NKX6 transcription factor activity is required for α - and β -cell development in the pancreas

well as β -cell development. We demonstrate that expression of Myt1 depends on overall Nkx6 gene dose, and therefore identify Myt1 as a possible downstream target of Nkx6 genes in the endocrine differentiation pathway. by the Bradford dye-binding assay. The concentrations of insulin and glucagon were determined by radioimmunoassay (RIA) using commercially available kits (Linco). To obtain a representative average of the number of hormone-positive cells, an entire pancreas was used for quantiﬁcation. Immunoﬂuorescence staining was performed on 10 µ m sections and positive cells counted on every ninth section throughout the pancreas at E16.5 and E18.5, every ﬁfth section at E12.5 and E14.5, and all sections at E10.5. The average cell number per section was determined for all sections counted from each individual pancreas. Mean differences were tested for statistical signiﬁcance using the Student’s t -test.


Introduction
The possibility of developing a cell-based therapy for treatment of diabetes mellitus has recently generated interest in identifying the molecular pathways that control development of endocrine cells in the pancreas. In mammalian pancreas, endocrine cells are clustered in the islets of Langerhans, in which a large core of insulin-producing β-cells is surrounded by α-, δand PP-cells that produce glucagon, somatostatin and pancreatic polypeptide (PP), respectively. Because of their role in the pathogenesis of diabetes mellitus, the mechanisms that underlie formation of β-cells have been studied in most detail. Although β-cell proliferation appears to be the predominant mechanism of β-cell regeneration during adulthood (Dor et al., 2004), β-cell formation from multipotent precursors accounts for the vast majority of β-cells formed during embryogenesis (Jensen, 2004;Kim and MacDonald, 2002). Therefore, most factors that control β-cell differentiation have been identified through studies in the embryo.
Formation of the mouse pancreas begins at embryonic day (E) 9.5 as separate dorsal and ventral evaginations from the foregut endoderm (Slack, 1995). At this stage, the epithelium contains multipotent progenitors that express the transcription factor PDX1 and have the potential to give rise to all pancreatic lineages, comprising endocrine and exocrine cells, as well as cells of the pancreatic ducts (Gu et al., 2002;Herrera, 2002). The first endocrine cells appear as early as E10.5 and produce glucagon-and/or insulin, but lack specific products characteristic of mature hormone-producing cells (Oster et al., 1998;Wilson et al., 2002). Mature insulin-and glucagonproducing cells, as well as cells expressing the exocrinespecific markers amylase and carboxypeptidase A are first detected around E13.5. The first δand PP-cells are found at E15.5 and E18.5, respectively (Pictet and Rutter, 1972;Slack, 1995).
It has been shown that specific transcription factors restrict the developmental potential of the initially multipotent pancreatic progenitors and promote their differentiation into specific cell types (Jensen, 2004). This is exemplified by the basic helix-loop-helix (bHLH) transcription factor NGN3, which limits the potential of a specific subset of pancreatic precursors to undergo endocrine differentiation while prohibiting an exocrine or ductal cell fate (Gu et al., 2002). The complete absence of hormone-producing cells from the pancreas of Ngn3 -/mice further supports a role for NGN3 in endocrine specification (Gradwohl et al., 2000). Interestingly, ectopic expression of NGN3 in all pancreatic precursors In diabetic individuals, the imbalance in glucose homeostasis is caused by loss or dysfunction of insulinsecreting β-cells of the pancreatic islets. As successful generation of insulin-producing cells in vitro could constitute a cure for diabetes, recent studies have explored the molecular program that underlies β-cell formation. From these studies, the homeodomain transcription factor NKX6.1 has proven to be a key player. In Nkx6.1 mutants, β-cell numbers are selectively reduced, while other islet cell types develop normally. However, the molecular events downstream of NKX6.1, as well as the molecular pathways that ensure residual β-cell formation in the absence of NKX6.1 are largely unknown. Here, we show that the Nkx6.1 paralog, Nkx6.2, is expressed during pancreas development and partially compensates for NKX6.1 function. Surprisingly, our analysis of Nkx6 compound mutant mice revealed a previously unrecognized requirement for NKX6 activity in α-cell formation. This finding suggests a more general role for NKX6 factors in endocrine cell differentiation than formerly suggested. Similar to NKX6 factors, the transcription factor MYT1 has recently been shown to regulate αas well as β-cell development. We demonstrate that expression of Myt1 depends on overall Nkx6 gene dose, and therefore identify Myt1 as a possible downstream target of Nkx6 genes in the endocrine differentiation pathway.
largely results in excess differentiation of α-cells, but not of other endocrine cell types. This indicates that although able to confer endocrine identity to early pancreatic progenitors, NGN3 requires additional factors for the differentiation of β-, δand PP-cells (Apelqvist et al., 1999;Schwitzgebel et al., 2000).
Among the transcription factors that are required for β-cell development are NKX2.2, PAX4, HB9 and NKX6.1. In Nkx2.2 and Pax4 mutants, loss of β-cells results from a switch to an alternative endocrine fate (Prado et al., 2004;Sosa-Pineda et al., 1997), whereas HB9 has been implicated in the control of β-cell maturation (Harrison et al., 1999;Li et al., 1999). Mice deficient for the NK-homeodomain factor NKX6.1 have a specific defect in β-cell neogenesis, while all other endocrine cell types develop normally (Sander et al., 2000). In Nkx6.1 mutants, β-cell development is disrupted only after E13.5, when the first mature β-cells appear. This suggests that the differentiation of early-appearing insulin-positive cells and mature β-cells is controlled by independent mechanisms and implies a selective role for NKX6.1 in just the major β-cell differentiation pathway.
The observation that a significant number of β-cells still form in the absence of NKX6.1 even after E13.5 points to the existence of an NKX6.1-independent pathway of β-cell development. In this study, we examined the Nkx6.1 paralog Nkx6.2 for its function and possible synergy with NKX6.1 in pancreatic development. In contrast to NKX6.1, we found NKX6.2 function to be dispensable for normal endocrine development. However, our analysis of Nkx6.1;Nkx6.2 double nullizygous mice not only revealed a partially compensatory role for NKX6.2 in β-cell development, but an as of yet unappreciated requirement for NKX6 activity in α-cell formation. We further discovered that NKX6 activity controls expression of the NGN3 co-factor MYT1 in a dose-dependent manner, thus providing a possible mechanism by which NKX6 activity may control endocrine development.

Mutant mice and BrdU injections
Nkx6.1 and Nkx6.2 mutant mice were generated by gene targeting as previously described (Sander et al., 2000;Vallstedt et al., 2001). Compound Nkx6 mutant embryos were obtained by crossing Nkx6.2 tlz/+ ;Nkx6.1 +/mice in timed matings. Mid-day of the day on which the vaginal plug was detected was considered as day E0.5. All genotyping was performed by Southern blot analysis as described (Sander et al., 2000;Vallstedt et al., 2001). For BrdU labeling experiments, pregnant females were injected with 50 µg BrdU per gram of body weight i.p., and embryos harvested 45 minutes after injection.
Immunohistochemistry, in situ hybridization, TUNEL assay and X-gal staining Pancreata were removed from adult mice and embryos at E15.5 and later stages; pancreatic tissue in embryos at earlier stages was studied in whole embryos. Samples were fixed in 4% paraformaldehyde in PBS and either paraffin-embedded or frozen in OCT.
TUNEL assays on tissue sections were performed using a commercially available kit (Oncor).

RNA preparation and real-time quantitative PCR
Total RNA from dissected pancreatic anlagen was extracted with the RNeasy kit (Qiagen) and treated with DNAse. cDNA was prepared by in vitro transcription using SuperscriptII reverse transcriptase (Invitrogen). PCR reactions were performed in triplicate in a total reaction volume of 50 µl, and amplifications performed in an ABI Prism 7700 sequence detecting system (Applied Biosystems). With the exception of Myt1, which was detected with Taqman Universal PCR Mastermix (Applied Biosystems), all transcripts were amplified with 1ϫ SYBR Green PCR master mix (Applied Biosystems) and 300 nM of each primer. To exclude contamination with non-specific PCR products, melting curves were analyzed for all PCR products. The following cycle was used for the amplification: 50°C for 2 minutes; 95°C for 10 minutes; followed by 40 cycles of denaturation at 95°C for 15 seconds; and primer extension at 60°C for 1 minute. For each reaction, a parallel reaction that lacked template was performed as a negative control. Relative changes in gene expression were calculated by the comparative ∆Ct method in which γ-actin was used for normalization with the SYBR Green method and HPRT with the Taqman method (Livak and Schmittgen, 2001). Listed 5′ to 3′, primer sequences were as follows: γ-actin forward, GCACCCGGTGCTTC-TGAC; γ-actin reverse, CCAGATGCATACAAGGAC; glucagon forward, TTCCCAGAAGAAGTCGCCATT; glucagon reverse, TC-CCTGGTGGCAAGATTATCC; insulin forward, CCACCCAGGC-TTTTGTCAAA; insulin reverse, CCCAGCTCCAGTTGTTCCAC. The Taqman gene expression IDs were Mm00456190_m1 for Myt1 and Mm00446968_m1 for HPRT.

Hormone quantification and cell counting
Protein was extracted from individual pancreata of E18.5 embryos using acid extraction and protein concentration was then determined by the Bradford dye-binding assay. The concentrations of insulin and glucagon were determined by radioimmunoassay (RIA) using commercially available kits (Linco).
To obtain a representative average of the number of hormonepositive cells, an entire pancreas was used for quantification. Immunofluorescence staining was performed on 10 µm sections and positive cells counted on every ninth section throughout the pancreas at E16.5 and E18.5, every fifth section at E12.5 and E14.5, and all sections at E10.5. The average cell number per section was determined for all sections counted from each individual pancreas. Mean differences were tested for statistical significance using the Student's t-test.

NKX6.2 is transiently expressed in the developing mouse pancreas
To determine the pattern of Nkx6.2 expression, we first performed enzymatic X-gal staining on mouse embryos, in which the Nkx6.2-coding sequence was replaced by a Tau-lacZ cassette (Nkx6.2 tlz mice) (Vallstedt et al., 2001). At E9.5, βgalactosidase (β-gal) activity was selectively detected in both the dorsal and the ventral pancreatic rudiments (Fig. 1A). By E15.5, β-gal was expressed at high levels in the dorsal and at lower levels in the ventral aspect of the pancreas (Fig. 1F). Additional β-gal activity was found in the distal part of the stomach and the duodenum (Fig. 1F). In contrast to Nkx6.1, which is maintained in adult β-cells (Sander et al., 2000), Nkx6.2 expression rapidly declined in late embryogenesis. Only few scattered β-gal-positive (+) cells were detected at E18.5 and no expression was seen in the adult pancreas (data not shown).
To examine which pancreatic cell types express NKX6.2, we performed co-immunofluorescence analyses with an anti-NKX6.2 antibody, together with antibodies against various pancreatic markers. Demonstrating specificity of the anti-NKX6.2 antibody, NKX6.2 and β-gal colocalized in cells of .5, lacZ expression is detectable in the dorsal (dp) and ventral (vp) pancreatic anlagen. (B) Anti-β-gal and anti-NKX6.2 staining show complete overlap in the stomach (s) and dorsal pancreas (dp) in Nkx6.2 tlz/+ embryos. In the pancreas, NKX6.2 is detected in a subset of PDX1 + cells (yellow cells in C), but not in glucagon + (Glc) cells at E10.5 (arrowhead in D). NKX6.2 is also found in the stomach epithelium (C). At E12.5, NKX6.2 expression is largely restricted to the glucagon + cells (E). At E15.5, lacZ expression is predominantly detected in the dorsal, but also in cells of the ventral pancreas (vp) (F), but also found in the distal stomach and duodenum (d) (F). (G,H) NKX6.2 is absent from both NGN3 + (G) and insulin + (Ins) cells (arrowhead in H). By contrast, NKX6.2 colocalizes with glucagon and amylase (Amy) (I). Not all glucagon + and amylase + cells express NKX6.2 (arrowheads in I). At E15.5, the majority of NKX6.2 + cells coexpress PDX1 (J). In contrast to NKX6.2, β-gal occasionally colocalizes with NGN3 (open arrowhead in K) and insulin (open arrowhead in L) in Nkx6.2 tlz/+ embryos. However, the majority of NGN3 + or insulin + cells are β-gal negative (K, and arrowhead in L). Like NKX6.2, β-gal is expressed in a subset of glucagon + cells (open arrowhead indicates co-expressing cell; arrowhead indicates non-co-expressing cell in M) and in exocrine cells, as shown by co-localization of β-gal and amylase on adjacent sections (arrow in M and N). Scale bar: 20 µm. the distal stomach epithelium and dorsal pancreas in Nkx6.2 tlz/+ embryos (Fig. 1B), whereas no NKX6.2 staining was detected in Nkx6.2 tlz/tlz embryos (see Fig. S1B in the supplementary material). At E10.5, when PDX1 marks the entire pancreatic epithelium as well as the prospective distal stomach and duodenum ( Fig. 1C; see Fig. S1C in the supplementary material) (Offield et al., 1996), NKX6.2 colocalized with PDX1 in the pancreatic and stomach epithelium (Fig. 1C). Notably, NKX6.2 was found in a large percentage (~70%) of, but not in all, PDX1 + cells, and was absent from the early glucagon-expressing cells at E10.5 (Fig. 1D). By E12.5, NKX6.2 expression became restricted to a few epithelial cells, of which a significant proportion expressed glucagon (Fig. 1E). At E15.5, the pancreatic epithelium contains undifferentiated pancreatic progenitors, which include the NGN3 + endocrine progenitors, as well as already differentiated α-, βand exocrine cells. Nkx6.2 is not part of the endocrine progenitor pool at E15.5 or prior, as NKX6.2 does not colocalize with NGN3 ( Fig. 1G, data not shown). Likewise, NKX6.2 was not detected in insulin-producing β-cells (Fig. 1H). Instead, we found that all NKX6.2 + cells co-expressed either glucagon or amylase (Fig. 1I). As both endocrine and exocrine cells express PDX1 at E15.5, this finding is consistent with the observation that the vast majority of NKX6.2 + cells were PDX1 + (Fig. 1J). Interestingly, only a subset of, but not all αand exocrine cells expressed NKX6.2 (Fig. 1I). Together with the finding that NKX6.2 expression disappears from the pancreas around birth, these observations indicate that maturation of αand exocrine cells coincides with the downregulation of NKX6.2.
As NKX6.2 is not expressed in NGN3 + endocrine progenitors or β-cells (Fig. 1G,H), we asked whether PDX1/NKX6.2 co-expressing progenitors could give rise to NGN3 + or insulin + cells. To test this issue, we used the highly stable β-gal protein as a tracer for the fate of Nkx6.2-expressing cells. Similar to NKX6.2 protein (Fig. 1I), we found β-gal to colocalize with glucagon and amylase (Fig. 1M,N). However, in contrast to NKX6.2 protein, a number of β-gal + cells coexpressed either NGN3 or insulin (Fig. 1K,L). Given the stability of β-gal protein, these data are consistent with the idea that Nkx6.2-expressing cells differentiate into NGN3 + endocrine progenitors and ultimately into β-cells.

NKX6.1 and NKX6.2 have distinct domains in the pancreas
The finding that NKX6.2 colocalizes with glucagon and amylase, but is absent from insulin-producing cells and NGN3 + endocrine progenitors suggests that it is expressed in a separate domain from NKX6.1, which is found in β-cells and NGN3 + cells (Fig. 3A,B). We directly tested this hypothesis by coimmunofluorescence staining for NKX6.1 and NKX6.2. At E10.5, we observed five different populations of cells in the pancreatic epithelium ( Fig. 2A,B): first, cells that exclusively expressed either NKX6.1 (~33% of all NKX6 + cells) or NKX6.2 (~16%); second, a small population of cells that expressed both NKX6 proteins at similar levels; and finally, cells that produced low levels of one NKX6 factor, but high levels of the other (all three populations of co-expressing cells account for ~51% of NKX6 + cells). At E12.5, the vast majority of pancreatic epithelial cells exclusively expressed NKX6.1 (~86% of all NKX6 + cells), and only a few, scattered cells coexpressed NKX6.1 and NKX6.2 (~3%) (Fig. 2C). Consistent with the absence of NKX6.2 from insulin + and NGN3 + cells (Fig. 1G,H), NKX6.1 and NKX6.2 were found in entirely separate domains at E15.5 (Fig. 2D). Thus, with the exception of the early, undifferentiated epithelium, NKX6.1 and NKX6.2 mark different cell populations in the pancreas.
Nkx6.2 is regulated by NKX6.1 Similar to our observations in the pancreas, a progressive segregation of the NKX6.1 and NKX6.2 expression domains was also noted during spinal cord development (Vallstedt et al., 2001). Moreover, it was observed that NKX6.1 represses the expression of NKX6.2 in the spinal cord, providing a possible mechanism for how their exclusive expression domains are established and maintained.
To determine if a similar mechanism operates in the pancreas, we examined whether absence of NKX6.1 affects expression of NKX6.2. Although we did not detect any difference in the number of NKX6.2 + cells between wild-type and Nkx6.1 mutants at E10.5 (Fig. 3Q), their number was increased in Nkx6.1 mutants at E12.5 (Fig. 3E,F,Q). Likewise, microarray experiments from whole pancreatic epithelium showed a significant upregulation of Nkx6.2 mRNA in Nkx6.1 mutants at E13.5 (5.2-fold) and E15.5 (2.8-fold) (data not shown). To test if there is mutual cross-repression between the two NKX6 factors, we also analyzed the expression of NKX6.1 in Nkx6.2 mutants, but did not detect an increase in the number of NKX6.1 + cells (data not shown). Therefore, similar to spinal cord (Vallstedt et al., 2001) includes the majority of NGN3 + endocrine progenitors ( Fig.  3A; data not shown). In addition, NKX6.1 was expressed in all insulin + cells, occasionally in glucagon + cells, but was absent from the exocrine lineage ( Fig. 3B-D). In Nkx6.1 mutants, NKX6.2 was not detected in insulin + or NGN3 + cells at E15.5 or prior (Fig. 3H To reveal the identity of the NKX6.2 + cells in Nkx6.1 mutants, we next examined whether ectopic NKX6.2 expression occurs in αor exocrine cells. This did not appear to be the case, as a similar fraction of glucagon-and amylaseproducing cells stained positive for NKX6.2 in wild type and Nkx6.1 mutants (Fig. 3K-N). Together, these findings suggest that NKX6.2 is not ectopically activated in differentiated cells. Therefore, we tested if NKX6.2 + cells co-expressed PDX1, which in addition to βand exocrine cells, marks pancreatic progenitors at E15.5. Indeed, more than 90% of NKX6.2 + cells in Nkx6.1 mutants were also PDX1 + (Fig. 3P). As we did not observe ectopic expression of NKX6.2 in βor exocrine cells, we can infer that upregulation of NKX6.2 in the absence of NKX6.1 most probably occurs in pancreatic progenitors.  Notably, at E18.5, NKX6.2 is no longer detected in either wildtype or Nkx6.1 mutant pancreas, suggesting that NKX6.2 is only transiently upregulated during development.

cells was not affected in
Nkx6.1 -/-;Nkx6.2 -/embryos (Fig. 5D,H). Similar to cell numbers, pancreatic insulin and glucagon mRNA levels were also significantly lower in Nkx6.1 -/-;Nkx6.2 -/than in Nkx6.1 -/embryos (Fig. 4F). Thus, our present findings reveal a requirement for NKX6 activity in the development of both the insulin and the glucagon lineages. The role of NKX6 proteins in α-cell development could not be revealed from the analysis of either Nkx6 single mutant, as the other NKX6 factor fully compensates.
No evidence for endocrine cell conversion in the absence of NKX6 activity A possible mechanism that could account for the reduction in insulin-and glucagon-producing cells is that progenitors differentiate along an alternate path. Such fate conversion has recently been shown to be the cause of β-cell loss in Nkx2.2 and Pax4 mutant mice (Prado et al., 2004), where cells producing the hormone ghrelin are formed at the expense of βcells. To determine if a similar fate conversion underlies the βand α-cell loss in the absence of NKX6 activity, we analyzed the expression of ghrelin in Nkx6 compound mutants, but did not detect an increase in ghrelin + cells (Fig. 5I-L). Together with our finding that the number of somatostatin-and PP-producing cells were normal in Nkx6.1 -/-;Nkx6.2 -/embryos (Fig. 5D,H), these results indicate that endocrine fate conversion does not account for the reduction in insulin-and glucagon-producing cells in NKX6-deficient mice.
An alternative explanation for the reduction in insulin and glucagon cell numbers is that endocrine cells are arrested in their final steps of differentiation and therefore fail to produce hormones. If endocrine precursors were arrested before their final differentiation, one would expect to detect cells that have initiated expression of some endocrine markers, such as IAPP, ISL1 or PAX6. However, we found that the number of IAPP-, ISL1-and PAX6-producing cells mirrored the number of hormone-positive cells in all Nkx6 compound mutants. Although Nkx6.2 mutants resembled wild-type embryos, a gradual reduction in the number of IAPP-, ISL1-and PAX6producing cells was observed with deletion of either Nkx6.1 alone or the combined deletion of both Nkx6.1 and Nkx6.2 (Fig.  6A-L). Likewise, expression of the β-cell markers HB9 and PDX1 was markedly reduced in Nkx6.1 -/and almost absent in Nkx6.1 -/-;Nkx6.2 -/embryos (see Fig. S2I in the supplementary material; data not shown). These results argue against a defect in terminal differentiation of βand α-cells in Nkx6 mutants.

NKX6 activity regulates expression of the NGN3 co-factor Myt1
We next addressed whether loss of both NKX6.1 and NKX6.2 affects the formation of NGN3 + endocrine progenitors. As determined by in situ hybridization and immunofluorescence staining, all Nkx6 compound mutants had similar numbers of NGN3-expressing cells as did wild-type embryos at E14.5 ( Fig. 7A-D; NGN3 + cell numbers as determined by immunofluorescence: Nkx6.2 -/-~8 0% of wild type; Nkx6.1 -/-80% of wild type; Nkx6.1 -/-;Nkx6.2 -/-~92% of wild type). Likewise, in our microarray experiments we found no reduction in Ngn3 mRNA levels in Nkx6.1 mutant pancreas at E13.5 or E15.5 (data not shown). At E16.5, we consistently observed a slight (~20%), but not statistically significant reduction in the number of NGN3 + cells in Nkx6.1 -/and Nkx6.1 -/-;Nkx6.2 -/embryos (Table 1). Notably, in all genotypes, more than 95% of all NGN3 + or PAX6 + cells were BrdU negative (data not shown), indicating that the majority of endocrine progenitors are not in S-phase of the cell cycle. It is therefore unlikely that differences in the proliferative rate of NGN3 + progenitors account for the loss of endocrine cells in Nkx6 mutants. Likewise, as there is no increase in the number of TUNEL + cells (Table 1), cell death does not appear to be the underlying mechanism.
Next, we analyzed whether NKX6 activity could be required for the expression of important endocrine differentiation factors. First, we tested whether the β-cell differentiation factors Pax4 and HB9 are expressed in NKX6-deficient mice. Both factors were normally expressed at E14.5 (see Fig. S2 in the supplementary material), therefore suggesting that Pax4 and HB9 are regulated independently of NKX6 factors.
Recent findings indicate that NGN3 alone is insufficient to induce endocrine differentiation, but requires the zinc-finger transcription factor MYT1 as a co-factor. Expression of a dominant-negative form of Myt1 (DnMYT1) reduces the ability of NGN3 to induce ectopic glucagon expression from chicken endoderm (Gu et al., 2004). Moreover, inhibition of MYT1 activity in endocrine progenitors by DnMYT1 in transgenic mice results in a severe reduction in the number of insulin-and glucagonproducing cells (Gu et al., 2004), suggesting that both NGN3 and MYT1 need to be present for normal αand β-cell development. As the phenotype caused by expression of DnMYT1 resembles our observations in Nkx6.1 -/-;Nkx6.2 -/embryos, we considered the possibility that Nkx6 genes function in a common genetic pathway with Myt1. A first hint that NKX6 factors regulate Myt1 came from our microarray experiments, which showed a significant reduction of Myt1 expression in Nkx6.1 mutants at E13.5 and E15.5 (1.7-and 1.8-fold, respectively; data not shown). To test whether Myt1 expression depends on overall Nkx6 gene dose, we analyzed Myt1 in pancreas of Nkx6 compound mutants by in situ hybridization. Corresponding to the microarray data, we found that the number of Myt1expressing cells was mildly reduced in Nkx6.1 mutants (Fig.  7E,G), while severely diminished in Nkx6.1 -/-;Nkx6.2 -/embryos (Fig. 7E,H). Such dose-dependent regulation of Myt1 expression by NKX6 activity was confirmed by quantitative RT-PCR, which showed a 1.5-fold and 3.3-fold reduction in Myt1 mRNA levels in Nkx6.1 -/and Nkx6.1 -/-;Nkx6.2 -/embryos, respectively (Fig. 7I). These findings suggest that NKX6 factors control αand β-cell differentiation by either directly or indirectly regulating the expression of the NGN3 co-factor MYT1.

Discussion
In this study, we demonstrate that the endocrine differentiation program in the mouse pancreas requires activity of NKX6 class proteins. Although a requirement for Nkx6.1 in β-cell differentiation has been previously shown (Sander et al., 2000), our present study provides evidence that redundant activity of the NKX6.1 paralog NKX6.2 masks a more general Development 132 (13) Research article   (Awatramani et al., 2000;Jorgensen et al., 1999;Mirmira et al., 2000), and function both as transcriptional repressors through interaction with Gro/TLE co-repressor proteins . Second, when transfected into the developing neural tube, NKX6.1 and NKX6.2 have qualitatively similar activities in inducing motoneurons (Vallstedt et al., 2001). A second potential mechanism that could account for the inability of NKX6.2 to compensate for NKX6.1 in β-cell differentiation is the difference in their spatial expression domains. In support of this idea, we found only NKX6.1, but not NKX6.2, to be expressed in NGN3 + endocrine progenitors and β-cells. It is therefore possible that normal development of the β-cell lineage requires sustained expression of NKX6 factors in endocrine progenitors and/or β-cells, while α-cell development requires only NKX6 activity in PDX1 + progenitors. Consistent with this view, ectopic expression of Ngn3 under control of the Pdx1 promoter leads to premature formation of αbut not βcells (Apelqvist et al., 1999;Schwitzgebel et al., 2000). Several observations indicate that compensation for NKX6.1 function by NKX6.2 occurs in PDX1 + pancreatic progenitors (Fig. 8). First, co-expression of NKX6.1 and NKX6.2 is only found in PDX1 + cells of the early pancreatic epithelium. Second, we show that derepression of NKX6.2 in Nkx6.1 mutants is limited to PDX1 + progenitors. Finally, sustained expression of β-gal in NGN3 + precursors of Nkx6.2 tlz/+ embryos suggests that NKX6.2/PDX1 co-positive cells are precursors for a subset of NGN3 + endocrine progenitors.

Regulation of Nkx6.2 by NKX6.1
The transcriptional repression of Nkx6.2 by NKX6.1 could either occur directly by binding of NKX6.1 to Nkx6.2 regulatory sequences or indirectly through regulation of additional genes. To test whether NKX6.1 could directly regulate Nkx6.2 transcription, we examined the mouse and human 5′ upstream sequences for conserved NKX6 consensus binding sites, but were unable to identify such elements within 10 kb 5′ of the Nkx6.2 transcription start site (K.D.H. and M.S., unpublished). Moreover, recent studies on Nkx6 gene function in the mouse hindbrain show that NKX6.1 and NKX6.2 are coexpressed in branchiomotor neurons and therefore do not have mutually exclusive expression domains (Muller et al., 2003). Together, these findings suggest that repression of Nkx6.2 by NKX6.1 may not involve direct transcriptional repression, but may be mediated through as yet unknown Nkx6 target genes.

NKX6 and MYT1 function in pancreatic endocrine development
Based upon the finding that Nkx6.1 mutants display a selective reduction in β-cells, it was suggested that NKX6.1 functions exclusively in the β-cell differentiation pathway (Sander et al., 2000). Our present results demonstrate a previously unrecognized requirement for NKX6 activity in α-cell formation, therefore suggesting a more general role for NKX6 factors in pancreatic endocrine development. A possible mechanism for how NKX6 activity may control αand β-cell differentiation is provided by our finding that NKX6 proteins regulate expression of the neurogenin (NGN) co-factor MYT1 in a dose-dependent manner. Both neurogenesis and pancreatic endocrine differentiation require the combined activities of MYT1 and NGN (Bellefroid et al., 1996;Gu et al., 2004). In the nervous system, NGN can induce the expression of Myt1, which suggests that they may function in a linear genetic pathway (Bellefroid et al., 1996). If regulation of these factors in the pancreas is similar to the nervous system, and NGN3 would be able to induce the expression of Myt1, it would explain why ectopic expression of Ngn3 alone is sufficient to induce pancreatic endocrine differentiation. However, from our results we can conclude that regulation of Myt1 by NKX6 factors is independent of NGN3, as Ngn3 is normally expressed in Nkx6 single and in Nkx6.1;Nkx6.2 double mutant mice. Thus, we propose that NGN3 is not sufficient to induce and/or maintain Myt1 in the pancreas, but that Myt1 expression requires the activity of NKX6 factors. Although we cannot exclude a cell-autonomous mechanism, our finding that NKX6.2 is expressed only in PDX1 + , but absent from NGN3 + progenitors, while Myt1 is enriched in Ngn3-expressing cells (Gu et al., 2004), suggests that NKX6 factors regulate Myt1 through a non-cell autonomous mechanism (Fig. 8).
If NKX6 activity is necessary for Myt1 expression, how can we reconcile the fact that Myt1 is still present in Nkx6.1;Nkx6.2 double mutant mice? One possible explanation is that NKX6.1 and NKX6.2 do not account for all NKX6 activity in the pancreas. The mouse genome indeed contains a third Nkx6 class gene that we found to be expressed in E10.5 and E14.5 pancreas by RT-PCR (K.D.H. and M.S., unpublished). This residual NKX6 activity could account for the low levels of Myt1 as well as for the small numbers of αand β-cells that still differentiate in the pancreas of Nkx6.1 -/-;Nkx6.2 -/embryos. It remains to be shown if deletion of all three Nkx6 genes in mice will result in a complete absence of Myt1 expression and a subsequent block of all endocrine cell differentiation, or whether, alternatively, NKX6 factors have a specific role in the development of just the αand β-cell lineages, while δand PP-cell differentiation are controlled by a NKX6-independent mechanism.
In contrast to Myt1, NGN3 expression appears to be independent of NKX6 activity. This is supported by our finding that NGN3 expression is normal in Nkx6 mutants until E15.5 and only slightly (~20%) reduced at E16.5. As the reduction in NGN3 cell numbers at E16.5 is small and observed to the same extent in Nkx6.1 -/and Nkx6.1 -/-;Nkx6.2 -/embryos, it is unlikely to account for the loss of glucagon + cells, which is only seen in Nkx6.1 -/-;Nkx6.2 -/embryos. This raises the issue of how absence of NKX6 activity affects the fate of NGN3 + cells. In our analyses, we did not find any evidence for persistence of NGN3 + cells at later developmental time points (data not shown) or a subsequent arrest in their endocrine differentiation. Moreover, our observation that the number of apoptotic, somatostatin-, PP-or ghrelin-producing cells was normal in Nkx6 mutants, argues against cell death or endocrine fate conversion as an underlying cause of αand β-cell loss. This leaves the possibility that absence of NKX6 activity leads to the differentiation of NGN3 + progenitors into non-endocrine cells; a hypothesis that we are currently testing. A factor (X) that mediates the signal between PDX1 + and NGN3 + progenitors remains to be identified. Myt1 in NGN3 + progenitors is required for their differentiation into αand β-cells.