Identification of HSPA8 as an interacting partner of MAB21L2 and an important factor in eye development

Abstract

Background:

Pathogenic variants in human MAB21L2 result in microphthalmia, anophthalmia and coloboma. The exact molecular function of MAB21L2 is currently unknown. We conducted a series of yeast two-hybrid (Y2H) experiments to determine protein interactomes of normal human and zebrafish MAB21L2/mab21l2 as well as human disease-associated variant MAB21L2-p.(Arg51Gly) using human adult retina and zebrafish embryo libraries.

Results:

These screens identified klhl31, tnpo1, TNPO2/tnpo2, KLC2/klc2, and SPTBN1/sptbn1 as co-factors of MAB21L2/mab21l2. Several factors, including hspa8 and hspa5, were found to interact with MAB21L2-p.Arg51Gly but not wild-type MAB21L2/mab21l2 in Y2H screens. Further analyses via 1-by-1 Y2H assays, co-immunoprecipitation and mass spectrometry revealed that both normal and variant MAB21L2 interact with HSPA5 and HSPA8. In situ hybridization detected co-expression of hspa5 and hspa8 with mab21l2 during eye development in zebrafish. Examination of zebrafish mutant hspa8^hi138Tg identified reduced hspa8 expression associated with severe ocular developmental defects, including small eye, coloboma, and anterior segment dysgenesis. To investigate the effects of hspa8 deficiency on the mab21l2^{Arg51_Phe52del} allele, corresponding zebrafish double mutants were generated and found to be more severely affected than single mutant lines.

Conclusion:

This study identifies heat shock proteins as interacting partners of MAB21L2/mab21l2 and suggest a role for this interaction in vertebrate eye development.

Introduction

Pathogenic variants in MAB21L1 and MAB21L2 of the male-abnormal 21 like family have been shown to cause human ocular developmental disorders. Dominant and recessive alleles in MAB21L2 were identified in several families with microphthalmia, anophthalmia or coloboma (MAC) ^1–5. For dominant alleles, various missense variants affecting the same arginine residue at position 51 were most common (4/7 dominant families) ^2,3,5. For MAB21L1, recessive alleles (both missense and presumed loss-of-function) were associated with cerebello-oculo-facio-genital syndrome in six unrelated families, where the main ocular features include corneal dystrophy/opacities and nystagmus ^6,7. More recently, our group reported additional MAB21L1 alleles in patients with microphthalmia, coloboma and aniridia, including a dominant allele affecting the same arginine residue at position 51, p.(Arg51Leu), conserved in all MAB21L proteins ⁸.

The functional roles of MAB21L proteins are still largely unknown. The solved crystal structure for MAB21L1 suggested possible nucleotidyltransferase (NTase) activity due to a high amount of structural overlap with a known NTase, cyclic GMP-AMP synthase (cGAS), involved in cytosolic DNA recognition and subsequent synthesis of second messenger cGAMP ^9–11. However, in vitro experiments have been unable to provide evidence supporting this activity ^5,10. A role in transcriptional regulation has also been proposed as both Mab21l1 and Mab21l2 have been shown to localize to the nucleus ¹² and to have mild affinity for nucleic acids, in particular single-stranded RNA ^5,10. In addition, MAB21L2 showed transcriptional repressor/co-repressor activity in one study based on in vitro luciferase assays ¹³. Further data has supported a role for Mab21l1/Mab21l2 in Tgf-β signaling ^13,14 and/or the Pax6 pathway ^15–18.

One way to gain insight into a protein’s function(s) is to discern its possible interactions using a large-scale unbiased approach. To identify critical interactions of MAB21L2 involved in disease pathogenesis, we used a yeast two-hybrid screen to compare the interactome of human and zebrafish wild-type MAB21L2/mab21l2 and p.(Arg51Gly) mutant and identified several heat shock proteins (HSP) including HSPA8/hspa8 and HSPA5/hspa5 as major interactants. Furthermore, we utilized the zebrafish model to reveal the importance of hspa8 in ocular development by itself and in a mab21l2-deficient background.

Results

Identification of candidate co-factors of MAB21L2 by Y2H screening

Yeast two-hybrid analyses were conducted as an unbiased approach to identify potential binding partners of the human/zebrafish MAB21L2/mab21l2 wild-type and human p.(Arg51Gly) mutant proteins (Table S1). MAB21L2 proteins are highly conserved with matching lengths (359 aa) and 97% (348/359) identify at the protein level between human and zebrafish. Two libraries were selected for this experiment, 1) human retina (adult), to identify interactions relevant to human retina and 2) zebrafish embryo (18-20 hpf), to capture important interactions taking place during vertebrate embryonic development.

In the screens, MAB21L2-WT versus human retina library tested 77.5 million interactions and isolated 156 positive clones; MAB21L2-WT versus zebrafish embryo (18-20 hpf) library tested 61.6 million interactions and isolated 356 positive clones; MAB21L2-p.(Arg51Gly) versus zebrafish embryo (18-20-hpf) library tested 36.3 million interactions and isolated 268 positive clones; and mab21l2-wt versus zebrafish embryo (18-20 hpf) library tested 58.6 million and 74 million interactions and identified 86 and 234 positive clones (using a LexA and GAL4 DNA binding domain, respectively) (Table S1). To note, two screens, MAB21L2-p.(Arg51Gly) versus zebrafish embryo library and mab21l2-wt (pB27 vector) versus zebrafish embryo, resulted in mild autoactivation of the HIS3 reporter gene and thus were treated with 3-AT, resulting in a reduced number of tested interactions due to higher selective pressure.

The results were first prioritized based on confidence score, with a focus on those with A-D scores (Table S2). Then, interactions which appeared in two or more independent wild-type screens were selected for further analysis, which included 30 factors (Table 1). In this list, we noted several interactions that appeared in three or more independent screens and had a confidence score of A or B in at least one of these screens, and that had overlapping selected interaction domains (SID; region that was shared for all prey mapping to the same reference protein) between screens, increasing the possibility that these were relevant findings. This included kelch-like protein 31 (klhl31), transportin-1 (tnpo1) and transportin-2 (TNPO2/tnpo2), kinesin light chain 2 (KLC2/klc2) and spectrin beta chain (SPTBN1/sptbn1) (Table 1). Previous data demonstrated that both wild-type MAB21L2 ^2,12 and MAB21L2-p.(Arg51Gly) ² variant display primarily nuclear localization. Analysis of available subcellular localization data revealed that half of all prioritized factors (15/30) and four out of the five most consistent interactants between screens, klhl31, tnpo1, tnpo2, and SPTBN1/sptbn1, were reported to have nuclear staining (uniprot.org; ¹⁹) and thus are likely to co-localize with MAB21L2 in the cell. The remaining protein, KLC2/klc2, has not been observed in the nucleus; however, since MAB21L2 also demonstrates some cytoplasmic staining ¹², this interaction cannot be excluded. All five of these most consistent interacting partners exhibited high confidence scores with both wild-type MAB21L2/mab21l2 and mutant MAB21L2-p.(Arg51Gly), indicating that these interactions are not disrupted by the pathogenic variant.

Table 1:

Summary of MAB21L2 interacting proteins independently identified in Y2H assays.

Gene Symbol	MAB21L2-WT v. Human Retina		MAB21L2-WT v. Zebrafish Embryo		mab21l2-wt v. Zebrafish Embryo		MAB21L2-Arg51Gly v. Zebrafish Embryo		Protein Name	Subcellular location1,2
	CS	SID	CS	SID	CS	SID	CS	SID
ARIH2/arih2	C	170-393	C	223-427					E3 ubiquitin-protein ligase	N
evi5b			D	169-682	D	169-682	D	169-682	Ecotropic viral integration site 5 homolog	N
hnf4a			C	159-378	B	132-351	C	159-381	Hepatocyte nuclear factor 4-alpha	N
klhl31			A	52-262	A	50-262	A	52-264	Kelch-like protein 31	N
mcm5			D	592-721	D	592-721	C	595-715	DNA replication licensing factor	N, C
nop56			D	62-319	B	167-319			Nucleolar protein 56	N
nup205			D	230-606	D	230-606			Nuclear pore complex protein	N
pdcd11			D	1449-1617	D	1448-1616			Protein RRP5 homolog	N
pparab			D	66-459	D	66-459			Peroxisome proliferator-activated receptor alpha	N
psme4b			D	1590-1827	D	1597-1827			Proteasome activator complex subunit 4	N, C
ranbp9			C	18-393	D	1-451	C	42-393	Ran-binding protein 9	N
STAT3/stat3	D	1-705	D	53-539	D	53-539			Signal transducer and activator of transcription 3	N
tnpo1			B	131-553	B	131-553	A	131-553	Transportin-1	N
TNPO2/tnpo2	A	212-484	A	283-474	A	198-478	A	209-478	Transportin-2	N
ANKFY1/ankfy1	D	617-949			D	134-468			Rabankyrin-5	C
ckap5			D	1299-1682	D	1299-1682	D	1299-1682	Cytoskeleton-assoc. protein 5	C; c-skel.
DSP/dspa	D	1814-1997	D	1150-1602					Desmoplakin	C-skel, M
eif3ba			D	53-339	D	125-387			Eukaryotic translation initiation factor	C
etf1			D	130-341	C	130-340	B	138-340	Eukaryotic peptide chain release factor	C
hook3			D	349-717	D	349-717			Protein Hook homolog 3	C; c-skel.
KLC2/klc2	B	1-150	A	1-130	A	58-155	A	58-130	Kinesin light chain 2	C; c-skel.
myo1eb			D	715-1096	D	715-1096	D	715-1096	Unconventional myosin-Ie	C; c-skel.
SPTBN1/sptbn1	D	507-681; 939-1326	B	590-667	D	254-710; 953-1345	D	474-667	Spectrin beta chain	C; c-skel., N
trim71			D	267-500	D	265-563			E3 ubiquitin-protein ligase	C
PNPLA6/pnpla6	D	761-930			D	712-903			Patatin-like phospholipase domain-containing protein 6	C; ER
VPS35/vps35	D	5-734			D	238-656			Vacuolar protein sorting-assos. protein	C
ephb6			D	621-798	D	723-886			Ephrin type-B receptor 6	ECM
lygl2			D	4-191	D	4-191			Lysozyme g	ECM
tnc			D	1245-1542	D	1094-1359	D		Tenascin	ECM
sdhb			D	40-237	D	40-237			Succinate dehydrogenase iron-sulfur subunit	Mito

¹uniprot.org;

²Tang et al., 2003;

CS: confidence score; SID: selected interaction domain; ECM extracellular matrix; ER: Endoplasmic reticulum; C: cytoplasm; c-skel.: cytoskeleton; M: membrane; Mito: mitochondrion; N: nuclear. Interactions identified in three or more independent screens AND with CS between A-C in at least one screen are indicated in bold.

Dominant mutations in MAB21L2 have been hypothesized to have gain-of-function effects ⁵. Therefore, to identify interactions unique to the mutant, factors interacting with MAB21L2-p.(Arg51Gly) with high confidence (A-D) scores but not showing a similar binding to the wild-type protein in all three independent screens were selected (Table 2). A total of 27 interactions were detected exclusively for the mutant; of these, only one A-level interaction was identified (hspa8), one B-level (gfm1) and four C-level (hspa5, sfpq, kif4 and blzf1). Notably, two heat shock proteins were present in this short list, heat shock cognate 71 kDa protein (hspa8) and heat-shock 70 kDa protein 5 (hspa5), both with noted nuclear staining (uniprot.org; ^20–23) and thus consistent with MAB21L2/MAB21L2-p.(Arg51Gly) primary subcellular localization ^2,12. The interacting fragment of hspa8 involved amino acids 255-420 (which contains two (out of four) ATP-binding subdomains and a part of the substrate-binding domain (uniprot.org; ²⁴)), while for MAB21L2-p.(Arg51Gly) it is unclear what specific region was involved in the interaction since the entire protein was used as a bait.

Table 2:

Summary of unique interactions identified for MAB21L2-Arg51Gly in Y2H assays.

Gene Symbol	MAB21L2-WT v. Human Retina	MAB21L2-WT v. Zebrafish Embryo	mab21l2-wt v. Zebrafish Embryo	MAB21L2-Arg51G v. Zebrafish Embryo		Protein Name	Subcellular location1
				CS	SID
gli2a	-	-	-	D	441-847	Zinc finger protein	N
cux2a	-	-	-	D	1-325	Cut-like homeobox 2	N
dnajc7	-	-	-	D	39-267	DnaJ homolog subfamily C member 7	N
hspa5	-	-	-	C	279-508	Heat-shock 70 kDa protein 5	C, N
hspa8	-	-	-	A	255-420	Heat shock cognate 71 kDa protein	N
ik	-	-	-	D	77-319	IK cytokine, downregulator of HLA II	N
kif4	-	-	-	C	871-970	Chromosome-associated kinesin	N
kntc2l	-	-	-	D	160-367	Kinetochore protein NDC80 homolog	N
kpnb3	-	-	-	D	517-795	Importin-5	N
mphosph8	-	-	-	D	574-856	M-phase phosphoprotein 8	N
prdm15	-	-	-	D	743-1047	PR domain zinc finger protein 15	N
prpf40a	-	-	-	D	382-602	Pre-mRNA-processing factor 40 homolog A	N
prr12a	-	-	-	D	2259-2493	Proline-rich protein 12	N
psme3	-	-	-	D	61-241	Proteasome activator complex subunit 3	N
rarab	-	-	-	D	65-359	Retinoic acid receptor alpha	N
sfpq	-	-	-	C	226-494	Splicing factor, proline- and glutamine-rich	N
trip12	-	-	-	D	314-585	E3 ubiquitin-protein ligase TRIP12	N
hsp90b1	-	-	-	D	418-750	Endoplasmin	C; ER
cita	-	-	-	D	1562-1947	Non-specific serine/threonine protein kinase	C; C-skel.
klc4	-	-	-	D	37-136	Kinesin light chain 4	C-skel.
stom	-	-	-	D	990-1359	Stomatin	C-skel., M
utrn	-	-	-	D	1855-2231	Utrophin	C-skel., M
blzf1	-	-	-	C	161-394	Golgin-45	Golgi
zgc:66014	-	-	-	D	237-507	RAB11-binding protein RELCH homolog	Endosome, Golgi
irig1	-	-	-	D	346-604	Leucine-rich repeats and immunoglobulin-like domains protein 1	M
gfm1	-	-	-	B	557-724	Elongation factor G, mitochondrial	Mito
fam83fa	-	-	-	D	63-555	Protein FAM83F	UNK

¹uniprot.org;

CS: confidence score; SID: selected interaction domain; ER: Endoplasmic reticulum; C: cytoplasm; C-skel: cytoskeleton; M: membrane; Mito: mitochondrion; N: nuclear; UNK: unknown

To verify the Y2H data, 1-by-1 interaction assays were conducted in yeast including the zebrafish hspa8 fragment isolated from the initial Y2H screen (nucleotides 762-1494) that was tested against human MAB21L2-WT and MAB21L2-p.(Arg51Gly). Interestingly, the interaction was confirmed for both MAB21L2-p.(Arg51Gly) (Figure 1B), along with MAB21L2-WT (Figure 1A). This is contrary to the results of the Y2H, where the interaction was only identified for the mutant protein. However, yeast growth was stronger for the mutant; this growth difference could be indicative of differing affinities between mutant and wild-type MAB21L2 with hspa8.

Figure 1: HSPA8/HSPA5 interaction with MAB21L2 proteins.

A, B. Growth plates for 1-by-1 Y2H interaction assay. pB66ø (empty GAL4 DNA-Binding Domain (DBD) vector) was used as a negative control. MAB21L2-WT (A) or MAB21L2-p.(Arg51Gly) (B) was tested against a zebrafish hspa8 fragment; yeast was grown on DO-2 (-tryptophan, -leucine; selects for presence of both bait and prey) and DO-3 selective media (-tryptophan, -leucine, -histidine; selects for interaction of bait and prey). Note the weakened yeast growth on DO-3 medium for MAB21L2-WT (A), in comparison to MAB21L2-p.(Arg51Gly) (B). C, D. Co-immunoprecipitation assay for human full-length FLAG-tagged MAB21L2-WT or MAB21L2-p.(Arg51Gly) and human full length myc-tagged HSPA8 (C) or HSPA5 (D). Cells were transfected with respective constructs, pcDNA empty plasmid was used as a control. Cell lysates were collected, with a portion reserved for ‘input’, and a portion immunoprecipitated with anti-FLAG antibodies. Western blot analysis with anti-myc antibodies for detection of HSPA8/5 and anti-FLAG antibodies for detection of MAB21L2-WT/p.(Arg51Gly) was performed. Both MAB21L2-WT and MAB21L2-p.(Arg51Gly) co-precipitated with HSPA8 and HSPA5.

Validation of Y2H data in mammalian cells

To explore MAB21L2 interactions in mammalian cells, we performed mass spectrometry analysis of proteins co-immunoprecipitated with either wild-type or mutant (p.(Arg51Gly)) FLAG-tagged MAB21L2 proteins in HLE-B3 human lens epithelial cells. 91 and 107 proteins were found to be significantly enriched in precipitates obtained with wild-type and mutant MAB21L2 proteins, correspondingly (Tables S4 and S5). When the mass spectrometry and Y2H data were compared, a total of 6 proteins were found to overlap, DSP, SPTBN1, VPS35, HSPA5, HSPA8 and HSP90B1 (Table S3). For HSPA8, an abundance ratio of 74.817 was identified when comparing MAB21L2-p.(Arg51Gly) transfected to untransfected cells and 58.74 when comparing MAB21L2 wild-type transfected to untransfected cells (1.274 mutant/WT abundance ratio), suggesting a stronger interaction between HSPA8 and the MAB21L2 mutant but not significantly different (P value 0.099) (Table S3). For HSPA5, abundance ratios of 13.927 for MAB21L2-p.(Arg51Gly) mutant and 17.106 for wild-type MAB21L2 transfected cells/untransfected controls were identified (0.814 abundance ratio; P value 0.813).

To further validate human full-length HSPA8 and HSPA5 interactions against MAB21L2-WT and p.(Arg51Gly), co-immunoprecipitation (co-IP) was performed. The myc-tagged HSPA8 (n=6 experiments; Figure 1C) and myc-tagged HSPA5 (n=2 experiments; Figure 1D) were both immunoprecipitated with FLAG-tagged MAB21L2-WT as well as MAB21L2-p.(Arg51Gly).

hspa8 and hspa5 expression in zebrafish

To investigate the expression of hspa8 and hspa5 during zebrafish ocular development in situ hybridization was performed on 24, 48 and 96-hpf embryos. Both hspa5 and hspa8 demonstrated broad expression patterns at all embryonic stages, including strong presence during eye development. The expression of hspa8 and hspa5 was enriched in the developing eye, retinal ciliary marginal zone, and lens, as well as the brain and craniofacial regions, overlapping mab21l2 expression domains (Figure 2). The co-expression of mab21l2 and hspa genes during embryonic development supports the possibility of functionally important interaction between the encoded proteins.

Figure 2: Expression pattern of both hspa5 and hspa8 overlaps with mab21l2 at embryonic stages.

RNA-scope analysis of hspa5 or hspa8 (green) and mab21l2 (yellow) in 24 (A-A” and K-K”), 48-(B-F” and L-O”) and 96-hpf (G-J” and P-S”) wild-type embryos. (l) lens, (r) retina, (mb) midbrain, (opt) optic tectum, (ba) branchial arches, (th) thalamus, (hth) hypothalamus. White arrows indicate the optic fissure (A-B, K-K”) and ciliary marginal zone (C-I and M-R).

hspa8^{hi138Tg/hi138Tg} embryos demonstrate reduced level of hspa8 transcript

To explore the role of hspa8 in zebrafish ocular development, we obtained a mutant line from the Zebrafish International Resource Center (ZIRC), hspa8^hi138Tg. This line was generated via a large-scale retroviral-mediated insertional mutagenesis screen ²⁵ using established protocols ²⁶. Genomic DNA sequencing confirmed the location of the retroviral insert 350 bps upstream of the hspa8 5’ untranslated region (UTR).

To identify the effect of the 5’ insertion on hspa8 expression, RT-PCR was performed on 48-hpf embryos. In comparison to wild-type embryos, homozygous hspa8^hi138Tg embryos exhibited a noticeable reduction in hspa8 expression (Figure 3A). To further validate these results, qRT-PCR was performed. Significant reduction in hspa8 expression was observed with both primer pairs tested with the following results: exon 2 – exon 3: −404.6 fold-change, P<0.0001; exon 2 – exon 4: −281.1 fold-change, P<0.0001 (Figure 3B).

Figure 3: Molecular and phenotypic characterizaton of homozygous hspa8^hi138Tg zebrafish embryos.

A. RT-PCR analysis of hspa8 transcript level comparing 48-hpf wild-type and homozygous hspa8^hi138Tg embryos. Two different primer sets were utilized; primer set one targeted exons 2 and 3, primer set two targeted exons 2 and 4. β-actin was used as a loading control. B. qRT-PCR analysis of the hspa8 transcript level utilizing two different primer sets (as utilized in RT-PCR experiments) comparing 48-hpf wild-type and homozygous hspa8^hi138Tg embryos. Error bars indicate standard error of the mean (SEM). **** indicates statistical significance with a P value <0.0001. C-I’. Gross morphological assessment of 24-, 48- and 72-hpf control and homozygous hspa8^hi138Tg embryos (stage is marked in the left corner and type (control/hspa8 mutant) in the right corner of each image). Homozygotes displayed systemic defects (small head, misshapen body); ocular defects included small, colobomatous eyes at all stages (red asterisks indicate retinal edges in F’, G’ and H’). At 72-hpf, homozygotes displayed severely reduced anterior chamber space and thickened corneas (compare black arrow in I to red arrow in I’). J-L’. H&E staining of transverse sections from 24- , 48- , and 72-hpf wild-type and homozygous hspa8^hi138Tg embryos. At 24-hpf embryos display abnormally shaped optic cup (red asterisk in J’) and lens-cornea attachment (red arrow in J’). At 48- and 72-hpf, small disordered retina, thick cornea and reduced anterior chamber space are observed (red arrow in K’ and L’). (r) retina, (l) lens, (c) cornea, (ac) anterior chamber. Size of scale bar for C-I’ is 200μm and for J-L’ is 25μm.

hspa8^{hi138Tg/hi138Tg} embryos show ocular abnormalities

Homozygous embryos were identified in ~25% of progeny from heterozygous adult crosses, consistent with the principles of Mendelian inheritance. Previous observations of homozygous hspa8^hi138Tg mutants at 5-dpf had identified gross abnormalities of the hindbrain, thinner body with a rounder yolk, underdevelopment of the liver/gut, small head with central nervous system necrosis, and small eyes ^25,27. We examined homozygous offspring at early developmental stages (24-72-hpf) with an emphasis on ocular structures and noted the following phenotypes: small eyes with variable coloboma and anterior segment anomalies (thickened developing cornea and absent anterior chamber space at 72-hpf) (Figure 3C–I’); ocular features were correlated with embryos with more severe coloboma displaying more significant anterior segment defects. Heterozygous eyes appeared normal at all stages. Progeny from heterozygous crosses were reared to adulthood; genotyping revealed no surviving homozygotes, suggesting lethality, with normal survival and morphology for heterozygous hspa8^hi138Tg adults.

To further assess the ocular phenotype, we performed histological analysis on 24-, 48- and 72-hpf wild-type (9, 10 and 9 eyes, respectively) and hspa8^hi138Tg homozygous mutants (4, 5 and 6 eyes, respectively) (Figure 3J–L’). At 24-hpf, smaller eyes with cornea-lenticular adhesions and abnormal retinal morphology were observed. At 48- and 72-hpf, mutant eyes show aberrant retina lacking any signs of retinal lamination, abnormally thick corneas, and absent anterior chamber.

Genetic interaction of hspa8 and mab21l2 alleles in zebrafish

In order to examine the genetic interaction between mab21l2 and hspa8, double mutants carrying hspa8^{hi138Tg 25} and mab21l2^mw702 (mab21l2^{Arg51_Phe52del} ) alleles were generated. The mab21l2^mw702 allele is an in-frame deletion of two amino acids including the arginine at position 51, c.151_156delCGTTTC, p.(Arg51_Phe52del); fish homozygous for this allele display coloboma ². Double heterozygous adults, hspa8^hi138Tg/+: mab21l2^mw702/+, were crossed to generate different combinations of hspa8^hi138Tg and mab21l2^mw702 alleles. Phenotyping of the ventral retina/optic fissure region and lens was completed for the following numbers of eyes for each genotype at 48-hpf: hspa8^+/+:mab21l2^+/+ 12 eyes, hspa8^hi138Tg/+:mab21l2^+/+ 10 eyes, hspa8^+/+:mab21l2^mw702/+ 15 eyes, hspa8^hi138Tg/+:mab21l2^mw702/+ 20 eyes, hspa8^{hi138Tg/hi138Tg}:mab21l2^+/+ 5 eyes, hspa8^+/+:mab21l2^mw702/mw702 32 eyes, hspa8^{hi138Tg/hi138Tg}:mab21l2^mw702/+ 10 eyes; hspa8^hi138Tg/+:mab21l2^mw702/mw702 11 eyes, hspa8^{hi138Tg/hi138Tg}:mab21l2^mw702/mw702 4 eyes).

Most single and double heterozygous embryos showed a complete fusion of the optic fissure (Figure 4). For single homozygous embryos, most/all fish demonstrated coloboma phenotypes: hspa8^+/+;mab21l2^mw702/mw702 fish displayed grade 1 (25%), grade 2 (46.9%) and grade 3 (21.9%) colobomas with small number of normal eyes (6.2%), while hspa8^{hi138Tg/hi138Tg}:mab21l2^+/+ embryos all had coloboma with equal numbers (40%) for grade 1 and 2, and 20% for grade 3 (Figure 4G). Embryos with heterozygosity for one gene and homozygosity for another one showed the following results: hspa8^hi138Tg/+:mab21l2^mw702/mw702 eyes had a similar distribution to hspa8^+/+:mab21l2^mw702/mw702 while hspa8^{hi138Tg/hi138Tg};mab21l2^mw702/+ embryos showed a more severe coloboma phenotype in comparison to hspa8 homozygotes in wild-type background with 10% grade 1, 50% grade 2, 20% grade 3 and 20% grade 4. Embryos double homozygous for hspa8 and mab21l2 demonstrated the most severe abnormalities, with moderate to severe coloboma in all (75% ranking as grade 4 and the remaining 25% grade 3) (Figure 4G).

Figure 4. Gross phenotypic analysis of hspa8^hi138Tg and mab21l2^mw702 (mab21l2^{Arg51_Phe52del}) double mutants.

A-F. Representative images of 48-hpf embryos with various hspa8^hi138Tg and mab21l2^mw702 allelic combinations; specific genotypes (left corner) and assigned grade for coloboma (right) are indicated; retinal margins are indicated with yellow outline. G. Graph of the proportion of embryos per grade to total embryos per allele combination. Grade 0 shown in black, grade 1 in green, grade 2 in blue, grade 3 in purple and grade 4 in red. H. Graph of average lens area per allele combination. Error bars indicate SEM. P values are indicated above each bracket, with significance determined by a P value <0.05. (*** <0.0001; ** <0.0005; * <0.05).

To characterize the lens phenotype, we measured lens area in different genotypic groups and observed significant differences. Single heterozygous or homozygous embryos for mab21l2 showed no significant difference from wild-type, consistent with previous observations ², while single heterozygous or homozygous embryos for hspa8 displayed smaller lenses (at 87.9% and 62.5% of wt, respectively) (Figure 4H). Interestingly, embryos with homozygosity for mab21l2 in hspa8 heterozygous background showed smaller lenses with a significant difference from wild-type, mab21l2 homozygotes in wild-type background and hspa8 heterozygotes, suggesting that interaction between the mutant alleles in these fish produced the abnormal lens size. Consistent with this, double homozygous embryos were the most affected and showed significant differences from all other groups.

Discussion

Previous work has shown that pathogenic variants in the MAB21L2 gene are associated with MAC-spectrum ocular phenotypes ^1–5. However, to date, the function(s) of the MAB21L2 protein remain elusive, and knowledge of binding partners is limited. In this study we have identified potential cofactors of MAB21L2 using a yeast two-hybrid screen followed by additional validations in both yeast and human cells.

Several interactions were identified in three independent MAB21L2/mab21l2 wild-type screens with high confidence scores. These included TNPO2/tnpo2/tnpo1, KLC2/klc2, SPTBN1/sptbn1 and klhl31. The identification of tnpo1 and TNPO2/tnpo2, with both transportins participating as nuclear transport receptors for nuclear import ^28,29, is consistent with previous data showing MAB21L2 is localized primarily to the nucleus ^2,12. KLC2 ³⁰ and SPTBN1 ¹⁹ have been associated with TGF-β signaling; MAB21L2 likewise has been implicated in this pathway ^13,14. Finally, Klhl31, a kelch-like family member, has been found to modulate canonical WNT signaling ³¹ and act as a repressor in MAPK/JNK signaling ³², of which both pathways have been implicated in eye development ^33–35. To note, these interactions were observed with high confidence for both wild-type and mutant (p.Arg51Gly) MAB21L2 proteins, indicating that they are not affected by the human disease-associated variant. If transportins are involved in MAB21L2 nuclear import, this would be consistent with previous reports that nuclear localization is uninterrupted for MAB21L2-p.(Arg51Gly) ².

Through the yeast two-hybrid assays, we also revealed interactions that were potentially unique to or favored by the MAB21L2-p.(Arg51Gly) mutant in comparison to the wild-type protein. This discovered high confidence interactions with hspa5 and hspa8, two different heat shock proteins. HSPA5 and HSPA8 genes encode for HSP70 proteins, with HSPA5 encoding for an endoplasmic reticulum-specific HSP70, otherwise known as BiP or GRP-78, and HSPA8 encoding constitutively active HSP70 (HSC70).

HSPA5 encoded GRP-78 (Bip) is known for its role in maintenance of proteostasis, specifically within the endoplasmic reticulum ^36–38. It has a major role in regulation of the unfolded protein response (UPR) ^36–38. However, it has also been found in several additional compartments throughout the cell. Reports have identified GRP-78 at the plasma membrane where it acts as a co-receptor in signal transduction ^39–41. Additional studies have also confirmed the presence of HSPA5 in the nucleus of human cells ^20,22. In mice, Hspa5 null animals demonstrate early embryonic lethality, while heterozygotes develop without any obvious abnormalities ⁴². Another study demonstrated an important role for Hspa5 in kidney development in Xenopus laevis via regulation of retinoic acid signaling ²². Retinoic acid (RA) signaling is also a pathway with known importance in ocular development ⁴³, and several RA pathway genes have been implicated in MAC-spectrum disease, including STRA6 ^44–51 ALDH1A3 ⁵², RARB ^52–54 and RBP4 ^55–57.

HSC70, encoded by HSPA8, is most well-known for its classical role as a molecular chaperone where it participates in the maintenance of proteostasis (via assisting protein folding, translocation, assembly/disassembly of complexes, and degradation) ⁵⁸. Additionally, it is also known to be located throughout the cell in various compartments, including the nucleus, MAB21L2’s primary subcellular compartment ^2,12. Previous work has shown an ability for HSC70 proteins to shuttle across the nuclear envelope ⁵⁹ and import other cytosolic proteins with nuclear localization signals (NLS) ⁶⁰. The HSPA8 gene also contains two of its own NLS’s ^21,23. Interestingly, previous work has also shown that in response to a stress stimuli, HSC70 nucleocytoplasmic shuttling becomes inhibited and the protein remains sequestered in the nucleus ⁶¹. Additionally, it has been demonstrated that if HSC70 is prohibited from nuclear release during the recovery phase post-stress, cell survival is negatively affected upon application of a secondary stressor event ⁶², suggesting it’s shuttling capability is vital. Finally, there have been various reports implicating Hspa8 (Hsc70) in transcriptional regulation and involvement in assembly of complexes at promoter regions ^63–65.

The interaction of MAB21L2 wild-type and p.(Arg51Gly) mutant proteins with HSPA5/hspa5 and HSPA8/hspa8 was validated by mass spectrometry, 1-by-1 assay (for zebrafish hspa8), as well as co-IP experiments in HLE-B3 cells (for both HSPA5 and HSPA8), thus confirming them as novel interacting partners of MAB21L2. Contrary to the results from Y2H, where hspa8 and hspa5 were unique interactions specific to the MAB21L2 p.(Arg51Gly) mutant, other assays demonstrated that both wild-type and mutant MAB21L2 proteins are able to interact with these heat shock proteins; of note, trends towards a weaker interaction for the wild-type were evident in two of the validation assays but with no statistically significant difference. While a difference in protein-binding affinity remains a possibility, other methods for its quantification will need to be employed in order to clarify this issue. Overall, a stronger association of HSPA8 with the MAB21L2 p.(Arg51Gly) mutant would be consistent with its role in facilitating correct protein folding, as well as stabilizing or degrading mutant proteins.

Neither HSPA5 nor HSPA8 are currently associated with distinct human phenotypes. Studies of a zebrafish hspa8 mutant, hspa8^hi138Tg, demonstrated its importance in normal embryonic development, including the eye. Ocular and systemic phenotypes were present at all examined stages; the ocular phenotype was severe and included small, underdeveloped eyes, with abnormal retina morphology, coloboma and anterior segment (lens and corneal) defects. These features overlap phenotypes observed in mab21l2-deficient lines ^2,66–68. Moreover, embryos with double deficiency for hspa8^hi138Tg and mab21l2^{Arg51_Phe52del} (mab21l2^mw702) demonstrated more pronounced phenotypes suggesting functional interaction between these factors: fish with homozygosity for hspa8 in mab21l2 heterozygous background showed stronger coloboma phenotypes than hspa8 homozygotes with wild-type mab21l2 alleles while embryos with mab21l2 homozygosity in hspa8 heterozygous background showed a significant reduction in lens size in comparison to single mab21l2 homozygotes; double homozygous embryos displayed the most severe phenotypes with the highest grade/incidence of coloboma and smallest lenses. This indicates a possible role for HSPA8/hspa8 in stabilization and correct folding of wild-type and mutant MAB21L2/mab21l2 proteins, supported by more severe disease in the mab21l2^{Arg51_Phe52del} mutant in hspa8-deficient background. Furthermore, these data suggest a possible role of HSPA5 and/or HSPA8 variants in the phenotypic variability associated with MAB21L2 alleles in human families that needs to be investigated.

Experimental Procedures

Yeast Two-Hybrid (Y2H) Screen and Interaction Confirmation

A full-length N-terminal tagged human MAB21L2 (NM_006439) wild-type (#EX-V1703-M11, Genecopoeia, Rockville, MD, USA) and a mutant human MAB21L2-p.(Arg51Gly) (all in a pEZ-M11 vector), previously described², as well as zebrafish mab21l2 were used in Y2H analysis ².

All plasmids were submitted to Hybrigenics Services (https://www.hybrigenics-services.com/; Evry, France) for use in the ULTImate Y2H^™ screening analysis which was conducted as previously described ⁶⁹. Briefly, in the initial set of screens, human MAB21L2 and MAB21L2-p.(Arg51Gly) cDNA were cloned into a pB66 ⁶⁹ vector to be expressed in-frame with a N-terminal GAL4 DNA binding domain. Constructs were tested for autoactivation of the HIS3 gene reporter and treated with 3-aminotriazol (3AT) if needed. The following prey libraries were tested: MAB21L2-WT vs. human retina, MAB21L2-WT vs. zebrafish embryo (18-20 hours post fertilization (hpf)) and MAB21L2-p.(Arg51Gly) vs. zebrafish embryo (18-20 hpf). In a second set of screens, zebrafish (Danio rerio) mab21l2 cDNA was cloned into both a pB27, (derived from pBTM116 ⁷⁰), to be expressed in-frame with a N-terminal LexA binding domain, and the pB66 vector, as described above. Both were screened against a zebrafish embryo (18-20 hpf) prey library.

Data was analyzed and each identified interaction was assigned a Predicted Biological Score (PBS^®), a statistically calculated confidence score ^71,72. Briefly, confidence scores were classified as follows: A- Very high confidence, B- High confidence, C- Good confidence, D- Moderate confidence (includes likely false-positives or interactions with low mRNA representation in prey library), E-unreliable due to non-specific interactions, F- technical artifacts.

To test individual interactions in yeast, a 1-by-1 assay was also conducted through Hybrigenics Services, which is based on the reconstitution of an active transcription factor to initiate expression of a HIS3 gene reporter. Briefly, full-length bait proteins (MAB21L2-WT or MAB21L2-p.(Arg51Gly), as used in the initial Y2H screen above) were cloned in frame into pB66 ⁶⁹, to be expressed as a C-terminal fusion to a GAL4 DNA binding domain. The prey fragment for hspa8 was extracted from the ULTImate Y2H^™ screen (specifically, MAB21L2-p.(Arg51Gly) against the Zebrafish Embryo (18-20 hpf) library) and was cloned into a pP6 plasmid ⁷³ to be expressed in frame with a GAL4 activation domain. The assay was conducted by transforming bait constructs into yeast haploid cells CG945 (mata) and prey constructs into YHGX13 (Y187 ade20101::loxP-kanMX-loxP, mata). Yeast cells were mated to obtain diploids ⁶⁹. To note, SMAD and SMURF were used as control bait and prey constructs, respectively ⁷⁴. Yeast cells were grown on two different selective mediums. DO-2 lacks tryptophan and leucine and was used as a control to determine that both bait and prey were present; DO-3 lacks tryptophan, leucine and histidine, where growth in the absence of histidine suggests an interaction between the bait and prey.

Co-Immunoprecipitation (co-IP) studies

HSPA8 and HSPA5 interactions with MAB21L2 were tested using co-IP. Briefly, human lens epithelial (HLE-B3) cells (CRL-11421^™, ATCC^®, Manassas, VA, USA) were transfected with 7.5ug of bait (N-terminal FLAG-tagged MAB21L2 or MAB21L2-p.(Arg51Gly), described above) and 7.5ug prey. HLE-B3 cells demonstrate endogenous expression of MAB21L2². Prey constructs were as follows: Full-length C-terminal Myc tagged HSPA8 (NM_006597; #EX-W1208-M09, GeneCopoeia) and HSPA5 (NM_005347; #EX-T3592-M09, GeneCopoeia) in a pEZ-M09 vector or control Invitrogen^™ pcDNA3.1 vector (V79020, ThermoFisher Scientific, Waltham, MA, USA). Transfections were conducted using Invitrogen^™ Lipofectamine^™ 2000 Transfection Reagent (11668019, ThermoFisher Scientific) with Opti-MEM (31985070, ThermoFisher Scientific). Cells were cultured as previously described ⁸. Two methods were used for co-IP: 1) whole-cell lysates were collected 48 hours post transfection and lysed with 1% Triton X-100 (with protease inhibitor and phosphatase inhibitor). An aliquot was reserved for input control. Then, lysates were incubated with a monoclonal anti-FLAG M2 antibody produced in mouse (F1804-200UG, Sigma Aldrich) rotating overnight at 4°C. Following, the antibody-lysate sample was incubated with protein A/G PLUS-Agarose beads (sc-2003, Santa Cruz Biotechnology, Inc., Dallas, TX, USA) rotating 2 hours at 4°C. Beads were washed with 1% Triton X-100 and proteins eluted from beads with 1X sample buffer and heating at 95°C for 5 minutes. 2) whole-cell lysates were collected 48 hours post-transfection and lysed with a 1X immunoprecipitation (IP) Buffer (1% Triton X-100, 150 mM NaCl, 20mM Tris-HCl pH7.5, 1mM EDTA). An aliquot was reserved for input control. Lysates were incubated with Anti-FLAG^® M2 Magnetic Beads (#M8823, Sigma-Aldrich, St. Louis, MO, USA) overnight at 4°C. The Beads were washed with 1X IP buffer and protein was eluted from beads via incubation with 4X Laemmli buffer at 50°C for 10 minutes. Western blot was conducted using a wet-tank transfer and membrane probed. The following antibodies were utilized: 1:1000 Myc-Tag (9B11) Mouse mAb (#2276S, Cell Signaling, Danvers, MA, USA), monoclonal mouse anti-FLAG (F1804-200UG, Sigma Aldrich), Myc-Tag (71D10) Rabbit mAb (#2278, Cell Signaling Technology) and polyclonal rabbit anti-FLAG (SAB1306078-400UL, Sigma-Aldrich) primary antibodies; 1:2000 goat anti-mouse (#PA1-74421, ThermoFisher Scientific) and goat anti-rabbit (#32460, ThermoFisher Scientific) HRP conjugated secondary antibodies.

Mass Spectrometry

HLE-B3 cells were transfected with 7.5μg of N-terminally FLAG-tagged MAB21L2 wild-type or MAB21L2-p.(Arg51Gly), as described above. Two days post-transfection cells were collected and incubated in lysis buffer (50mM HEPES pH7.4, 150mM NaCl, 150mM egtazic acid (EGTA), 10% glycerol, 1% Triton X-100) for 30 minutes on ice. Then, cells were spun down (12,000 X g, 10 minutes, 4°C) and the supernatant collected. Cell lysate was then incubated overnight at 4°C with monoclonal mouse anti-FLAG (F1804-200UG, Sigma Aldrich), phosphatase inhibitor and protease inhibitor. The following day, cell lysate mixture was incubated with Dynabeads^™ Protein G (10003D, ThermoFisher Scientific) for 45 minutes at 4°C. Then, a series of washes were performed, 3 times with IP washing buffer (50mM HEPES pH7.5, 150mM NaCl, 1mM EGTA, 1.5mM MgCl2, 0.1% IgePal), and then twice with a no detergent IP buffer (50mM HEPES, 150mM NaCl, 1mM ethylenediaminetetraacetic acid (EDTA)). The samples were then submitted to the Medical College of Wisconsin Mass Spectrometry Core for analysis.

Studies of zebrafish lines

The care and use of zebrafish (Danio rerio) was approved by the Institutional Animal Care and Use Committee at the Medical College of Wisconsin. Housing, care and breeding was carried out as previously described ⁷⁵. Age was determined by hours post fertilization (hpf) and days post fertilization (dpf), along with a morphological assessment ⁷⁶.

The hspa8^hi138Tg (transgenic insertion upstream of the hspa8 coding region) zebrafish line was obtained from the Zebrafish International Resource Center (ZIRC) and used for characterization of the ocular phenotype. The line was generated using retroviral-mediated insertional mutagenesis ²⁵ using previously established protocols for zebrafish ²⁶. The exact upstream position of the transgene to the hspa8 start codon was determined using PCR to amplify a region containing the transgene and subsequent genomic region of hspa8 (primers F-5’-CAAACCTACAGGTGGGGTCTTTC-3’ anneals to retroviral insertion, R-5’-GGGACTTCAACCGACAAGAACC-3’ anneals to hspa8 genomic region) followed by Sanger sequencing (primer R-5’-TGAATGAAATCACCCTGCAC-3’). Genotype was determined using two separate PCRs, 1) to determine the presence of the mutant allele (F-5’-CAAACCTACAGGTGGGGTCTTTC-3’, R-5’-GGGACTTCAACCGACAAGAACC-3’; wild-type will have no amplification), 2) to determine the presence of wild-type allele and thus discriminate between heterozygous or homozygous hspa8 mutants (F-5’-TCAAAAGCCATTCGTGATGA-3’, R-5’-GGGACTTCAACCGACAAGAACC-3’; homozygotes will have no amplification).

To generate hspa8^hi138Tg/mab21l2-deficient double heterozygous fish, a previously generated heterozygous mab21l2^{Arg51_Phe52del} line (mab21l2^mw702 as catalogued in ZFIN (Zebrafish Information Network)) ², was crossed with heterozygous hspa8^hi138Tg fish. Embryos were raised to adulthood and genotyped, using the above protocol for hspa8, and as previously described for mab21l2 ².

Whole-mount images of 24, 48 and 72-hpf zebrafish were taken using a ZEISS SteREO Discovery.V12 microscope (Carl Zeiss Inc., Thornwood, NY, USA) with a ZEISS AxioCam MRc or AxioCam 305 Color (Carl Zeiss Inc.). Coloboma severity grading for 48-hpf hspa8^hi138Tg;mab21l2^{Arg51_Phe52del} embryos was performed blinded by two individuals for all genotype combinations using the grading protocol in Brown et al. (2009) as a guide, with modification for a 48-hpf timepoint ⁷⁷. For grade 0, a normal eye with a fully closed optic fissure was observed, for grade 1- no gap but a visible line likely indicating the presence of intact basement membrane was present, for grade 2- an obvious small gap between the opposing retinal margins was observed, for grade 3- a medium sized gap between the disorderly aligned retinal margins (equal or exceeding pupil radius in any part of the gap) was present, and for grade 4- a large gap between the disorderly aligned retinal margins (greater than pupil diameter in any part of the gap) was noted. Measurements of lens area were taken utilizing ImageJ software (https://imagej.nih.gov/ij/) by outlining the lens boundary three times and calculating an average area. All graphs were generated using GraphPad Prism 9 (San Diego, CA, USA’ https://www.graphpad.com/scientific-software/prism/).

Histological analysis of 24-, 48- and 72-hpf wild-type and homozygous hspa8^hi138Tg embryos and 24-hpf hspa8^hi138Tg;mab21l2^{Arg51_Phe52del} embryos, reared in either E2 embryo medium or 1X phenylthiourea, were performed as previously described ⁷⁵, where embryos were fixed in Davidson’s solution overnight and then transferred to 70% ethanol; preserved embryos were encapsulated in histogel (HG-4000-012, ThermoFisher Scientific) and submitted to the Medical College of Wisconsin Histology Core for processing, sectioning and H&E staining.

In situ hybridization and qRT-PCR analyses

In situ hybridization was performed to determine the expression pattern of hspa5 and hspa8, as previously described ⁷⁵ using the following probes: Dr-mab21l2-C2 (499581-C2, Advanced Cell Diagnostics (ACD), Newark, CA, USA), Dr-hspa8-C3 (499661-C3, ACD), and Dr-hspa5-C3 (499651-C3, ACD).

For qRT-PCR analysis of hspa8 transcript levels, two biological replicates (5 embryos each) of 48-hpf homozygous hspa8^hi138Tg embryos were collected from a hspa8^hi138Tg/+ X hspa8^hi138Tg/+ cross. Tails were used for genotyping, and the heads/trunks processed for RNA. RNA was isolated using the Direct-zol^™ RNA MiniPrep kit (R2052, ZymoResearch, Irvine, CA, USA). cDNA was synthesized using SuperScript III First Strand Synthesis System (18080051, ThermoFisher Scientific). Zebrafish hspa8 and actb1 were amplified from cDNA. For hspa8 the following primer pairs were utilized: 1) F-5’-TTGATCTCGGGACCACCTAC-3’ (exon 2), R-5’-TCAGACTGAACAACGCCATC-3’ (exon 3) (210 base pairs (bp) intron between; expected product size 232 bp); 2) F-5’-TTGATCTCGGGACCACCTAC-3’ (exon 2), R-5’-CAGCAGCAGTTGGTTCATTG (exon 4) (includes exon 3 and 2 introns (210bp and 212 bp); expected product size 513 bp). For actb1, the following primer pair was utilized: F-5′-GAGAAGATCTGGCATCACAC-3′ (exon 3), R-5′-ATCAGGTAGTCTGTCAGGTC-3′ (exon 4) (311 bp intron between; expected product size from cDNA 323 bp). qRT-PCR was then conducted where samples were run in triplicate and a no template control was included; fold change was calculated. Graphs were generated using GraphPad Prism 9. Statistical significance was determined using an unpaired sample t-test with a P value of <0.05.

Supplementary Material

Supinfo

Table S1. Summary of screening parameters for yeast two-hybrid analyses.

Table S2. Summary of all positive clones identified through four separate yeast two-hybrid screens with A-D confidence.

Table S3. Abundance ratios and corresponding P values for mass-spectrometry identified interactions overlapping Y2H protein orthologues.

Table S4. Summary of proteins interacting with WT MAB21L2 in B3 cells and identified by mass-spectrometry.

Table S5. Summary of proteins interacting with MAB21L2-R51G variant in B3 cells and identified by mass-spectrometry.