Pancreatic SEC23B deficiency is sufficient to explain the perinatal lethality of germline SEC23B deficiency in mice

In humans, loss of function mutations in SEC23B result in Congenital Dyserythropoietic Anemia type II (CDAII), a disease limited to defective erythroid development. Patients with two nonsense SEC23B mutations have not been reported, suggesting that complete SEC23B deficiency might be lethal. We previously reported that SEC23B-deficient mice die perinatally, exhibiting massive pancreatic degeneration and that mice with hematopoietic SEC23B deficiency do not exhibit CDAII. We now show that SEC23B deficiency restricted to the pancreas is sufficient to explain the lethality observed in mice with global SEC23B-deficiency. Immunohistochemical stains demonstrate an acinar cell defect but normal islet cells. Mammalian genomes contain two Sec23 paralogs, Sec23A and Sec23B. The encoded proteins share ~85% amino acid sequence identity. We generate mice with pancreatic SEC23A deficiency and demonstrate that these mice survive normally, exhibiting normal pancreatic weights and histology. Taken together, these data demonstrate that SEC23B but not SEC23A is essential for murine pancreatic development. We also demonstrate that two BAC transgenes spanning Sec23b rescue the lethality of mice homozygous for a Sec23b gene trap allele, excluding a passenger gene mutation as the cause of the pancreatic lethality, and indicating that the regulatory elements critical for Sec23b pancreatic function reside within the BAC transgenes.

and sutural cataracts 17 , while multiple loss of function SEC23B mutations have been identified in patients with congenital dyserythropoietic anemia type II (CDAII) 18 . CDAII is an autosomal recessive disease characterized by moderate anemia, increased bi/multi-nucleated erythroblasts in the bone marrow (BM), a double red blood cell (RBC) membrane by electron microscopy, and a faster than normal migration of the RBC membrane protein band 3 on SDS-PAGE 1,[19][20][21][22] . No other non-hematologic abnormalities have been reported to result from SEC23B mutations, though a recent report suggested that germline heterozygous SEC23B variants are associated with Cowden syndrome and apparently sporadic thyroid cancer 23 . Patients with two nonsense SEC23B mutations have never been reported 1,24 , raising the possibility that complete SEC23B deficiency might be lethal.
We previously reported that mice homozygous for a Sec23b gene-trap allele (Sec23b gt/gt ) die perinatally, with analysis at embryonic day 18.5 (E18.5) demonstrating degeneration of the pancreas and other professional secretory tissues 25 . Lethally irradiated mice transplanted with hematopoietic stem cells (HSC) deficient for SEC23B did not exhibit anemia or other CDAII characteristics 24 , and Sec23b gt/gt HSC exhibited no competitive disadvantage at reconstituting the BM erythroid lineage. We also recently reported that mice homozygous for a Sec23a gene trap allele die at mid-embryogenesis, exhibiting a neural tube defect and impaired collagen secretion, reminiscent of the human phenotype 26 .
We now confirm and extend our previous findings, demonstrating that mice with erythroid specific (Epo-R Cre) and pan-hematopoietic (Vav1-Cre) Sec23b deletion support a normal erythroid and hematopoietic compartment. We also show that two BAC transgenes comprising Sec23b rescue the phenotype of Sec23b gt/gt mice, ruling out a passenger gene mutation and indicating that key Sec23b regulatory elements reside within these BAC transgenes. Finally, we demonstrate that loss of SEC23B expression exclusively in the pancreas is sufficient to explain the lethality of mice with germline deletion of Sec23b, and show that pancreatic SEC23A, unlike SEC23B, is dispensable for normal murine pancreatic development.

Results
The key regulatory sequences required for Sec23b pancreatic function reside within a 127 kb spanning the Sec23b gene. Mice homozygous for a Sec23b gene-trap allele (Fig. 1A) were previously reported to exhibit massive pancreatic degeneration with uniform perinatal lethality 25 . In contrast, pancreatic abnormalities have not been observed in SEC23B deficient humans 1 . Two independent mouse BAC transgenes (Fig. 1B) spanning the Sec23b gene were crossed onto the Sec23b gt line. Both Sec23b BAC transgenes demonstrated full rescue of the Sec23b gt/gt lethality, with the expected numbers of Sec23b gt/gt Tg + observed at weaning (Table 1). Sec23b gt/gt Tg + mice exhibited normal survival (observed for 300+ days), and fertility, with no apparent pancreatic or other abnormalities on routine necropsy ( Supplementary Fig. 1). BAC transgene expression in the relevant tissues is inferred from the dramatic phenotype rescue. These data demonstrate that the key regulatory elements necessary for Sec23b pancreatic function are located within the minimum ~127 kb region shared by the two BAC transgenes (Fig. 1B).
Loss of pancreatic SEC23B expression is sufficient to account for the perinatal mortality observed in Sec23b −/− mice. Pancreas-specific Sec23b deficient mice were generated, using the Pdx-Cre transgene. Out of 74 progeny generated from a Sec23b +/fl Pdx-Cre + × Sec23b +/− cross, only 2 Sec23b −/fl Pdx-Cre + mice were observed at weaning (compared to 9/74 mice expected, p < 0.016, Table 2). Analysis of pancreas tissues from these 2 mice detected a high level of residual functional non-excised Sec23b fl alleles ( Supplementary  Fig. 3), likely explaining their extended survival, either due to incomplete excision of Sec23b and/or selection of non-excised cells during pancreatic development.
Taken together, these data demonstrate that loss of SEC23B exclusively in the pancreas is sufficient to explain the perinatal mortality observed in mice with germline deletion of Sec23b.
Pancreatic SEC23B deficiency results in loss of pancreatic acini. Histologic evaluation of pancreatic tissues harvested from E18.5 Sec23b −/fl p48-Cre + embryos demonstrated hypoplastic pancreatic remnants with degeneration of pancreatic acini ( Fig. 2A, supplementary Table 2, p < 0.0001). Acinar cells were smaller than normal, with scant to minimal eosinophilic cytoplasm that was often vacuolated, and separated by clear space and prominent interlobular stroma. Though pancreatic islets could not be identified by H&E staining due to parenchymal collapse, immunostaining for insulin or glucagon demonstrated normal islet cell morphology, while immunohistochemistry for amylase confirmed loss of acinar cells (Fig. 2B).
Mice with pancreas-specific SEC23A deficiency survive to adulthood and lack a pancreatic phenotype. Mice homozygous for a Sec23a gene trap allele die at mid embryogenesis exhibiting neural tube opening and abnormal development of extra-embryonic membranes 26 . This early lethality precluded the evaluation of the effect of SEC23A deficiency on pancreatic development. To assess the impact of pancreatic SEC23A deficiency on pancreatic development, we generated mice heterozygous for a conditional Sec23a floxed allele (Sec23a +/fl ) and for a Sec23a null allele (Sec23a +/− ) ( Fig. 1D-G). Mice with pancreatic deficiency of SEC23A were generated by crossing either Sec23a +/fl p48-Cre(+ ) or Sec23a +/− p48-Cre(+ ) mice to Sec23a fl/fl mice. These crosses yielded the expected number of Sec23a fl/fl p48-Cre(+ ) and Sec23a −/fl p48-Cre(+ ) offspring (Table 3). Sec23a fl/fl p48-Cre + and Sec23a −/fl p48-Cre + mice exhibited normal survival (mice observed for > 300 days of age), development, and fertility. SEC23A protein was undetectable by western blotting in pancreas tissues harvested from Sec23a −/fl p48-Cre + mice (Fig. 3A), consistent with complete excision of the Sec23a fl allele by the p48Cre-transgene. Pancreas tissues dissected from Sec23a −/fl p48-Cre + mice exhibited normal weights (Fig. 3B) and were histologically indistinguishable from WT controls (Fig. 3C). SEC23B protein demonstrated mildly increased steady state levels in SEC23A-deficient pancreata (Fig. 3D), while SEC23B mRNA was not increased (Fig. 3E). These results demonstrate that SEC23A, in contrast to SEC23B, is dispensable for normal pancreatic development and function.
No RBC abnormalities observed in mice with erythroid-specific and hematopoietic specific SEC23B deficiency. Mice with erythroid-specific or pan-hematopoietic SEC23B-deficiency were generated by crossing the Sec23b fl allele to mice expressing Cre-recombinase driven by the EpoR promoter or the Vav1 promoter, respectively. Sec23b −/fl EpoR-Cre(+ ) and Sec23b −/fl Vav1-Cre(+ ) mice were observed in the expected Mendelian ratios at weaning (Table 2), appeared grossly normal, and exhibited normal survival and fertility, as well as normal RBCs, with none of the features of human CDAII. Specifically, these mice exhibited no anemia The Sec23a wild type allele is referred to as Sec23a + . The Sec23a conditional gene trap (Sec23a cgt ) allele was crossed to a FLP mouse, resulting in the Sec23a floxed allele (Sec23a fl ). Subsequent Cre mediated excision gives rise to the Sec23a null allele (Sec23a − ). The locations of genotyping primers are depicted by arrows. (E) Competitive PCR assay with 1 forward primer (primer A) and 2 reverse primers (primers B and B4) yields a 179 base pair (bp) product from the wild type allele (primers A and B4), and a 260-bp product from the Sec23a cgt allele (primers A and B). (F) PCR assay using primers A and E2 results in a 231 bp product from the wild type allele and a 335 bp product from the Sec23b fl allele. (G) PCR assay using primers A and D results in a 482 bp product from the Sec23b − allele. No PCR product is generated with these primers from a WT allele due to the large distance between the primers.

Murine hematopoietic SEC23B deficiency does not result in a block of B-lymphocyte development.
In a previous report 24 (Fig. 4F,G), thereby arguing against a major defect in B-lymphocyte development resulting from SEC23B deficiency.

Discussion
In humans, homozygous or compound heterozygous loss of function mutations in SEC23B result in a phenotype limited to the erythroid compartment, with no other non-hematologic abnormalities reported. In contrast, mice homozygous for a Sec23b genetrap allele die perinatally exhibiting extensive pancreatic degeneration, and mice with hematopoietic deficiency for SEC23B exhibit no RBC phenotype.
In this report, we confirm the perinatal mortality and pancreatic phenotype of SEC23B deficient mice using an independent Sec23b null allele (Sec23b − ). We also show that two independent BAC transgenes spanning Sec23b rescue the lethality and pancreatic phenotype of Sec23b gt/gt mice, thereby confirming that the latter phenotype observed in these mice is indeed a result of SEC23B deficiency and definitively excluding an off-target genetrap effect on a nearby gene (passenger gene mutation) segregating with the Sec23bgt allele as a cause of the phenotype. We also show that mice with either erythroid specific or pan-hematopoietic Sec23b deletion do not exhibit a CDAII phenotype. These data are consistent with previously reported HSC transplantation results 24 , and exclude an artifact from HSC transplantation in our prior report as the cause of the lack of RBC phenotype in mice transplanted with SEC23B deficient HSCs 24 . Furthermore, we demonstrate that the stages of B-lymphocyte development in the BMs of mice transplanted with SEC23B deficient HSCs are indistinguishable from those in BMs of control mice transplanted with wild type HSCs, thereby ruling out a defect in B-lymphocyte development resulting from SEC23B deficiency, as previously suggested 24 .
We also show that mice with pancreatic SEC23A deficiency survive normally and are indistinguishable from their wild type littermate controls, exhibiting normal pancreas weight and histology. These data demonstrate that SEC23B, but not SEC23A, is essential for murine pancreatic development. SEC23B protein but not SEC23B mRNA was increased in SEC23A-deficient pancreata. These data suggest that relative SEC23A and SEC23B protein levels may be regulated in part by competition for SEC24 subunits 24 , with increased stability of the SEC23 subunit within a SEC23-SEC24 heterodimer.
We show that deletion of Sec23b exclusively in the pancreas is sufficient to account for the lethality of mice with germline deficiency for SEC23B. These data exclude a previously unidentified pathology, ectopic to the pancreas, as a cause of the pancreatic phenotype and suggest that the corresponding pancreatic destruction is due to a cell-autonomous defect in the pancreatic cell itself. Immunohistochemical stains demonstrate morphologically abnormal acinar cells but normal islet cells. These data suggest that the observed pancreatic degeneration is a result of an acinar cell defect, potentially due to delayed ER exit of a specific secretory cargo(s), possibly one or more exocrine pancreatic enzymes, which when retained in the ER results in pancreatic degeneration.
The rescue of the SEC23B deficiency phenotype by either of two BAC transgenes demonstrates that the regulatory elements critical for physiologic pancreatic expression of Sec23b reside within the minimum region shared by the BAC transgenes. These findings have important implications for future work aiming at defining therapeutic approaches to modifying the expression of the SEC23 paralogs in CDAII and other SEC23 disorders 17,18,23 .

Materials and Methods
Generation of Sec23b mutant mouse lines. Two independent Sec23b mutant mouse lines, one with a gene trap insertion in intron 19 of the gene (Sec23b gt ), and another line with a conditional gene trap (flanked by FRT sites) insertion in intron 4 of Sec23b (Sec23b cgt ) were previously described 22,23 (Fig. 1A,C). Mice expressing FLP recombinase driven by the human β-actin promoter (β-actin FLP) (Jackson laboratory stock # 005703) were crossed to the Sec23b cgt allele to excise the gene trap cassette, resulting in the Sec23b floxed allele (Sec23b fl ), with exons 5 and 6 flanked by loxP sites. The Sec23b fl allele was crossed to mice ubiquitously expressing Cre recombinase under the control of an EIIa promoter (EIIa-Cre) (Jackson laboratory stock # 003724) to excise exons 5 and 6 and generate a null Sec23b allele (Sec23b − ), resulting in a frameshift and early stop codon in exon 7 (Fig. 1C). Sec23b +/− mice were back-crossed with C57BL/6 J mice to remove the EIIA-Cre transgene. Mice with a pancreas-specific SEC23B knock-out were generated by crossing Sec23b fl/fl or Sec23b −/fl mice to mice expressing Cre recombinase driven by either the p48 promoter (generous gift from Dr. Christopher Wright) 27 or the Pdx1 promoter 28 . Mice with erythroid-specific and pan-hematopoietic SEC23B-deficiency were generated by crossing the Sec23b fl allele to EpoR-Cre mice 29 (generous gift from Dr. Ursula Klingmüller) and Vav1-Cre mice (Jackson laboratory stock # 008610), respectively. Sec23b gt mice used in this study were backcrossed to C57BL/6 J mice for > 10 generations. The Sec23b cgt allele was derived from a C57BL/6 ES cell, and the Sec23b fl and Sec23b − alleles were maintained on a pure C57BL/6 J background. All Cre mice were back-crossed to C57BL/6 J mice for > 6 generations. Mice were housed at the Life Sciences Institute of the University of Michigan, and all experiments were approved by and performed in accordance with the regulations of the University Committee on Use and Care of Animals.

Generation of BAC transgenic mice. Two bacterial artificial chromosome (BAC) clones spanning the
Sec23b gene, RP23-70J9 (RP23) and RP24-371A4 (RP24), were purchased from the BACPAC Resources Center at Children's Hospital Oakland Research Institute. BAC DNA constructs were expanded in One Shot TOP10 Escherichia coli and purified using the NucleoBond BAC100 kit (Machery-Nagel). BAC DNA was injected into zygotes generated from a cross between C57BL/6JxSJL F1 females and Sec23b +/+ male mice. RP23 and RP24 transgenic founders (Sec23b +/+ Tg + ) were crossed to Sec23b +/gt mice, and the Sec23b +/gt Tg + progeny were crossed to Sec23b +/gt C57BL/6 J mice to generate potential Sec23b gt/gt Tg + "rescue mice". Mice were genotyped for the Sec23b allele and for the presence of the BAC transgene. Standard genotyping methods are unable to differentiate between the endogenous Sec23b allele and the Sec23b gene present on the BAC transgene. Therefore, microsatellite genotyping was used (see below) to distinguish Sec23b +/gt Tg + mice from Sec23b gt/gt Tg + mice.
PCR genotyping. DNA was isolated from mouse tail biopsies and genotyping was performed using the Go-Taq Green Master Mix (Promega). Genotyping for the Sec23b cgt , Sec23b fl , and Sec23b − alleles and for the various Cre transgenes was performed as previously described 24,27,28 . Location of the Sec23a genotyping primers is shown in Fig. 2A, and their sequences are shown in Supplementary Table 1. The Sec23a cgt allele was genotyped in a three-primer PCR assay using a forward primer (primer A) located in Sec23a intron 2, upstream of the gene trap insertion site and two reverse primers, one (primer B) located in the gene trap insertion cassette between the two FRT sites and the second (primer B4) located in intron 2 downstream of primer A (the genomic sequence to which primer B4 anneals is absent from Sec23a cgt ). This PCR product was resolved by 2% (weight/ volume) agarose gel electrophoresis (Fig. 1E). Genotyping for the Sec23a fl allele was performed with a PCR assay consisting of the forward primer A and a reverse primer located in intron 2 between the two LoxP sites (primer E2) (Fig. 1F). This reaction does not yield a PCR product for the Sec23a − allele. Confirmation of the excision of    exon 3 (Sec23a − allele) was performed using a PCR assay with primer A and a reverse primer located in intron 3 downstream of the LoxP site (primer D) (Fig. 1G).

Microsatellite genotyping.
To distinguish Sec23b gt/gt Tg + from Sec23b +/gt Tg + mice, a microsatellite genotyping assay was designed that differentiates the endogenous Sec23b wild type allele from the Sec23b Tg + allele originally targeted on the 129/SvImJ background. The gene trap is expected to be 129/SvImJ within the congenic interval, in contrast to the wild-type allele, which should be either C57BL/6 J or SJL/J (SJL/J was introduced with the transgenic founders). A similar microsatellite genotyping strategy was previously described 30  Fetal Liver cells transplants. Timed matings were performed by intercrossing Sec23b +/− mice as previously described 24 . Pregnant females were euthanized at E16.5 postcoitus, and fetal liver cells were isolated and transplanted into lethally irradiated recipient mice as previously described 24 .
Complete blood counts (CBC) and bone marrow (BM) analysis. Blood (20 or 70 microliters) was collected from the retro-orbital venous sinuses of anesthetized mice via anticoagulant-coated capillary tubes. Blood was diluted in 5% bovine serum albumin in phosphate buffered saline (pH 7.4) to a total volume of 200 microliters. CBCs were determined as previously described 24 . BM cells were collected from hind limbs of euthanized mice, cytospins prepared and stained as previously described 24 , and the latter examined under light microscopy by an investigator blinded to the BM genotype.
RBC ghosts were prepared from peripheral blood RBC and stored at − 80 °C in lysis buffer, as previously described 24 .
Western blot and qRT-PCR. Total cell lysates were prepared as previously described 24 . Western blots (film visualization with chemiluminescent detection) and quantitative western blot (Infrared fluorescent detection) were performed as previously described 22 . For quantitative western blots, band intensities were quantified using the Image Studio software (LI-COR Biosciences) and normalized to GAPDH, and the secondary antibodies utilized were IRDye 680RD or IRDye 800CW. qRT PCR was performed as previously described 24 . Antibodies. Rabbit Anti-mouse SEC23B and anti-mouse SEC23A antibodies were generated as previously described 24,25 . Mouse anti-GAPDH and anti-Band3 antibodies were purchased from Millipore. Anti-Actin antibody was purchased from Santa Cruz.
Hematoxylin and eosin staining and Immunohistochemistry. At necropsy, tissues were collected and fixed in aqueous buffered zinc formalin (Z-fix, Anatech) for histologic and immunohistochemical analysis. Tissues were routinely processed, embedded in paraffin, sectioned at 4 um, and stained with hematoxylin and eosin (H&E). For immunohistochemistry, rabbit polyclonal antibodies to pancreatic amylase (ab21156, 1:1000; Abcam, Cambridge MA), insulin (C27C9, 1:800; Cell Signaling Technology, Danvers MA), and glucagon (D16G10, 1:100; Cell Signaling Technology, Danvers MA) were used. Following antigen retrieval, quenching of endogenous peroxidases and rodent block, primary antibodies were applied. After primary antibody incubation and washing, rabbit polymer HRP secondary antibody (Biocare, Concord CA), was applied. Negative controls were obtained by substitution of the primary antibody with Universal Negative reagent (Biocare, Concord CA). Following washing, 3,3-diaminobenzidine (DAB) was applied to visualize all reactions. Sections were counterstained with hematoxylin, dehydrated through graded alcohols, immersed in xylene, and mounted with coverslips. Histologic evaluation was performed by an investigator blinded to the genotype of the evaluated mice.
Statistical analysis. The Chi-square test was used to calculate the statistical significance of the deviation of the genotypes of a given cross from the expected Mendelian ratios. The statistical significant differences between blood count parameters in test cohorts and controls were calculated using the student's T-test.