Analysis of MCFD2- and LMAN1-deficient mice demonstrates distinct functions in vivo

Coagulation factor V (FV) and FVIII are large plasma glycoproteins that share similar domain structures and play pivotal roles in hemostasis and thrombosis.¹  Activated FV and FVIII serve as nonenzymatic cofactors for FXa and FIXa, respectively, with mutations in FV resulting in inherited FV deficiency and mutations in FVIII resulting in hemophilia A. Combined deficiency of FV and FVIII (F5F8D) is an autosomal recessive bleeding disorder that is characterized by simultaneous reduction of both cofactors^2-3  resulting from loss-of-function mutations in LMAN1 or MCFD2.^4-5  LMAN1 is a transmembrane lectin that contains a carbohydrate binding domain localized to the endoplasmic reticulum (ER) lumen, as well as cytoplasmic motifs for ER exit and retrieval.^6-8  MCFD2 is a small soluble protein with 2 EF-hand domains at the C terminus and an unstructured sequence at the N terminus.^5,9-10  The carbohydrate-recognition domain of LMAN1 and the EF-hand domains of MCFD2 interact to form a Ca²⁺-dependent protein complex that cycles between the ER and the ER–Golgi intermediate compartment (ERGIC), an organelle that is unique to higher eukaryotic cells.

The LMAN1–MCFD2 complex is thought to function as a cargo receptor required for efficient secretion of FV and FVIII.^5,11-12  Cargo receptors capture and concentrate specific cargo proteins into COPII-coated vesicles in the ER.^13-14  After budding, these vesicles fuse to form the ERGIC, where cargo proteins are separated from the cargo receptor and continue their journey through the secretory pathway. The dissociation of FV/FVIII from the LMAN1–MCFD2 complex may be triggered by a drop in Ca²⁺ concentration in the ERGIC.¹⁵  Cargo receptors are recycled to the ER via COPI-coated vesicles for additional rounds of cargo transport. LMAN1 and MCFD2 interact with FV/FVIII, suggesting distinct and essential roles for both subunits of the protein complex in FV/FVIII transport.^10-12  LMAN1–MCFD2 is the first known example of a specific ER-to-Golgi cargo receptor in vertebrates, with the requirement for a transmembrane component (LMAN1) and a soluble cofactor (MCFD2) suggesting a more complex trafficking mechanism than previously characterized cargo receptors in yeast.

Residual FV and FVIII secretion in subjects with LMAN1 or MCFD2 mutations results in plasma levels of 5-30% compared with normal controls, although the range of levels observed in MCFD2-deficient subjects is slightly lower than in subjects with LMAN1 mutations.¹⁶  Previously, we reported that LMAN1-deficient mice exhibit a strain-specific partial lethal phenotype, decreased levels of plasma FV and FVIII, and decreased platelet FV, although both are more moderately reduced than in humans.^17-18  Hepatocytes of LMAN1-deficient mice also accumulate α1-antitrypsin (AAT), an inhibitor of neutrophil elastase and an acute-phase reactant¹⁹  previously identified as a potential cargo of LMAN1.²⁰  Studies of conditional LMAN1-knockout (KO) mice demonstrate that FVIII is exclusively synthesized in endothelial cells.¹⁸  We now report the characterization of gene-targeted mice deficient in MCFD2, demonstrating a further reduction in plasma FV and FVIII levels compared with LMAN1-deficient mice as in humans, as well as a defect in AAT secretion, with doubly deficient mice exhibiting higher FV and FVIII levels comparable to LMAN1-deficient mice.

We designed a targeting vector in pCKO²¹  that would replace exons 2 and 3 of the murine Mcfd2 gene with a neomycin expression cassette. This vector was electroporated into 129/svImJ mouse embryonic stem (ES) cells. ES cell clones were initially screened using a TaqMan-based polymerase chain reaction (PCR) method,²²  which identified 1 clone with homologous recombination that results in reduction to only a single copy of the wild-type (WT) Mcfd2 allele. After further confirmation by Southern blot analysis (Figure 1), ES cells were expanded and injected into blastocysts derived from C57BL/6J mice by the Transgenic Mouse Core at the University of Michigan, which were then implanted into pseudopregnant female mice. The resulting male chimeric offspring were mated with female C57BL/6J or 129/svImJ mice to generate F1 mice with germline transmission. MCFD2-deficient mice were also crossed with previously reported LMAN1-deficient mice with a gene-trap allele (gt1)¹⁷  to generate offspring with double deficiencies. All mice had free access to food and water and were kept in cages in a 12-hour light/dark cycle. All animal experimental protocols were approved by the Institutional Animal Care and Use Committees at the University of Michigan and the Cleveland Clinic.

Genotyping was carried out by a 3-primer PCR assay of genomic DNA prepared from tail clippings of pups. Primers flanking the 3′ end of the left arm within intron 1 of Mcfd2 (L1, AGGAGAGCAAGCCAATTCTG; L2, TGAAAACCCCTCACTCTGCT) were combined with a primer located in the vector sequence inserted into the genome (V1, CTAAAGCGCATGCTCCAGAC) to generate different-sized PCR products from the WT (201 bp) and the targeted allele (248 bp) (Figure 1).

Mouse plasma was collected as previously described.¹⁷  We developed a murine FV-specific enzyme-linked immunosorbent assay (ELISA) using 2 monoclonal anti-murine FV antibodies from Green Mountain Antibodies (Burlington, VT). Clone GMA-752 was used as a capture antibody, and biotinylated clone GMA-754 was used as a detection antibody. Previously tested antibodies were unable to detect intracellular FV in mouse liver.^17,23-24  Clone GMA-753 was found to detect intracellular FV in mouse liver extracts by western blot analysis. FVIII activity was measured using a chromogenic assay kit (Chromagenix, Milano, Italy) that detects human and murine FVIII.²⁵  Pooled WT mouse plasma was used as a standard. Measurements of FV and FVIII were carried out on 3 different days with 3 different batches of plasma samples, with each batch containing all 4 genotypes (WT, LMAN1-deficient, MCFD2-deficient, and doubly deficient mice).

Mouse plasma AAT levels were measured using an ELISA kit (ICL Laboratory, Portland, OR). Western blot analysis of intracellular and secreted AAT was performed using a rabbit anti-AAT (Antibodies-online, Atlanta, GA).

WT, Lman1^gt1/gt1, Mcfd2^−/−, and Lman1^gt1/gt1Mcfd2^−/− mouse liver sections were embedded in Tissue-Tek optimal cutting temperature compound (Sakura Finetek, Torrance, CA), as previously described.^26-27  Cryosections (6 µm) were permeabilized with 0.1% Triton X-100 and incubated with chicken anti-mouse AAT (ICL Laboratory), followed by an Alexa Fluor 488–conjugated secondary antibody (Thermo Fisher, Waltham, MA). Cover slips were mounted using mounting medium (Vector Laboratories, Burlingame, CA), and images were captured using an Olympus FluoView confocal microscope.

A 2-tailed Student t test was used to assess the significance of differences between 2 groups of data. The χ² test was used to evaluate the significance of differences between the expected and observed genotype distributions. To investigate whether doubly deficient mice were unrepresented, we used logistic regression to compute the odds ratio (OR) and confidence intervals (CIs) between observations and expectations.

The targeting construct was designed to delete exons 2 and 3 of Mcfd2 (Figure 1A), which contain the majority of the coding sequence. Southern blot analysis confirmed correct integration of the targeting vector into the Mcfd2 locus of the ES cell clone (Figure 1B). Mating of male chimeric mice with female C57BL/6J mice generated pups with germline transmission of the Mcfd2 deletion allele that should be F1 for the C57BL/6J and 129/svImJ genetic background, whereas mating of male chimeric mice to female 129/svImJ mice generated pups with germline transmission of the Mcfd2 deletion allele on the pure 129/svImJ background. Intercrosses of F1 mice generated homozygote F2 mice (Figure 1C). Western blot analysis confirmed the absence of MCFD2 protein expression in Mcfd2^−/− mice (Figure 1D). F1 C57BL/6J/129/svImJ Mcfd2^+/− mice were further backcrossed onto the C57BL/6J background for >10 generations.

We performed intercrosses of Mcfd2^+/− mice backcrossed onto the C57BL/6 background for >10 generations, as well as Mcfd2^+/− mice on the 129/svImJ background and the mixed C57BL/6 and 129/svImJ genetic background (F1 and F2 mice). The expected Mendelian ratio of genotypes was observed in the offspring of all intercrosses at weaning (Table 1), indicating that MCFD2 is not required for murine development or survival. No spontaneous bleeding was observed in Mcfd2^−/− mice, and bleeding time from tail vein transection^28-29  was indistinguishable between Mcfd2^−/− mice (62 ± 24 seconds, n = 14) and WT controls (57 ± 17 seconds, n = 11) on the C57BL/6 background, consistent with a previous report using Lman1^−/− mice.¹⁷  Previously reported LMAN1-deficient mice with a gene-trap allele (gt1)¹⁷  that had been backcrossed with C57BL/6J mice for >16 generations were mated with Mcfd2^+/− mice on the 129/svImJ background to produce double-heterozygous mice on a mixed C57BL/6 and 129/svImJ genetic background or with Mcfd2^+/− mice backcrossed with C57BL/6J mice for 10 generations to produce double-heterozygous mice on the C57BL/6J background. To assess the viability of LMAN1/MCFD2 double-KO (DKO) mice, we first performed intercrosses of Lman1^+/gt1Mcfd2^+/− mice on a mixed C57BL/6 and 129/svImJ genetic background (Table 2). Western blot analysis confirmed the absence of LMAN1 and MCFD2 in Lman1^gt1/gt1Mcfd2^−/− mice (Figure 1D). No significant decrease in the number of Lman1^gt1/gt1Mcfd2^−/− mice was observed (P = .670; OR, 0.82; 95% CI, 0.34-1.95).

Next, we performed intercrosses of Lman1^+/gt1Mcfd2^+/− mice on the C57BL/6 background (Table 2). We observed ∼42% of the expected number of Lman1^gt1/gt1Mcfd2^−/− mice (P = .028; OR, 0.41; 95% CI, 0.18-0.88), suggesting a partial lethal phenotype of DKO mice on this genetic background. In this mating, the Lman1^gt1/gt1 genotype (14.0%, P < .001) is underrepresented, whereas the Mcfd2^−/− genotype (26.5%, P > .7) is of the expected number, suggesting that the partial lethal phenotype of DKO mice can be attributed to the Lman1^gt1/gt1 genotype. We also performed an Lman1^+/gt1Mcfd2^−/− × Lman1^+/gt1Mcfd2^+/− cross using the offspring of intercrosses of Lman1^+/gt1Mcfd2^+/− mice and again observed a lower than expected number of DKO mice (P = .004; OR, 0.43; 95% CI, 0.24-0.74) and underrepresentation of the Lman1^gt1/gt1 genotype (14.9%; P < .001), with expected numbers for the Mcfd2^−/− genotype (48%, P > .8). MCFD2-KO and surviving DKO mice exhibit normal growth, fertility, and gross appearance; no apparent abnormalities on standard necropsy evaluation; and normal overall survival (up to 1.5 years of follow-up).

Mice were generated from intercrosses of Lman1^+/gt1Mcfd2^+/− mice on the C57BL/6 background and intercrosses of the resulting offspring. Mean FV and FVIII levels in LMAN1-KO mice were decreased to 58% and 54% of WT levels, respectively (P < .001; Figure 2A), consistent with previous reports.¹⁷  Average FV and FVIII levels in MCFD2-KO mice were further decreased to 35% and 33% of WT levels (P < .0001 for FV and P < .005 for FVIII between Lman1^gt1/gt1 and Mcfd2^−/− mice). Surprisingly, average FV and FVIII levels in DKO mice are similar to Lman1^gt1/gt1 mice, at ∼60% of WT levels (Figure 2A), and are significantly higher than for Mcfd2^−/− mice (P < .0001 for FV and FVIII). In vitro studies by Nyfeler et al suggest that AAT is also an LMAN1 cargo.²⁰  However, plasma AAT levels are not significantly decreased in LMAN1-KO mice, as determined by western blot analysis.¹⁷  We analyzed plasma AAT levels using a more quantitative ELISA method. Plasma AAT levels were higher in male mice than in female mice, consistent with previous reports.³⁰  We observed small, but statistically significant, decreases in AAT levels in male mice of all 3 KO mouse lines (P < .05) but not in female mice (Figure 2B).

Intracellular FVIII expression in vivo is too low to be detected by currently available methods. Therefore, we focused our analysis on FV, reasoning that functions of LMAN1 and MCFD2 should be similarly reflected in FV and FVIII. Using an antibody that detects intracellular FV by western blot analysis, we found elevated FV levels in liver of all 3 KO mouse lines compared with WT mice (Figure 3A), consistent with a defect in ER exit. Similar to LMAN1-KO mice, higher intracellular levels of AAT were also observed in MCFD2-KO and DKO mice (Figure 3). Intracellular AAT was increased in livers of all 3 KO mouse lines in male and female mice, with higher levels observed in males (Figure 3B), consistent with the higher plasma AAT levels detected in male mice. To distinguish between proteins accumulating in the ER and proteins that have moved into the Golgi and beyond, we treated liver extracts with endoglycosidase H (endo H), which cleaves high-mannose oligosaccharides from N-linked glycoproteins in the ER and the cis Golgi. Glycoproteins that have moved into the medial Golgi are resistant to endo H. FV bands shifted to faster migration positions in sodium dodecyl sulfate-polyacrylamide gel electrophoresis after endo H treatment, consistent with the removal of N-linked glycans, indicating that the majority of intracellular FV is located in the ER (Figure 4A). Thus, higher FV levels in MCFD2 and DKO liver cell extracts represent increased FV protein retained in the ER.

Intracellular AAT was detected as 2 bands by immunoblotting (Figure 3). The upper band of AAT migrates at the same position as plasma AAT (Figure 3B) and is resistant to endo H treatment (Figure 4A), therefore corresponding to a postmedial Golgi fraction of the protein. The lower band of AAT is sensitive to endo H treatment, indicating that this is the ER species (Figure 4A). The intensity of the lower band increased in all 3 KO mouse lines, suggesting that AAT accumulates in the ER of hepatocytes of these mice (Figure 3A). Consistent with the immunoblotting results, immunofluorescence staining of frozen liver sections demonstrated increased intracellular accumulation of AAT in hepatocytes of all 3 KO mouse lines (Figure 4B).

Similar to LMAN1 deficiency,¹⁷  MCFD2 deficiency in mice leads to less marked reductions in plasma FV and FVIII levels than in human F5F8D patients with MCFD2 mutations.¹⁶  FV and FVIII levels are lower in MCFD2-KO mice compared with the previously reported LMAN1-KO mice,¹⁷  which are consistent with observations in humans with mutations in the MCFD2 versus LMAN1 genes.¹⁶  Murine FV is produced in hepatocytes and megakaryocytes,^24,29  whereas FVIII is synthesized in endothelial cells.^18,31  In addition, FV protein levels are nearly 2 orders of magnitude higher than those of FVIII.³²  Thus, the LMAN1–MCFD2 cargo receptor must function in different cell types and transport proteins of vastly different abundances. MCFD2 deficiency has no effect on LMAN1 level, and a small amount of MCFD2 exists in LMAN1-deficient cells.⁵  LMAN1 and MCFD2 have been shown to interact with FVIII independently of each other,^11-12  raising the possibility that deficiency of either protein alone may not completely abolish cargo receptor function. However, ablation of both components of the LMAN1–MCFD2 cargo receptor did not further reduce FV and FVIII levels, suggesting that an alternative secretory pathway may be responsible for residual FV/FVIII secretion in DKO mice. Surprisingly, FV and FVIII in DKO mice match the higher levels found in LMAN1-KO mice and not the lower levels of MCFD2 mice. Mechanisms that might explain this puzzling observation include a compensatory increase in an alternative FV/FVIII-secretion pathway (potentially mediated by an alternative cargo receptor) in the absence of LMAN1, but not MCFD2, although few ER-to-Golgi cargo receptors have been identified in mammalian cells.^33-36  Previous immunoprecipitation studies, with and without a chemical cross-linker, were unable to identify any other interaction partners for LMAN1 or MCFD2.^5,10-12 

LMAN1 and MCFD2 are both widely expressed, including in many cell types that do not express FV and FVIII,¹⁷  suggesting functions in addition to FV and FVIII transport. In vitro studies suggest that MCFD2 may function as an autocrine/paracrine factor for the maintenance of neural stem cells.^37-38  However, the lack of neurological and cognitive manifestations in F5F8D patients and mice argues against an essential role in neural stem cell survival in vivo. In contrast to the partial lethality observed in C57BL/6J LMAN1-KO mice, MCFD2-KO mice exhibit normal survival. These data suggest the existence of an additional cargo(s) that is dependent on LMAN1 but not MCFD2. The LMAN1-KO partial lethal phenotype is unlikely to be due to bleeding, given the only modest reductions in FV and FVIII levels. These data suggest a requirement in late embryogenesis for another, as yet unknown, LMAN1 cargo(s), with variable levels of expression for this cargo, or another component of the associated pathway, across mouse strains.

Although the ER-to-Golgi transport of FV and FVIII requires LMAN1 and MCFD2, previous reports suggest that MCFD2 is dispensable for certain cargo proteins, such as the lysosomal enzymes cathepsin C and cathepsin Z in vitro,³⁹  although no decrease in steady-state levels of cathepsin Z was detected in plasma of Lman1^gt1/gt1 mice.¹⁷  However, we observed AAT accumulation in the ER of liver hepatocytes in single-KO and DKO mice, suggesting that LMAN1 and MCFD2 are also required for ER-to-Golgi transport of AAT. ER accumulation of AAT, although similar to that of FV, is only accompanied by a slight reduction in plasma AAT levels in male mice, suggesting that the steady-state AAT level in plasma is subject to additional regulation. Residual secretion of AAT and FV/FVIII in DKO cells suggests the existence of an alternative ER-to-Golgi transport pathway for this glycoprotein, as well as for FV/FVIII. The mechanism responsible for a mild decrease in plasma AAT only in male mice is unknown, although AAT levels are upregulated by testosterone and downregulated by estradiol,³⁰  suggesting the potential for a slight shift in the balance of AAT synthesis and degradation in male vs female mice. Taken together, our studies suggest additional nonoverlapping functions of LMAN1 and MCFD2 in vivo.

The authors thank Bilge Ozel for assistance with statistical analysis.

This study was supported by grants from the National Institutes of Health, National Heart, Lung, and Blood Institute (R01HL094505 and R35HL135793) (D.G.) and National Cancer Institute (R03CA202131) (B.Z.) and the Natural Science Foundation of Xinjiang Uygur Autonomous Region (2016D01A020 [M.Z.]). D.G is an investigator at the Howard Hughes Medical Institute.

Contribution: B.Z., C.Z., M.Z., and D.G. designed the study; C.Z., M.Z., L.E., W.W., and B.Z. performed research; C.Z., M.Z., and B.Z. analyzed data and wrote the manuscript; and all authors provided critical comments.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Bin Zhang, Genomic Medicine Institute, Lerner Research Institute of Cleveland Clinic, 9500 Euclid Ave/NE50, Cleveland, OH 44195; e-mail: zhangb@ccf.org.