Functional overlap between the mammalian Sar1a and Sar1b paralogs in vivo

Significance

SAR1A and SAR1B are components of the COPII coat that mediate protein transport from the endoplasmic reticulum to the Golgi. In humans, mutations in SAR1B cause chylomicron retention disease (CMRD), whereas mice with SAR1B deficiency exhibit embryonic lethality. No human condition or mouse phenotype has yet been associated with mutations in SAR1A. Here, we show that both Sar1a and Sar1b are required for murine survival. Expression of SAR1A in place of SAR1B appears to fully rescue the lethality of Sar1b deficiency in mice, demonstrating a high degree of functional overlap between these two paralogs. This suggests that transcriptional subfunctionalization is the primary evolutionary force maintaining both paralogs in mammals and therapeutic upregulation of SAR1A could be leveraged to treat CMRD.

Abstract

Proteins carrying a signal peptide and/or a transmembrane domain enter the intracellular secretory pathway at the endoplasmic reticulum (ER) and are transported to the Golgi apparatus via COPII vesicles or tubules. SAR1 initiates COPII coat assembly by recruiting other coat proteins to the ER membrane. Mammalian genomes encode two SAR1 paralogs, SAR1A and SAR1B. While these paralogs exhibit ~90% amino acid sequence identity, it is unknown whether they perform distinct or overlapping functions in vivo. We now report that genetic inactivation of Sar1a in mice results in lethality during midembryogenesis. We also confirm previous reports that complete deficiency of murine Sar1b results in perinatal lethality. In contrast, we demonstrate that deletion of Sar1b restricted to hepatocytes is compatible with survival, though resulting in hypocholesterolemia that can be rescued by adenovirus-mediated overexpression of either SAR1A or SAR1B. To further examine the in vivo function of these two paralogs, we genetically engineered mice with the Sar1a coding sequence replacing that of Sar1b at the endogenous Sar1b locus. Mice homozygous for this allele survive to adulthood and are phenotypically normal, demonstrating complete or near-complete overlap in function between the two SAR1 protein paralogs in mice. These data also suggest upregulation of SAR1A gene expression as a potential approach for the treatment of SAR1B deficiency (chylomicron retention disease) in humans.

The intracellular secretory pathway transports proteins from the endoplasmic reticulum (ER) to the Golgi for subsequent secretion into the extracellular space, insertion into the plasma membrane, or transport to various intracellular organelles including lysosomes and endosomes (1). Following proper folding in the ER, cargo proteins are transported to the Golgi apparatus via coat protein complex II (COPII) vesicles/tubules (2, 3). The COPII coat is a highly conserved structure consisting of five core proteins that are present in all eukaryotes: SAR1, SEC23, SEC24, SEC13, and SEC31 (4, 5). GDP-bound SAR1 is recruited to the ER exit site (ERES) by membrane-bound SEC12 to initiate coat assembly. SEC12 also functions as the guanine exchange factor for GDP-SAR1. GTP-bound SAR1 subsequently recruits the SEC23-SEC24 heterodimer complex to the ERES on the cytoplasmic face of the ER membrane via direct interaction between SAR1 and SEC23. SEC23 also acts as the guanine activating factor for SAR1, resulting in GTP hydrolysis. Following cargo recruitment and concentration in the ER lumen mediated by SEC24, SEC13 and SEC31 are recruited to form the outer coat and complete coat assembly (6).

Mammalian genomes contain multiple COPII paralogs as a result of gene duplication events, including two paralogs each for SAR1, SEC23, and SEC31 and four paralogs for SEC24 (6). Gene duplications are common evolutionary events contributing to genome expansion. One of the duplicated copies typically accumulates inactivating mutations over evolutionary time, transitioning to a pseudogene, which is eventually lost from the genome. Occasionally, one duplicate copy undergoes neofunctionalization (acquisition of new function(s) distinct from those of the ancestral gene) or subfunctionalization (division of the ancestral gene’s functions among the paralogs), resulting in maintenance of both copies in the genome. Subfunctionalization can occur at either the level of protein function or transcription (7). The expansion of COPII paralogs through the course of eukaryotic evolution is thought to have accommodated increasing secretory demand. Humans and mice deficient in one paralog typically exhibit distinct phenotypes from those deficient in the other (6, 8). In mice, loss of SEC23A results in lethality due to a neural tube closure defect, whereas SEC23B deficiency leads to perinatal death from pancreatic degeneration (9, 10). However, substitution of the Sec23a coding sequence at the Sec23b locus completely rescues the perinatal lethality exhibited by Sec23b null mice (11), demonstrating a high degree of functional overlap between these two paralogs. Mice with complete loss of SEC24A exhibit normal growth and fertility with marked hypocholesterolemia, whereas mice that lack either SEC24B, SEC24C, or SEC24D exhibit lethality at various time points during development (12–15). Both SEC24A/B and SEC24C/D exhibit at least a partial degree of functional overlap (12, 16).

The SAR1A and SAR1B paralogs differ by only 20 out of 198 amino acids in humans, and both appear to be expressed in a wide range of human and mouse tissues (Fig. 1) (17, 18). Mutations in SAR1B are associated with the rare autosomal recessive disorder chylomicron retention disease (CMRD, or Anderson’s disease) (19), whereas a disorder due to human mutations in SAR1A has not been reported. Melville and colleagues previously identified three amino acid clusters differing between human SAR1A and SAR1B that alter GTPase kinetics and SEC23 affinity (20), with a recent report further demonstrating that only human SAR1A is inhibited by the alarmone ppGpp (21). While these findings suggest unique biochemical functions for these two paralogs at the molecular level, there is also evidence for some overlap in function. Genetic deletion of either SAR1A or SAR1B in Caco-2/15 cells results in decreased chylomicron secretion with a less dramatic phenotype observed in SAR1A null cells. Combined inactivation of both paralogs in Caco2/15 cells is required to recapitulate the severe phenotype seen in cells from patients with CMRD (22, 23). Similarly, while disruption of SAR1A is tolerated in human retinal pigment epithelial (RPE-1) cells, additional attempts to delete SAR1B in SAR1A-null cells were unsuccessful, suggesting some degree of functional compensation between these two paralogs (24).

Fig. 1.

SAR1A and SAR1B are highly conserved proteins ubiquitously expressed but at variable levels across mouse and human tissues. (A) Protein sequence alignment of human and mouse SAR1A (NP_001136120.1 and NP_001345416.1) and SAR1B (NP_001028675.1 and NP_079811.1). Amino acid residues that differ between SAR1A and SAR1B are highlighted in gray, and residues that are different between mouse and human are colored in red. (B) Relative SAR1A and SAR1B mRNA expression in multiple human and mouse tissues. The x axis represents SAR1A/SAR1B mRNA abundance ratio in each tissue. Raw RNA sequencing data were obtained from refs. 17 and 18.

In mice, germline deletion of Sar1b results in late-gestational lethality in homozygous null fetuses, with haploinsufficient animals demonstrating a defect in chylomicron secretion similar to that observed in homozygous CMRD patients (25). Liver-specific deletion of Sar1b results in severe hypocholesterolemia due to a lipoprotein secretion defect (26). The consequences of Sar1a inactivation in mice is unknown. We now report analysis of mice engineered to be genetically deficient in Sar1a or Sar1b as well as mice carrying a modified Sar1b locus at which the Sar1a coding sequence has been substituted for Sar1b. Our results demonstrate a high degree of functional overlap between the SAR1A and SAR1B paralogous proteins.

Results

Germline Deletion of Sar1a in Mice Results in Midembryonic Lethality.

To investigate SAR1A function in vivo, we generated mice carrying a Sar1a targeted allele in which Sar1a expression is disrupted by insertion of a gene trap cassette into intron 5 (27) (Fig. 2A). We designed a three-primer PCR assay to differentiate between the wild-type and gene trap alleles (Fig. 2B), with DNA sequence analysis confirming the expected configuration. Quantitative PCR (qPCR) analysis of liver cDNA generated from Sar1a^+/+ and Sar1a^gt/+ littermates demonstrated an ~50% reduction of Sar1a transcripts in the latter, whereas Sar1b mRNA abundance remained unchanged, consistent with inactivation of a single Sar1a allele (Fig. 2C). To generate mice with homozygous loss of Sar1a, we performed an intercross between Sar1a^gt/+ mice, with genotyping results shown in Table 1. No Sar1a^gt/gt mice were identified among the 65 offspring genotyped at postnatal day 21 (P21). The results of genotyping for embryos harvested at embryonic days E10.5, E11, and E11.5 post coitus (pc) are shown in Table 1. Although the expected number of Sar1a^gt/gt embryos was observed at E10.5, no viable Sar1a^gt/gt embryos remained by E11.5. Histologic evaluation of E10.5 embryos revealed that Sar1a^gt/gt embryos were smaller than controls but failed to identify any other obvious abnormalities (SI Appendix, Fig. S1).

Fig. 2.

Haploinsufficiency of SAR1A is tolerated in mice. (A) Schematic of the wild-type Sar1a(+) and gene trap Sar1a(gt) alleles The gene trap (gt) cassette is inserted into intron 5 of the Sar1a gene. P1, P2, and P3 denote the binding sites of genotyping primers. Rectangles represent exons, and solid line segments represent introns. Exons and introns are not drawn to scale. (B) Genotyping assay using the primers denoted in A and genomic DNA isolated from embryonic yolk sacs. The Sar1a(+) allele produces a PCR amplicon of 489 bp, whereas the Sar1a(gt) allele produces an amplicon of 352 bp. (C) Liver Sar1a and Sar1b mRNA abundance in Sar1a^gt/+ and littermate controls determined by qPCR (normalized to the mean level of Sar1a^+/+ samples). (D and E) Body mass and plasma cholesterol levels of wild-type and heterozygous Sar1a^gt/+ mice. For panels C–E, statistical significance was determined by a two-sided t test.

Table 1.

Germline loss of SAR1A results in lethality during midembryogenesis

Mating genotype expected	Sar1agt/+ X Sar1agt/+			P-value (χ2 test)
	Sar1a+/+	Sar1agt/+	Sar1agt/gt
	25%	50%	25%
E10.5	10 (28%)	16 (44%)	10 (28%)	0.800
E11.0	5 (26%)	10 (53%)	4 (21%)	0.924
E11.5	7 (32%)	15 (68%)	0 (0%)*	0.025
Weaning (P21)	17 (26%)	48 (74%)	0 (0%)	7.22 × 10−6

^*Two Sar1a^gt/gt embryos appeared dead and were partially resorbed.

Heterozygous Sar1a^gt/+ mice were present at the expected Mendelian ratio at weaning (P21) and exhibit normal growth and fertility. At 8 wk of age, no significant differences in body mass or plasma cholesterol levels were observed between Sar1a^+/+ and Sar1a^gt/+ mice (Fig. 2 D and E).

Germline Homozygous Sar1b Deletion Results in Perinatal Lethality.

We generated heterozygous Sar1b-deficient mice (Sar1b^+/−) by crossing mice carrying a conditional Sar1b allele (Sar1b^fl/+) with mice carrying a ubiquitously expressed Cre (EIIa-Cre) transgene. Cre-mediated excision of Sar1b exon 5 (Fig. 3A) was confirmed with a three-primer PCR assay clearly differentiating the wild-type and null alleles (Fig. 3B). As shown in Fig. 3C, Sar1b mRNA level in livers collected from Sar1b^+/− mice was reduced to approximately 50% of that in Sar1b^+/+ littermates, with no difference in Sar1a transcript abundance between the two groups. The results of a heterozygous Sar1b^+/− intercross are shown in Table 2. Though no homozygous Sar1b^−/− mice were observed at weaning, the expected numbers of Sar1b^−/− embryos were present at E11.5 and E18.5, with most Sar1b^−/− neonates appearing to die shortly after birth (Table 2). Histologic analysis of E18.5 embryos failed to identify any obvious abnormalities in Sar1b^−/− embryos (SI Appendix, Fig. S2).

Fig. 3.

Mice with heterozygous loss of Sar1b demonstrate normal growth and plasma cholesterol levels. (A) Schematic of the wild-type Sar1b(+), condition gene trap Sar1b(cgt), conditional Sar1b(fl), and the null Sar1b(−) alleles. The gene trap (gt) cassette is inserted into intron 4 of the Sar1b gene. Orange circles represent the two FRT sites flanking the gene trap cassette. The Flp recombinase excises the gene trap cassette, resulting in the Sar1b(fl) allele. In the Sar1b(fl) allele, exon 5 of the Sar1b gene is flanked by two loxP sites (brown triangles), which undergo recombination in the presence of Cre recombinase, leading to excision of exon 5 in the Sar1b(−) allele. P4, P5, and P6 denote the binding sites of genotyping primers. Rectangles represent exons and solid line segments represent introns. Exons and introns are not drawn to scale. (B) Genotyping assay using primers denoted in A and genomic DNA isolated from tail clips. The Sar1b(+) allele generates a PCR product of 136 bp, whereas the Sar1b(−) allele produces a PCR amplicon of 169 bp. (C) Relative liver Sar1a and Sar1b mRNA abundance in Sar1b^+/− mice and littermate controls determined by qPCR (normalized to the mean level of Sar1b^+/+ samples). (D and E) Body mass and plasma cholesterol levels of wild-type and heterozygous Sar1b^+/− mice. For panels C–E, statistical significance was determined by a two-sided t test.

Table 2.

Genetic inactivation of Sar1b leads to perinatal lethality

Mating genotype expected	Sar1b+/− X Sar1b+/−			P-value (χ2 test)
	Sar1b+/+	Sar1b+/−	Sar1b−/−
	25%	50%	25%
E11.5	3 (19%)	8 (50%)	4 (31%)	0.779
E18.5	14 (26%)	24 (44%)	16 (30%)	0.665
P0	4 (31%)	7 (54%)	2 (15%)	0.707
P1	4 (57%)	3 (43%)	0 (0%)	0.095
Weaning (P21)	29 (35%)	54 (65%)	0 (0%)	9.21 × 10−7

Heterozygous Sar1b^+/− mice were observed at the expected numbers at weaning and exhibited normal growth and fertility (Table 2). At 3 to 4 mo of age, no differences in body mass or plasma cholesterol levels were observed between Sar1b^+/− and Sar1b^+/+ mice (Fig. 3 D and E).

Hypocholesterolemia Resulting from Hepatic SAR1B Deficiency Is Rescued by Overexpression of either SAR1A or SAR1B.

We next inactivated Sar1b in hepatocytes by combining an Alb-Cre transgene and the Sar1b^fl allele. Sar1b^fl/fl Alb-Cre⁺ mice were viable and exhibited a marked reduction in plasma cholesterol levels to approximately 20 mg/dL (compared to ~75 mg/dL in littermate controls, Fig. 4A). No reduction in cholesterol was observed in mice with hepatic haploinsufficiency for Sar1b (Sar1b^fl/+ Alb-Cre⁺, Fig. 4A), consistent with the normal plasma cholesterol level of Sar1b^+/− mice (Fig. 3E).

Fig. 4.

Hypocholesterolemia in hepatic SAR1B-deficient mice is rescued by overexpression of either SAR1A or SAR1B. (A) Plasma cholesterol levels in controls (Sar1b^+/+ Alb-Cre⁻, Sar1b^fl/fl Alb-Cre⁻, and Sar1b^+/+ Alb-Cre⁺—brown), hepatic Sar1b haploinsufficient (Sar1b^fl/+ Alb-Cre⁺—pink), and hepatic Sar1b null (Sar1b^fl/fl Alb-Cre⁺—red) mice. Statistical significance was determined by a pairwise two-sided t test followed by Bonferroni adjustment for multiple hypothesis testing. (B) Schematic of the SAR1B deficiency rescue experiment. Sar1b^fl/fl Alb-Cre⁺ mice were injected with adenovirus (AV) expressing GFP, mouse SAR1A, or mouse SAR1B via the tail vein. Blood was sampled on days 0, 3, and 7 postinjection and assayed for cholesterol levels. (C) Plasma cholesterol levels of mice injected with AV as illustrated in B. Two different AV doses were used (low and high—see Materials and Methods; n = 4 per dose per AV). Statistical significance was determined by a two-way ANOVA test with interaction between AV type, time, and dose followed by Tukey’s post hoc test. P-values for the labeled comparisons are listed in the table on the right; n.s., not significant. P-values for all comparisons are listed in SI Appendix, Table S1.

To examine potential overlap in function between SAR1A and SAR1B, we next tested the capacity of an adenovirus (AV) expressing either mouse SAR1A, mouse SAR1B, or GFP (as a control) to rescue the hypocholesterolemia of Sar1b^fl/fl Alb-Cre⁺ mice. AVs were injected via the tail vein into Sar1b^fl/fl Alb-Cre⁺ mice (Fig. 4B) and plasma cholesterol levels were determined on days 0, 3, and 7 postinjection. Injection of the SAR1B AV resulted in a rapid increase in plasma cholesterol to wild-type levels by day 3, with minimal increase observed with the control GFP AV, demonstrating the efficacy of the AV delivery and efficient expression of Sar1b in the targeted hepatocytes. Injection of the SAR1A AV resulted in a similar rescue of the hypocholesterolemia phenotype, indistinguishable from that observed with SAR1B AV (Fig. 4C). These data demonstrate that SAR1A can compensate for SAR1B deficiency in this assay, suggesting significant overlap in function between the SAR1 paralogs, at least in hepatocytes.

Expression of the Sar1a Coding Sequence at the Sar1b Endogenous Locus Rescues the Lethal Sar1b Null Phenotype in Mice.

To further examine the overlap in function between SAR1A and SAR1B in vivo, we generated a Sar1b^a allele in which the Sar1b coding sequence is replaced with that of Sar1a at the Sar1b endogenous locus, beginning immediately downstream of the ATG start codon in exon 2 of Sar1b (Fig. 5A). Mice homozygous for this allele should be unable to express the SAR1B protein, with SAR1A expressed under the transcriptional control of both the Sar1a and Sar1b endogenous regulatory elements. A three-primer PCR genotyping assay was developed that could clearly differentiate the wild-type and Sar1b^a alleles (Fig. 5B).

Fig. 5.

SAR1A expressed under control of the Sar1b endogenous locus can compensate for complete loss of SAR1B. (A) Schematic of the Sar1b(a) allele. Dashed lines mark the homology arms for DNA recombination. Exons and introns are not drawn to scale. CDS, coding sequence; WPRE, woodchuck hepatitis virus posttranscriptional regulatory element; pA, polyA sequence. P7, P8, and P9 denote binding sites for genotyping primers. (B) Genotyping assay using primers depicted in A with genomic DNA isolated from tail clips. The wild-type Sar1b(+) allele is expected to produce a PCR product of 358 bp (primer pair P7 and P9), whereas the Sar1b(a) allele generates a PCR amplicon of 268 bp (primer pair P7 and P8). (C) Schematic of the endogenous Sar1a, endogenous Sar1b, and Sar1b(a) transcript. The table on the right shows whether each transcript is present in the indicated genotype. Each number (with arrows on either side) below the transcript represents a particular primer pair used for detecting that transcript by qPCR. Primer pairs (1)–(3) are specific for each of the endogenous Sar1a, endogenous Sar1b, and Sar1b(a) transcript, respectively. Primer pair (4) amplifies both the endogenous Sar1a and Sar1b(a) transcripts (total SAR1A-coding). Primer pair (5) amplifies both the endogenous Sar1b and Sar1b(a) transcripts (Sar1b locus transcript). (D–H) Relative abundance for transcripts as determined by qPCR using the primers illustrated in C (normalized to the mean level of Sar1b^+/+ samples) in livers and small intestines collected from Sar1b^+/+, Sar1b^+/a, and Sar1b^a/a littermates. Statistical significance was determined by a one-way ANOVA test. (I) Body mass of male and female offspring from Sar1b^+/a X Sar1b^+/a intercrosses between 3 and 9 wk of age (n = 23, 35, and 6 for from Sar1b^+/+, Sar1b^+/a, and Sar1b^a/a, respectively; error bars represent the SD for each group). (J) Plasma cholesterol levels of Sar1b^+/+, Sar1b^+/a, and Sar1b^a/a mice at 3 to 4 mo of age. Statistical significance was determined by a one-way ANOVA test.

To confirm expression of SAR1A under control of Sar1b endogenous regulatory elements in Sar1b^a/a mice, we isolated RNA from the liver and small intestine of these mice and designed qPCR assays that could distinguish the endogenous Sar1a(+) and Sar1b(+) transcripts, the Sar1b(a) transcript, as well as primers that amplify both the Sar1a(+) and Sar1b(a) transcripts (total SAR1A-coding mRNA), or the Sar1b(+) and Sar1b(a) transcripts (total mRNA from the Sar1b locus) (Fig. 5C). No significant differences were observed in the abundance of endogenous Sar1a mRNA between Sar1b^+/+, Sar1b^+/a, Sar1b^a/a mice (Fig. 5D). However, Sar1b^+/a and Sar1b^a/a mice demonstrated a dose-dependent decrease in endogenous Sar1b and a corresponding increase in total SAR1A-coding transcript levels as compared to Sar1b^+/+ mice (Fig. 5 E and G). Although Sar1b^a transcripts were only detected in Sar1b^+/a and Sar1b^a/a (and not Sar1b^+/+) mice (Fig. 5F), Sar1b^a mRNA expression was reduced to ~60% of the wild-type Sar1b gene (Fig. 5H).

As shown in Table 3, intercrosses between Sar1b^+/a mice generated homozygous Sar1b^a/a mice that are present at weaning (though at a slightly lower percentage than the expected Mendelian ratio) and demonstrate normal growth (Fig. 5I) and fertility. Subsequent Sar1b^+/a X Sar1b^a/a crosses produced Sar1b^a/a offspring at the expected ratio (Table 3). Histologic examination of multiple organs did not identify any abnormalities in Sar1b^a/a mice. Sar1b^a/a mice also demonstrated plasma cholesterol levels indistinguishable from wild-type Sar1b^+/+ and heterozygous Sar1b^+/a mice (Fig. 5J).

Table 3.

Expression of SAR1A under the control of the Sar1b locus rescues the lethality in Sar1b null mice

Mating genotype expected	Sar1b+/a X Sar1b+/a			P-value (χ2 test)
	Sar1b+/+	Sar1b+/a	Sar1ba/a
	25%	50%	25%
Weaning	80 (30%)	148 (55%)	43 (15%)	0.002
Mating genotype expected	Sar1b+/a X Sar1ba/a
	Sar1b+/a		Sar1ba/a
	50%		50%
Weaning	23 (52%)		21 (48%)	0.763

Discussion

In this study, we investigate the in vivo function of the two closely related SAR1 paralogs in mice, demonstrating a high degree of overlap. Though both paralogs are required for survival to weaning, complete deficiency for SAR1A results in lethality at midembryogenesis, while Sar1b null mice die shortly after birth. Humans with homozygous or compound heterozygous loss-of-function mutations in SAR1A have not been reported to date, consistent with an early embryonic lethality phenotype similar to that seen in mice. The perinatal death of Sar1b null mice contrasts with the human phenotype, with homozygous SAR1B loss-of-function mutations in humans resulting in CMRD, characterized by plasma lipid abnormality and survival to adulthood. We did not identify any structural abnormalities in Sar1a^gt/gtembryos at E10.5 or Sar1b^−/− fetuses at E18.5. Embryonic lethality is frequently observed in mouse knockout models, with a previous systemic analysis of ~500 such models documenting death at or before birth in approximately 29%, with half of these dying before midembryogenesis (28). The death of Sar1a^gt/gt embryos at midembryogenesis could result from defective development of the placenta or cardiovascular system, as observed in other genetic models demonstrating similar embryonic loss (29). This midembryonic lethality without overt histologic abnormalities is similar to that observed in mice expressing a hypomorphic Sec24d allele (15), potentially consistent with a common mechanism related to impaired protein secretion. Though our analysis of both SAR1A and SAR1B haploinsufficient mice was consistent with normal growth and fertility, deeper phenotyping of Sar1b^+/− mice by Auclair et al. revealed that these mice do recapitulate some features of CMRD, including lowered plasma levels of triglycerides, decreased chylomicron secretion, and inefficient intestinal fat absorption (25). Thus, Sar1b^−/− pups could suffer from a more severe phenotype leading to perinatal lethality. Of note, the expected prevalence of heterozygous loss-of-function mutations in either SAR1A or SAR1B among a large cohort of human subjects in the gnomAD database (30), suggests that haploinsufficiency for these genes is well tolerated in humans.

Despite the high degree of amino acid sequence similarity between the two SAR1 paralogs, complete loss of each paralog in mice results in death at distinct developmental time points. Taken together with previously reported differences in GTP hydrolysis and affinity for SEC23B (20), these data suggest that SAR1A and SAR1B have diverged in protein function. However, we observed that overexpression of SAR1A in the liver rescued the hypocholesterolemia phenotype of mice with hepatocyte-specific SAR1B deficiency (Sar1b^fl/fl Alb-Cre⁺), demonstrating at least partial overlap in SAR1A and SAR1B function in hepatocytes, although confounding effects from SAR1A overexpression or an indirect effect of AV infection cannot be excluded. Of note, there was a small but significant increase in cholesterol levels in mice treated with the control AV (expressing GFP) in this experiment. To address these issues, and to examine the potential for paralog-specific function outside of hepatocytes, we generated mice expressing the SAR1A coding sequence under the transcriptional control of both its endogenous locus and the Sar1b locus. The complete rescue of the lethality caused by complete loss of SAR1B expression by the Sar1b^a allele demonstrates extensive, and potentially complete, overlap in function between the SAR1A and SAR1B proteins in vivo.

Three potential mechanisms have been proposed to explain the maintenance of paralogs such as SAR1A and SAR1B in the genome following gene duplication: neofunctionalization, subfunctionalization at the transcription level, or subfunctionalization at the protein level (7). Our findings suggest that subfunctionalization at the transcription level likely explains the maintenance of the two SAR1 paralogs in mammalian genomes. Expression of both paralogs is required, with loss of a single paralog sufficient to reduce total SAR1 below a critical level in select cell types, resulting in the unique patterns of lethality for SAR1A and SAR1B deficiency in mice and the difference in SAR1B-deficient phenotypes between mice and humans.

The phenotypic differences between humans and mice with SAR1B deficiency is reminiscent of that observed for humans and mice lacking SEC23B (10, 31). Divergence in gene expression patterns across tissues within the same species and across mammalian species could explain this phenomenon for both paralog pairs. The difference in the relative expression of SAR1A and SAR1B in mice and human small intestine, with SAR1B being the dominant paralog in mouse small intestine whereas SAR1A is expressed at a higher level than SAR1B in human small intestine (Fig. 1B), could explain the survival of humans carrying homozygous loss-of-function mutations in SAR1B while Sar1b^−/− mice exhibit perinatal lethality.

As noted above, human SAR1A and SAR1B have been shown to exhibit distinct biochemical properties in vitro (20, 21). Despite these differences at the molecular level, our data demonstrate that the corresponding mouse paralogs are sufficiently similar in function such that SAR1A can compensate for the loss of SAR1B in vivo. We cannot exclude greater similarity in function for the mouse versus human paralogs or subtle phenotypic differences between Sar1b^+/+ and Sar1b^a/a mice. Of note, we did observe a modest loss of Sar1b^a/a mice in the initial Sar1b^+/a X Sar1b^+/a intercross (Table 3). However, a subsequent Sar1b^+/a X Sar1b^a/a cross yielded the expected number of Sar1b^a/a offspring. These data could be the result of mosaicism in the founder mice, stochastic variation in gene expression between individual mice, minor differences in the genetic background of the founder mice and their progeny (C57BL/6N vs. C57BL/6J), or a passenger gene effect (32).

Finally, upregulation of a closely related gene paralog represents a potential strategy for the treatment of select genetic diseases. For example, upregulation of γ-globin expression has been demonstrated to provide dramatic clinical improvement in patients with sickle cell disease and β-thalassemia due to genetic abnormalities in its closely related paralog, β-globin (33, 34). Our data suggest that therapeutic upregulation of one SAR1 paralog could potentially rescue deficiency of the other. Indeed, CRISPR-mediated activation of SEC23A was shown to rescue the erythroid differentiation defect exhibit by a SEC23B-deficient erythroid cell line (35). A similar strategy could potentially be leveraged to treat patients with CMRD due to SAR1B deficiency through the activation of SAR1A.

Materials and Methods

Animal Care.

All animal care and use complied with the Principles of Laboratory and Animal Care established by the National Society for Medical Research. Mice were housed in a controlled lighting (12 h light/dark cycle) and temperature (22 °C) environment and had free access to food (5L0D, LabDiet) and water. All animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Michigan (protocol number PRO00011038). Both male and female mice were used in this study.

Generation of Sar1a^gt/+ Mice.

The Bay Genomics ES cell clone CSH949 (27) was obtained from the Mutant Mouse Resource & Research Centers (MMRRC). This clone is heterozygous for the Sar1a^gt allele in which a gene trap cassette is inserted into Sar1a gene intron 5 (Fig. 2A). ES cell culture, expansion, and microinjection to generate ES cell chimeric mice were performed as previously described (14, 15). Chimeric mice were bred with C57BL/6J (0006640, Jackson Laboratory) to obtain germ-line transmission. Mice carrying the Sar1a^gt allele were maintained by continuous backcrossing to C57BL/6J mice.

Generation of Sar1b^+/− Mice.

Mice carrying the Sar1b^{tm1a(EUCOMM)Wtsi} allele (36) were obtained from the European Conditional Mouse Mutagenesis (EUCOMM) program (Wellcome Trust Sanger Institute). These mice carry a conditional gene trap (Sar1b^cgt) allele (Fig. 3A) in which the gene trap cassette flanked by FRT sites followed by a loxP site is inserted into intron 4 of the Sar1b gene. An additional loxP site is inserted downstream of exon 5 of Sar1b. Mice carrying the Sar1b^cgt allele were maintained by continuous backcrossing to C57BL/6J mice.

To generate the conditional Sar1b allele (Sar1b^fl, Fig. 3A), Sar1b^cgt/+ mice were crossed with mice carrying the Flp1 recombinase gene under control of the Actb promoter (005703, Jackson Laboratory). Resulting mice carrying the Sar1b^fl allele were then bred with mice carrying an EIIa-Cre transgene (003724, Jackson Laboratory) to generate the Sar1b null allele (Sar1b⁻) (Fig. 3A). Sar1b^fl/+ and Sar1b^+/− mice were maintained by continuous backcrossing to C57BL/6J mice.

Generation of Sar1b^fl/+ Alb-Cre⁺ Mice.

To generate Sar1b^fl/+ Alb-Cre⁺ mice, the Sar1b^fl/+ mice generated above were crossed with mice carrying an Alb-Cre transgene (003574, Jackson Laboratory) (14). Sar1b^fl/+ Alb-Cre⁺ mice were maintained by continuous backcrossing to C57BL/6J mice. Sar1b^fl/fl Alb-Cre⁺ mice were obtained by crossing Sar1b^fl/+ Alb-Cre⁺ mice with Sar1b^fl/+ or Sar1b^fl/fl mice.

Generation of Sar1b^+/a Mice.

Mice carrying the Sar1b^a allele, in which the mouse Sar1a coding sequence is inserted into the Sar1b locus immediately downstream of the ATG codon in exon 2 of the gene, were generated by Biocytogen Boston Corp. using CRISPR/Cas9 (Fig. 5A). Cas9 and an sgRNA targeting exon 2 of Sar1b along with the DNA repair template were injected into C57BL/6N zygotes. Founder mice with confirmed germ-line transmission were bred to C57BL/6N mice to obtain F1 (Sar1b^+/a) mice. Sar1b^+/a mice were backcrossed to C57BL/6J mice for at least three generations before experimentation. Sar1b^+/a mice were maintained by continuous backcrossing to C57BL/6J mice.

Mouse Genotyping.

Tail clips were obtained from 2- to 3-wk-old mice for genomic DNA isolation and genotyping, as previously described (10). PCR was performed using Go-Taq Green Master Mix (Promega) and resulting products were resolved by 3% agarose gel electrophoresis. All primers used for genotyping are listed in SI Appendix, Table S2.

Timed Mating and Embryo Collection.

Timed matings were performed by intercrossing heterozygous mice (Sar1a^gt/+ or Sar1b^+/−). Embryos were collected at E10.5, E11.0, E11.5, or E18.5 for genotyping and histologic analyses. Genomic DNA was isolated from the embryonic yolk sacs of E10.5-E11.5 embryos or tail clips of E18.5 embryos and genotyped as described above. Embryos were immediately fixed in Z-fix solution (Anatech Ltd.) for subsequent histologic analyses.

Blood and Tissue Collection.

Blood and tissues were collected and sera separated as previously described (37). Tissues were immediately fixed in Z-fix solution or frozen in liquid nitrogen. Sera and tissues were stored at −80 °C until experimentation.

Histologic Analyses.

Tissue processing, embedding, sectioning, and hematoxylin and eosin (H&E) staining were performed at the University of Michigan In-Vivo Animal Core. Slides were reviewed by the investigators and a veterinarian pathologist blinded to genotype.

Cholesterol Assays.

Total cholesterol in sera was determined using a colorimetric assay (SB-1010-225, Fisher Scientific) according to the manufacturer’s instructions.

qPCR Assays.

Total RNA was isolated from tissues and converted to cDNA as previously described (37). Primers specific for each target (SI Appendix, Table S3) were designed using NCBI Primer-BLAST. qPCR were performed using Power SYBR Green PCR Master Mix (4367659, Invitrogen). Gapdh, B2m, Rpl37, and Rpl38 were used as housekeeping gene controls. Normalized transcript abundance was calculated by the 2^−ΔΔCt method using the gene with the lowest SD across all samples as the endogenous control.

Adenovirus-mediated Expression of SAR1A or SAR1B.

Recombinant adenoviruses expressing mouse SAR1A or SAR1B were constructed using pAdTrack-CMV and the AdEasy adenoviral vector system (Agilent Technologies). Adenoviruses were amplified in Ad293 cells and purified using CsCl gradient ultracentrifugation. Mice were transduced with Ad-GFP, Ad-SAR1A, or Ad-SAR1B via tail vein injection at 0.1 OD per mouse (low dose) or 0.3 OD per mouse (high dose), as described previously (38). Plasma cholesterol concentrations were measured prior to and 3 or 7 d after transduction as described above.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.