster of endocrine cells of the principal islet. Upon expansion of the pancreas and cytodifferentiation of exocrine cells, +1ptf1a:egfp activity persisted in a limited region surrounding the principal islet at least until 5dpf (Suppl. Fig. 3). The persistent fluorescence may represent GFP perdurance in slowly dividing cells, or perhaps correspond to ongoing newly generated exocrine precursors in which the early enhancer continues to be active. The +1ptf1a element spans 5kb immediately downstream of the ptf1a gene and has a high degree of sequence conservation to other species, including human and mouse, suggesting a conserved functional role. Specifically, this interval contained a 527 bp sequence that is 79% identical between zebrafish and human, and 96% identical between mouse and human (Fig. 3K). We tested the highly conserved shorter sequence for its ability to direct expression to the pancreas, but found that, in isolation, it lacks enhancer activity.
To further characterize the spatial and temporal activities of the two pancreatic enhancers, we compared expression regulated by them to that of the intact ptf1a locus. Embryos transgenic for the early enhancer, the autoregulatory enhancer or the entire BAC were assayed by in situ hybridization for pancreatic egfp expression at 4 hour intervals between 30hpf and 46hpf (Fig. 4). Consistent with previous reports, expression from the intact BAC was first seen at 34hpf (24/24 embryos, Fig, 4A). While expression in the +1ptf1a:egfp transgenics was variable, a fraction of embryos had pancreatic expression at 34hpf (Fig. 4B) and expression was observed in the majority of embryos (28/36) by 38hpf (Fig. 4D). At all time points, the expression regulated by the +1ptf1a enhancer was in a subset of cells in the domain in which the intact locus showed expression. In contrast, the autoregulatory enhancer did not regulate detectable pancreatic expression until 42hpf (Fig. 4H), also in a smaller group of cells than those expressing the intact locus.
We identified two regulatory elements apparently controlling ptf1a expression in the pancreas: an early enhancer that initiates expression, and an autoregulatory enhancer that maintains or augments expression. Because the early enhancer has restricted spatial and temporal activity, and because we failed to identify additional enhancers that control ptf1a expression in the mature pancreas, we hypothesized that autoregulation has a major role in development of the pancreas following its initial specification. To test this hypothesis, we examined the effect of the lack of autoregulatory ptf1a control on pancreatic development. We altered a BAC containing ptf1a through the introduction of synonymous substitutions in the coding sequence, rendering the transcript resistant to a translation–blocking morpholino (MO) that efficiently depletes endogenous PTF1a and inhibits development of the ventral pancreatic bud . we further deleted the entire autoregulatory element from the BAC, generating the Δ-5-3 BAC transgene that expresses ptf1a under control of the full complement of characterized regulatory elements minus the autoregulatory enhancer. We subsequently established stable transgenic lines with the two BAC transgenes.
We tested the ability of the Δ-5-3 BAC to rescue pancreatic development, by MO injection into embryos transgenic for the ptf1a:egfp BAC. In control embryos, MO injection almost completely abolished visible exocrine pancreas tissue at 3 or 5 dpf (4 partial pancreas/137 embryos). Presence of the BAC encoding MO-resistant ptf1a rescued pancreas development, wholly or partially, in a substantial fraction of morphants (22/88 embryos, or 50% of expected 44 BAC carriers). In contrast, the Δ-5-3 BAC failed to rescue pancreas development (1 partial pancreas/90 embryos, or 45 expected BAC carriers). The Δ-5-3 BAC rescues represent the combined results of rescues performed in 2 independent transgenic lines. These findings indicate that ptf1a autoregulation has a very critical role during early pancreatic development, and lack of the relevant transcriptional activity leads to an early failure of the pancreas to expand.
To address the role of the +1ptf1a enhancer in pancreas specification, we similarly engineered a BAC encoding for MO-resistant ptf1a transcript and lacking the 5kb region corresponding to the +1ptf1a enhancer, the Δ+1+6 BAC. As above, we depleted endogenous ptf1a through MO injections in ptf1a:egfp transgenics. We then compared the abilities of the Δ+1+6 BAC and the intact BAC to rescue pancreas formation in 5dpf mosaic transgenic larvae. In MO-injected embryos, none (0/63) had visible exocrine pancreas tissue at 5 dpf, In embryos injected with MO and the intact BAC, a fraction of larvae had pancreas tissue (16/196) at 5dpf, consistent with the ability of the same BAC to rescue the MO phenotype in established transgenic lines. Surprisingly, pancreas tissue was observed in a larger fraction (43/209) of embryos injected with MO and the Δ+1+6 BAC. While different degrees of mosaicism between embryos leads to variability in these transient assays, comparison of the best pancreas rescues seen for each BAC demonstrates that the control BAC was able to fully rescue pancreas formation (fig. 6C,E,G) while the BAC lacking the early enhancer was only able to promote the formation of a hypoplastic pancreas (fig. 6D,F,H); the mean pancreas size in the embryos receiving the BAC lacking the early enhancer was only 35% the average size of the pancreas rescued by the intact BAC (Fig 6B).
Ptf1a has demonstrated importance in neuronal specification and development of the exocrine pancreas. However, prior to our work, identified cis–regulatory elements could only account for part of the transcriptional regulation of the gene. Therefore, we undertook a comprehensive, unbiased analysis of the zebrafish ptf1a locus to fully characterize its regulatory context. Previous work had established the presence of an autoregulatory enhancer upstream of mouse Ptf1a [12, 24]. Based on identification of consensus binding site sequences for PTF1, we identified a similarly functioning enhancer upstream of the zebrafish gene. The region we identified as the autoregulatory enhancer contains three potential PTF1 binding sites, which we showed are both necessary and sufficient for its function. We also computationally identified similar regions upstream of multiple teleost ptf1a genes, suggesting that autoregulation of ptf1a is a deeply conserved feature. It is important to note that sequence conservation tools did not identify the autoregulatory sequences in teleosts, and therefore the absence of recognizable sequence conservation does not preclude the presence of functional conservation.
Ptf1a has an important role in promoting inhibitory fates in several neuronal populations. Previous work in mice identified a downstream regulatory element with activity in the hindbrain, spinal cord and retina [12, 24], although it did not direct expression to the cerebellum and additionally showed ectopic activity in populations of excitatory neurons of the spinal cord. Here we report the identification of several independent enhancers in zebrafish that recapitulate different subsets of the endogenous expression pattern in the hindbrain, spinal cord, and the cerebellum. Surprisingly, although the +6ptf1a sequence is highly similar to the orthologous sequence in the mouse, a larger transgene encompassing the orthologous sequence showed no cerebellar activity in transgenic mice. This discrepancy could be due to divergence of the regulatory sequences in the two species or the presence of additional regulatory elements in the zebrafish sequence. We additionally identified multiple enhancers with activity in different regions of the hindbrain, suggesting the presence of diverse regulatory mechanisms activating ptf1a expression in different regions of the hindbrain. Whether these enhancers with partially overlapping expression patterns synergize in vivo to create exact borders of expression as suggested elsewhere  or whether these enhancers operate in a modular fashion will be interesting to determine in the future.
The region immediately downstream of zebrafish ptf1a was able to drive expression in the pancreatic progenitors, an activity not found for the orthologous mouse sequence . While this discrepancy could be due to divergence between the mouse and zebrafish sequences, other reasons could account for the differences. First, the mouse downstream regulatory sequence excluded 330bp corresponding to the 3′ UTR and a further 536bp of poorly conserved sequence directly downstream of the mouse 3′UTR. It is possible that the early pancreatic activity requires sequences in this region. We also found that the activity of this enhancer was sensitive to position effects, necessitating the screening of a relatively large number of founders to identify those with early pancreatic expression. The extremely high degree of conservation of this downstream region suggests that the mechanisms controlling initiation of ptf1a expression in the pancreas are conserved between teleosts and mammals. There are predicted binding sites for Hnf1, Sox9, and Pdx1 conserved in this region from teleosts to mammals, all attractive candidates for regulators of early ptf1a expression and specification of the exocrine progenitor cells. With the identification of this enhancer, the in vivo role for all of these factors in inducing ptf1a and subsequent pancreatic development can be addressed directly.
We tested the requirement for the identified early enhancer region in exocrine pancreas specification. Somewhat surprisingly, we found that a BAC lacking the early enhancer region was capable in mosaic injected larvae of rescuing the formation of exocrine pancreas tissue, and in fact did so at a higher frequency than the intact BAC, although the average size of rescued pancreatic tissue was significantly smaller. These data suggest that there are additional sequences outside of the deleted enhancer region that can lead to a low level of ptf1a expression; once the expression meets the threshold to activate the autoregulatory enhancer, it is self-sustaining. Interestingly, injection of the Δ+1+6 BAC into ptf1a:egfp transgenics in the absence of MO leads to ectopic GFP expression in a fraction of embryos, in some cases even apparent development of pancreatic tissue bilaterally (data not shown). We hypothesize that the deleted region contains both positive regulatory elements that increase the level of ptf1a in the early pancreas, but also negative regulatory sequences that help to prevent ectopic expression in adjacent non-pancreatic endoderm.
It has been hypothesized that Ptf1a autoregulation plays a role in late pancreatic development and in the mature pancreas to stabilize commitment to the exocrine lineage . As a direct test of this hypothesis, we removed the autoregulatory enhancer from a BAC transgene containing the entire regulatory region of ptf1a. In contrast to the previously described ptf1a hypomorphic mutant , Ptf1a encoded by the BAC lacking the autoregulatory enhancer is fully functional to activate all of its target genes. However, in the absence of expression from the endogenous ptf1a locus, as in a null mutant or morphant, the ptf1a locus on the BAC is incapable of regulating its own expression. In the absence of this autoregulatory loop, we find that ptf1a cannot support development of the exocrine pancreas. Our results thus reveal a requirement for ptf1a autoregulation much earlier during embryonic development than expected. These results demonstrate that ptf1a autoregulation has a critical role in pancreatic development and unlike other developmental regulatory elements, whose deletion might cause no effect or only produce effects under environmental stress [35, 36], the autoregulatory function is strictly required for ventral pancreas expansion. The presence of sequences with predicted autoregulatory function in the teleost loci described here and in other vertebrates  suggests that autoregulation is a well conserved and critical aspect of ptf1a expression, a hypothesis supported by our functional BAC rescue assays.
Our work has produced a more complete picture of Ptf1a regulation, including the first evidence of regions regulating initial expression in the pancreas and cerebellum, the two structures most severely affected in mutants. It will now be possible to test specific hypotheses about transcription factors and signaling pathways required for Ptf1a induction. In addition, we have directly tested the requirement for autoregulation at the locus, and shown that it is critical for early development of the exocrine pancreas. The transgenic model we have generated will also be useful in better understanding the role of Ptf1a autoregulation in the allocation and commitment of pancreatic cells to the exocrine and endocrine lineage. Previous studies have shown that maintaining Ptf1a expression during embryonic development blocks endocrine differentiation  and antagonizing ptf1a function postembryonically can lead to an exocrine–to–endocrine conversion . Our BACs lacking the autoregulatory enhancer can further be used to examine the requirement for ptf1a autoregulation in processes of pancreatic development, such as the expansion and maintenance of pancreatic progenitors and the balance of exocrine and endocrine progenitors. These efforts will ultimately help in understanding cell fate choices in the pancreas and enable manipulations of these processes towards the generation of desired cell types.
A) Expression from the transgene containing the two proximal PTF1 binding sites, in conjunction with the heterologous cFos minimal promoter at 2 dpf. Mutation of the E and TC boxes in the two PTF1 binding sites results in the loss of any tissue-specific activity of the enhancer (B).
All embryos were doubly transgenic for the ptf1aΔdlgap1:mCherryBAC, and the indicated neural enhancer regulating expression of egfp. The +6ptf1a enhancer (B) properly recapitulates the activity of the entire BAC (C) along the dorso-ventral axis of the hindbrain. By comparison, the +11ptf1a enhancer (F) shows ectopicactivity in more dorsal regions of the hindbrain (arrowhead) compared to the intact BAC (G). Similarly, the spinal cord activity of the +24ptf1a enhancer (J) extends more dorsally (arrow) compared to the entire BAC (K). All embryos were imaged at 26hpf.
At 38hpf the early enhancer is active in pancreatic progenitors (A) similar to the entire BAC (B), with significant overlap of their expression domains (C). At 4dpf the early enhancer has limited activity, restricted to a domain surrounding the principal islet (D), compared to broader expression from the intact BAC (E).
The work was funded by NIH grants HG005039 to S.F and DK61215 and DK56211 to S.L. We thank Michael Parsons for the kind gift of insulin:mCherry transgenic fish. We thank Laura Del Collo, Paula Roy, and Gui Hu for excellent fish care and technical assistance.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.