Mutant MESD links cellular stress to type I collagen aggregation in osteogenesis imperfecta type XX

Abstract

Aberrant forms of endoplasmic reticulum (ER)-resident chaperones are implicated in loss of protein quality control in rare diseases. Here we report a novel mutation (p.Asp233Asn) in the ER retention signal of MESD by whole exome sequencing of an individual diagnosed with osteogenesis imperfecta (OI) type XX. While MESD^D233N has similar stability and chaperone activity as wild-type MESD, its mislocalization to cytoplasm leads to imbalance of ER proteostasis, resulting in improper folding and aggregation of proteins, including LRP5 and type I collagen. Aggregated LRP5 loses its plasma membrane localization to disrupt the expression of WNT-responsive genes, such as BMP2, BMP4, in proband fibroblasts. We show that MESD is a direct chaperone of pro-α1(I) [COL1A1], and absence of MESD^D233N in ER results in cytosolic type I collagen aggregates that remain mostly not secreted. While cytosolic type I collagen aggregates block the intercellular nanotubes, decreased extracellular type I collagen also results in loss of interaction of ITGB1 with type I collagen and weaker attachment of fibroblasts to matrix. Although proband fibroblasts show increased autophagy to degrade the aggregated type I collagen, an overall cellular stress overwhelms the proband fibroblasts. In summary, we present an essential chaperone function of MESD for LRP5 and type I collagen and demon-strating how the D233N mutation in MESD correlates with impaired WNT signaling and proteostasis in OI.

Introduction

Osteogenesis imperfecta (OI) is a rare, heterogeneous group of inherited disorders of bone and connective tissue [1]. It is also known as “brittle bone disease” because it causes generalized osteopenia leading to multiple fractures [2]. Low bone mass, bone fragility, and growth deficiency are the primary clinical manifestations of OI [3]. The severity of the disease ranges from mild or moderate to perinatally fatal [3]. Currently, variations in 22 different genes are known to cause OI [4, 5]. Genetic variations in COL1A1 and COL1A2 are observed in 85% of cases of OI [6]. The general characteristics of OI include Wormian bones, blue sclerae, hearing loss, dentinogenesis imperfecta, generalized osteopenia, scoliosis, and deformities of long bones [1].

Physiologically, loss of collagen deposition in bone and inflammation are the main causes of weak bone in people with OI [7]. However, molecular mechanisms show that aberrant expression, stability, localization, and functionality of not only collagen, but also other proteins disrupt bone development and homeostasis in OI [8]. For example, OI types I-IV are caused by autosomal dominant mutations in the COL1A1 and COL1A2 genes [9, 10, 11]. Autosomal recessive mutations in chaperones (such as SERPINH1, MESD) [12, 13], post-translational modifiers (such as P3H1, PPIB, FKBP10, SERPINF1) [14, 15, 16, 17], and scaffold protein (CRTAP) [18] also cause different types of OI. In addition to impaired proteostasis, transcriptional imbalance caused by mutations in transcription factors (e.g., SP7, CREB3L1) [19, 20] and functional inactivation of mutant membrane transporters (e.g., IFITM5, TMEM38B, SPARC) [21, 22,23] and receptors (e.g., KDELR2, MBTPS2) [24, 25] also lead to various types of OI. Mutations in other proteins (such as BMP1, WNT1, CCDC134) [26, 27, 28] involved in osteogenesis have also been shown to cause OI through abnormal bone formation and development.

For all the diversity of mechanisms, the induction and progression of the certain types of OI have been associated with an increased proteotoxic burden due to the accumulation of misfolded collagens in affected individuals [29]. Thus, one of the central problems in OI is the misfolding and anomalous secretion of type I collagen. Under normal conditions, collagen biogenesis is controlled by a series of events, including folding, disulfide bridging, proline hydroxylation, and assembly of helical peptides coordinated by the functions of several ER proteins. Mutational inactivation of one or more proteins in this pathway leads to qualitative and quantitative defects of collagens. Decreased secretion of collagens, disruption of the matrix by secretion of misfolded collagens, and overwhelming proteotoxicity due to accumulation of unsecreted collagens underlie collagenopathies, including OI [30].

Type XX is a recently characterized form of OI that has been shown to be caused by autosomal recessive mutations in the mesoderm development LRP chaperone (MESD) protein [13, 31]. While the first five mutations (frameshift and missense) associated with OI type XX are localized in exon 3 (corresponding to the region downstream of the chaperone domain of MESD) [13], compound heterozygous frameshift mutations were later found in exon 2 and exon 3 in three individuals with severe forms of OI type XX [32]. Fusion of SENP1 with MESD was reported in one individual with infantile teratoma [33]. Progressive OI in these individuals was characterized by skeletal fragility, oligodontia, and blue sclera [13, 32].

MESD is a chaperone of low-density lipoprotein receptor-related proteins (LRP2, LRP4, LRP5, and LRP6), which function as coreceptors of WNT [34]. MESD has also been shown to promote phagocytosis in retinal epithelial cells and activate MUSK [35]. MESD consists of four regions - an N-terminal disordered signal peptide followed by a chaperone and an escort domain and an extreme C-terminal ER retention signal. The functional core of the chaper-one domain adopts a ferredoxin-like β-α-β-β-α-β fold [36, 37] that mediates mainly hydrophobic interactions with specific domains [38], such as the beta-propellor of LRP6 [39], of substrate proteins. The dynamic interaction of an N-terminal α-helix with the beta-sheets of the globular core determines the ‘open’ and ‘closed’ conformations of MESD [40]. LRPs bound to MESD lack the ability to bind to WNT. Therefore, MESD is separated from LRPs in the Golgi by a histidine switch in the beta-propeller of LRPs [39]. Overall, the folding of LRPs by MESD is involved in embryonic polarity and osteogenesis [34]. Mutations in MESD are thought to cause bone malformations primarily by disrupting the WNT/β-catenin pathway [13]. However, it is not known how MESD mutations can induce cellular proteotoxicity and cause severe collagen loss in the bones of affected individuals with OI.

In this study, we decipher the molecular mechanisms of OI type XX due to the homozygous D233N mutation in the ER retention signal of MESD. We find that mislocalization of MESD^D233N in the cytoplasm causes ER stress leading to protein aggregation and ubiquitination in fibroblasts of the proband. Mislocalization of MESD^D233N causes aberrant trafficking and aggregation of LRP5, leading to perturbed WNT signaling in the proband fibroblasts. Furthermore, we show that MESD is a direct chaper-one of type I collagen and that loss of MESD from ER leads to aggregation of type I collagen. Aggregated type I collagen is not secreted from cells, leading to the induction of severe proteotoxicity. However, some aggregated type I collagen is cleared from the cell by autophagy. The perturbed WNT signaling and overall proteotoxicity is manifested by cell cycle arrest, impaired attachment of cells to the extracellular matrix, and a reduction in membrane dynamics in proband fibroblasts. Therefore, deregulated LRP/WNT signaling and intracellular proteotoxicity are simultaneously contributing to the pathogenicity in OI type XX.

Results

Exome sequencing identifies a missense, p. Asp233Asn, variant of MESD associated with OI type XX

Clinical report

A girl of south Indian origin showed vertebral fracture in her eighth year of age. She was born to a nonconsanguineous married couple (Fig. 1A) at full-term via noninvasive vaginal delivery with a birth weight of 3 kg (-1.2 SD). Her antenatal history was unremarkable. At eighth year of her age, her height, weight, and head circumference were 89.5 cm (+1.1 SD), 22.8 kg (-2.4 SD), and 46 cm (-2.5 SD), respectively. Clinical examination from head-to-toe showed skin laxity, joint hypermobility without blue sclera and dentinogenesis imperfecta. Further examination of the eyes revealed a bilateral retinal hole. Her radiographs showed generalized osteopenia, vertebral fracture, multiple axial compression of D2, D7-L1, and dumbbell-shaped bilateral femoral metaphysis (Fig. 1B). Dual-energy X-ray absorptiometry (DEXA) revealed a T-score of -7 SD (the T-score of the age-matched control group is -5.5 SD) (Fig. S1). Her biochemical examinations revealed elevated serum alkaline phosphatase levels and vitamin D deficiency (Table S1).

Fig. 1. Whole exome sequencing identified a c.697G>A in exon 3 of MESD of an individual diagnosed with osteogenesis imperfecta type XX.

(A) Three generation pedigree of the proband. (B) Radiographs of the proband show vertebral fractures (arrows) and osteopenia. (C) Sanger sequencing and segregation analysis show homozygous state of c.697G>A variant in the proband and carrier status in her parents. (D) Representative diagram of the MESD (top) depicting exons and introns and protein domains (bottom). (E) Length of the PCR products of the reverse transcribed coding region of mature MESD mRNAs from control and proband fibroblasts. (F) The relative abundance of the transcripts of MESD in control and proband fibroblasts was analyzed by quantitative PCR. (G) Multiple sequence alignment of MESD protein of different organisms, showing conserved acidic amino acid in the third position of ER retention motif. (H) Poly-Phen-2 score predicts the pathogenicity of the D233N missense variant in MESD. (I) Theoretical estimation of the free energy change of protein folding due to mutation (ΔΔG), accounting for mutation-induced stability changes.

Exome sequencing of proband revealed a novel and homozygous missense variant, c.697G>A, in exon 3 of MESD (NM_015154.3) (Fig. 1C, Fig. S2, Table S2), resulting in a p.Asp233Asn (D233N) mutation in the ER retention motif of the protein (Fig. 1D, Table S3). This variant has not been previously reported in any of the control population databases. Sanger sequencing confirmed the homozygous state for the identified missense variant in the proband and also confirmed that the proband’s parents were carriers of the identified variant (Fig. 1C). The length of open reading frame of MESD was unaltered in transcripts of control and proband fibroblasts (Fig. 1E), indicating that the c.697G>A mutation did not generate a splice site in the mutant transcript. The proband fibroblasts did not show significantly different levels of the mutant MESD transcript compared with expression of the wild-type MESD transcript in control fibroblasts (Fig. 1F).

The aspartate residue in the ER retention motif of MESD is conserved across species (Fig. 1G, Fig. S3). Evolutionarily, this residue can only be replaced by a glutamate, suggesting that an acidic amino acid is essential and irreplaceable at the 233^rd position of human MESD. While the in silico tools predicted that the D233N mutation could damage the protein function (Table S4), the predictive pathogenicity of the D233N mutation was sublethal or benign (Fig. 1H). In addition, the mutation was predictably nondestabilizing because changes in the free energy of folding of MESD^D233N indicated a stable protein (Fig. 1I).

Mislocalization of MESD^D233N triggers endoplasmic reticulum stress

Previously reported OI type XX-related MESD mutants completely lacked the ER-retention signal [13, 32]. While the effect of Such MESD mutants on folding of LRPs were investigated, the global effect of mislocalization of the truncation mutants of MESD on ER homeostasis was unexplored. To understand the causal relationship between MESD^D233N and ER proteostasis, we first examined the localization of MESD^D233N in proband fibroblasts. In control fibro-blasts, wild-type MESD was found to be a fully ER-resident protein, as represented by high colocalization of MESD and an ER marker protein (CANX) (Fig. 2A, and B). In contrast, a large proportion of MESD^D233N was mislocalized in the cytoplasm of proband fibroblasts (Fig. 2A, and B). MESD^D233N was less enriched in the ER fraction of proband fibroblasts than in the enriched ER fraction of control fibroblasts, indicating a loss of MESD^D233N in ER of proband fibroblasts (Fig. 2C, and D). However, both colocalization and ER enrichment showed the presence of a small fraction of MESD^D233N in ER, signifying that the ER retention capacity of MESD^D233N is drastically reduced but not completely abolished.

Fig. 2. Mislocalization of MESD^D233N from ER to cytosol causes ER stress.

(A) Control and proband fibroblasts were immunostained for MESD and CANX. Representative confocal immunofluorescence microscopy images showing respective localization of MESD and MESD^D233N in ER and cytosol of control and proband fibroblasts. (B) Correlation coefficient of colocalization of MESD and MESD^D233N with CANX in control and proband fibroblasts (50 cells in each group). (C) ER was enriched from lysate of control and proband fibroblasts. Representative immunoblots of CANX, HSPA5 and MESD/MESD^D233N in cell lysate and enriched ER fraction. (D) Densitometric quantification of level of MESD relative to CANX in cell lysate and enriched ER of the immunoblot of 2C. (E) Relative quantification of the transcripts of HSPA5, PDIA2, EIF2AK3, and CALR, in control and proband fibroblasts. (F) Representative immunoblots of PDIA2, HSPA5 and HSP90B from the cell lysate of control and proband fibroblasts. (G) Densitometric quantification of level of PDIA2, HSPA5 and HSP90B relative to TUBA of the immunoblot of 2F. (H) Control and proband fibroblasts were stained with proteostat. Representative confocal fluorescence microscopy images showing high proteostat-positive protein aggregates in proband fibroblasts. (I) Quantification of number of protein aggregates per cell in control and proband fibroblasts (100 cells in each group). (J) Control and proband fibroblasts were immunostained for ubiquitin. Representative confocal immunofluores-cence microscopy images showing more ubiquitinated puncta in proband fibroblasts. (K) Quantification of number of ubiq-uitinated protein aggregates in control and proband fibroblasts (100 cells in each group). (L) Control fibroblasts were transiently transfected with control-siRNA or MESD-siRNA. These cells were stained with proteostat or immunostained for ubiquitin after 48 h. Representative confocal fluorescence microscopy images of MESD knockdown control fibroblasts showing higher number of proteostat-positive puncta and ubiquitinated protein aggregates. (M and N) Quantification of number of proteostat-positive puncta and ubiquitinated protein aggregates per cell of 2L (100 cells in each group). Quantifications are shown as mean ± S.D.; P values are indicated. TUBA is the loading control in the immunoblots. Scale bar in confocal microscopy images: 5 μm. Experiments were triplicated for biological and technical replicates.

Loss of MESD^D233N from ER resulted in ER stress, as evidenced by higher expression of stress responsive ER chaperones such as HSPA5, EIF2AK3, PDIA2, CALR, and HSP90B in proband fibroblasts (Fig. 2E, F, and G, S4A, S4B). Because ER stress has been reported to induce protein aggregation [46], we investigated whether loss of presence of MESD^D233N at ER correlated with unfolded protein response and subsequent protein aggregation. Indeed, proteostat staining showed increased accumulation of protein aggregates in proband fibroblasts compared with control fibro-blasts (Fig. 2H, and I). Moreover, the aggregated proteins in proband fibroblasts were also ubiquitinated (Fig. 2J, and K). In order to understand if the protein aggregation and their ubiquitination was direct effects of aberrant functioning of MESD^D233N, we knocked down MESD in control fibroblast cells. The MESD-siRNA treated control fibroblast cells showed proteostat-positive proteins aggregates and higher number of ubiquitinated protein aggregates compared to control-siRNA treated control fibroblast cells (Fig. 2L, M, and N). The MESD-siRNA treated control fibroblast cells were phenotypically similar to proband fibroblasts in terms of number of protein aggregates and ubiquitination of aggregated proteins, representing that loss of function of MESD^D233N in ER directly results the ER stress and proteotoxicity. Together, these results suggest that mislocalization of MESD^D233N ubiquitously induced ER stress in fibroblast cells.

D233N mutation has no effect on the stability and chaperone function of MESD

Having found the D233N mutation in the ER-retention signal of MESD causes loss of ER retention of MESD^D233N, we investigated whether the mutation alters the structure, stability and chaperone function of MESD^D233N compared with wild-type MESD. To analyze and compare the structural properties of the wild-type and mutant proteins, we estimated the secondary structures of MESD and MESD^D233N by circular dichroism (CD) spectra of bacterially purified recombinant proteins. The mutation showed no effect on the secondary structure of MESD^D233N compared with MESD. Both proteins showed nearly identical CD spectra (Fig. 3A), indicating the same content of different secondary structures in the proteins (Fig. 3B). Accordingly, slow thermal unfolding-coupled CD spectroscopy showed almost equal melting temperature (T_m) of the proteins (Fig. 3C, and D), indicating that both MESD and MESD^D233N have similar stability under in vitro conditions.

Fig. 3. The structure, stability, and chaperone activity of MESD^D233N is similar to wild-type MESD.

(A) Mean residue ellipticity changes (θ) of MESD and MESD^D233N probed by CD spectroscopy in the far-UV wavelength. (B) CDSSTR analysis of the percentage of secondary structures of MESD and MESD^D233N from the respective CD spectrum of the proteins. (C) Slow thermal unfolding CD spectra at λ=222 nm of MESD and MESD^D233N. (D) Melting temperatures of MESD and MESD^D233N. (E) Root mean square deviation (RMSD) values of the backbones of the MESD and MESD^D233N during molecular dynamics simulation. (F) Root mean square fluctuation (RMSF) values of the residues of MESD and MESD^D233N in the molecular dynamics simulation. (G) Folding assay/luciferase activity assay of Luc2 in presence of MESD, MESD^D233N and GFP in normal and heat-treated conditions. MESD and MESD^D233N showed similar chaperone activity. (H) Control and proband fibroblasts were immunostained for LRP5 and CANX. Representative confocal micros-copy images showing the localization of MESD in ER of control fibroblasts, and aggregated LRP5 in cytosol of proband fibroblasts. (I) Quantification of the number of control and proband fibroblast cells showing aggregates of LRP5 (100 cells in each group). (J) Representative immunoblot of LRP5 from whole cell lysate of control and proband fibroblasts. Quantifications are shown as mean ± S.D.; P values are indicated. Scale bar in confocal microscopy images: 5 μm. Experiments were triplicated for biological and technical replicates.

Unconstrained molecular dynamics simulation of the structures of mouse MESD and model MESD^D223N (mouse MESD is 224 amino acid long and the ²²¹REDL²²⁴ represent the ER-retention signal) in aqueous (buffer-like) environment showed that both proteins achieved similar stability over time. However, the root mean square deviation (RMSD) values of the backbone of MESD fluctuated intermittently before finally reaching RMSD values similar to those of the backbone of MESD^D223N (Fig. 3E). Root mean square fluctuation (RMSF) analysis showed almost similar values for the residues of MESD and MESD^D223N, except that the RMSF values of amino acids downstream of the signal sequence (29-34^th amino acid region) and a helix of the chaperone domain (helix-3: 155-168^th amino acid region) of MESD had relatively higher fluctuations than the corresponding amino acids of MESD^D223N (Fig. 3F). It was observed that the fluctuation of these residues led to transient destabilizing conformations of MESD. Concomitantly, we observed differential intramolecular crosstalk between regions of the structures of MESD and MESD^D223N. Under pre-simulation condition, the unstructured region (29-51^st amino acid region) upstream of the first helix of the chaperone domain was observed to interact simultaneously with two helical regions - helix-3 and the helix upstream of the ER-retention signal (helix-6: 210-218^th amino acid region) (Fig. S5A). Specifically, A29, D30, and T31 of the unstructured region interacted with S167 and Q168 of the helix-3 of globular domain, while R47, D48, and Y49 of the unstructured region interacted with residues A210, N215, and G218 of helix-6, respectively. This conformation probably provided the overall stability of the structures of MESD and MESD^D223N. At several time points during the simulation of MESD, the unstructured region positioned itself in such a conformation that its interactions with helix-3 of the globular domain was lost, resulting in a higher fluctuation of several residues of MESD. Such fluctuations led to transient destabilization of MESD. In contrast, the interaction of the unstructured region with the two helices in MESD^D223N was stable over time, resulting in the preservation of the compact and stable structure of MESD^D223N. A more flexible structure of MESD led to its extended conformation represented by a higher radius of gyration of the MESD structure (Fig. S5B). However, most of these residues of MESD and MESD^D223N showed similar number of interactions during the simulation time (Fig. S5C). It was found that the secondary structures of MESD and MESD^D223N remained similar throughout the simulation time (Fig. S5D), and this was consistent with the observations of CD spectroscopy.

MESD was reported to show chaperone activity for LRP5/6 [34]. Interestingly, the structure of its chaperone domain represents two consecutive beta-alpha-beta domain similar to some canonical bacterial chaperones [41]. Therefore, we speculated that the chaperone function of MESD might not be limited to LRP5, but the chaperone activity of MESD could extend to other proteins. To test this hypothesis and evaluate the chaperone activity of MESD^D233N, we used a refolding assay of heat-denatured Luc2 protein. Bacterially expressed recombinant MESD was able to maintain folded Luc2 and its catalytic (luciferase) activity (Fig. 3G). Recombinant MESD^D233N also maintained luciferase activity of Luc2 (Fig. 3G), indicating that the mutant protein has similar chaperone activity to the wild-type protein. However, we observed aggregated LRP5 dispersed in the cytoplasm of proband fibroblasts but not in control fibroblasts (Fig. 3H, I, and J). This result suggests that the intracellular chaperone function of MESD^D233N was suboptimal in proband, not because MESD^D233N was an inactive chaperone but probably because MESD^D233N was spatially excluded from its active site (ER).

Expression of bone morphogenic proteins is downregulated in proband fibroblasts

WNT signaling and its target genes, such as genes of bone morphogenic proteins (BMPs), are essential for bone developmental process [42, 43]. Disruption of BMP1 and WNT1 signaling are observed in osteogenesis imperfecta type XIII and type XV respectively [26, 27]. Since LRPs function as coreceptors of WNT [44], loss-of-function of LRP5, LRP6 and their upstream regulators, such as MESD, disrupt WNT reception by Frizzled. Because we found that mislocalization of MESD^D233N caused aggregation of LRP5, we tested if the aggregated LRP5 had any effect on downregulating the WNT signaling and expression of BMP2 and BMP4 proteins in proband. To test this, we checked the localization of LRP5 in the control and proband fibroblasts. Although, LRP5 showed significantly high colocalization with MESD in control fibroblasts, the aggregated LRP5 in proband fibroblasts showed loss of colocalization with the MESD^D233N (Fig. 4A, and B). Similarly, whereas LRP5 showed strong colocalization with membrane-bound FZD1 in control fibroblasts (Fig. 4C, and D), the colocalization of LRP5 with FZD1 was significantly lower in proband fibroblasts (Fig. 4C, and D). In fact, localization of LRP5 to the plasma membrane was not observed at all in proband fibroblasts. Sorting of LRP5 (and other LRPs) from ER to the plasma membrane through Golgi requires proper folding of LRP5 in the lumen of ER. If LRP5 is unfolded or misfolded in the ER because of a functional failure of its quality control proteins, such as MESD, the unfolded or misfolded LRP5 is unlikely to undergo anterograde transport to the Golgi and subsequently to the plasma membrane. Similar to many other unfolded and misfolded proteins in the lumen of ER, unfolded or misfolded LRP5 could undergo retrotranslocation from the ER followed by aggregation in the cytoplasm and degradation through the proteasomal and autophagy pathways. Because the proband’s fibroblasts lack sufficient chaperone activity of MESD in ER, LRP5 is not properly folded in ER, and this phenomenon may have prevented the localization of LRP5 in the membrane of the proband fibroblasts. MESD knockdown in human bone osteosarcoma epithelial cells (U2OS) also showed cytosolic LRP5 aggregates and loss of LRP5 colocalization with FZD1 at plasma membrane (Fig. S6A, S6B, S6C), indicating that functionally active MESD at ER is required for proper LRP5 folding and its localization to plasma membrane. Similar to the phenomenon that ER stress increases the expression of ER chaperones [45], the expression of LRP5 and other LRP proteins, such as LRP4 and LRP6, was higher in proband fibroblasts than in control fibroblasts (Fig. 4E). Thus, the lower localization of LRP5 in the membrane was not due to its lower expression but may have been due to the entrapment of LRP5 aggregates in the cytosol of proband fibroblasts. The expression of BMP2 and BMP4 was also significantly lower in proband fibroblasts compared with control fibroblasts (Fig. 4F, G, and H, S6D, S6E). Thus, the lower expression of WNT-responsive bone morphogenic proteins could be partly responsible for the lower strength and increased deformity of vertebral bone in proband.

Fig. 4. Co-association of LRP5 with MESD and FZD1 is perturbed and expression of BMPs is suppressed in proband fibroblasts.

(A) Control and proband fibroblasts were immunostained for LRP5 and MESD. Representative confocal immunofluorescence microscopy images showing colocalization of MESD and LRP5 in control fibroblasts, and loss of colocalization of MESD^D233N and LRP5 in proband fibroblasts. (B) The correlation coefficient of colocalization of MESD and LRP5 in control and proband fibroblasts(50 cells in each group). (C) Control and proband fibroblasts were immunostained for LRP5 and FZD1. Representative confocal immunofluorescence microscopy images showing colocalization of FZD1 and LRP5 in control fibroblasts, and loss of colocalization of FZD1 and LRP5 in proband fibroblasts. (D) The correlation coefficient of colocalization of FZD1 and LRP5 in control and proband fibroblasts (50 cells in each group). (E) Relative quantification of the transcripts of LRP4, LRP5 and LRP6 in control and proband fibroblasts. (F) Relative quantification of the transcripts of BMP2 and BMP4 in control and proband fibroblasts. (G) Representative immunoblot of BMP2 from whole cell lysate of control and proband fibroblasts. (H) Densitometric quantification of the level of cellular BMP2 relative to TUBA of the immunoblot of 4E. Quantifications are shown as mean ± S.D.; P values are indicated. TUBA is the loading control in the immunoblots. Scale bar in confocal microscopy images: 5 μm. Experiments were triplicated for biological and technical replicates.

MESD is a collagen I chaperone

Deficiency of collagen deposition in bone is characteristic of the fragile bone phenotype in OI. It is plausible that collagen protein formation, folding, and secretion are impaired in different types of OI. Since the loss of ER localization of MESD^D233N leads to ER stress, we tested whether ER proteomic stress has effects on specific collagen proteins, such as type I collagen, in proband fibroblasts. Interestingly, we found two different types of aggregates of type I collagen in the proband fibroblasts (Fig. 5A, and B), whereas the control fibroblasts did not show significant accumulation of type I collagen aggregates (Fig. 5A, and B). A significant population of proband fibroblasts showed small and scattered cytosolic aggregates (Fig. 5A, and B). Some proband fibroblasts also showed large and clumped aggregates (Fig. 5A, and B). While the large aggregates of type I collagen likely resulted from cohesion of smaller aggregates, the smaller aggregates may have originated from the misfolded entities of type I collagen. Concomitantly, the aggregated type I collagen lost their localization in ER of proband fibro-blasts, whereas a significant fraction of type I collagen showed localization in ER of control fibro-blasts (Fig. 5C, and D).

Fig. 5. MESD shows direct chaperone activity for pro-α1(I).

(A) Control and proband fibroblasts were immunostained for pro-α1(I) and pro-α 2(I). Representative confocal immunofluorescence microscopy images showing small-scattered and large-clumped aggregates of type I collagen in proband fibroblasts. (B) Quantification of percent of control and proband fibroblasts showing small-scattered and large-clumped aggregates of type I collagen (100 cells in each group). (C)Control and proband fibroblasts were immunostained for pro-α1(I) and CANX. Representative confocal immunofluo-rescence microscopy images showing high colocalization of pro-α1(I) with CANX at ER of control fibroblasts, whereas cytosolic aggregated pro-α1(I) significantly lose colocalization with CANX. (D) The correlation coefficient of colocalization of pro-α1(I) and CANX in control and proband fibroblasts (50 cells in each group). (E) Representative immunoblots of pro-α1(I) and MESD from the nondenaturing immune complexes of MESD from lysate of control fibroblasts. (F) Turbidity assay of pro-α1(I) in presence of MESD, MESD^D233N and GFP in normal and heat-treated conditions. MESD and MESD^D233N showed similar chaperone activity for pro-α1(I). (G) Relative quantification of the transcripts of COL1A1 and COL1A2 by qPCR in control and proband fibroblasts. (H) Left: Control fibroblasts were transiently transfected with control-siRNA or MESD-siRNA. These cells were immunostained for pro-α1(I) 48 h after transfection. Representative confo-cal immunofluorescence microscopy images showing type I collagen aggregates in MESD-siRNA treated control fibroblasts. Right: Proband fibroblasts were transiently transfected with empty pcDNA3.1 or a MESD-pcDNA3.1. Cells were immunostained for pro-α1(I) 48 h after transfection. Representative confocal immunofluorescence microscopy images showing decrease of type I collagen aggregates in MESD overexpressing proband fibroblasts. (I) Quantification of number of type I collagen aggregates per cell as described in 5H (100 cells in each group). Quantifications are shown as mean ± S.D.; P values are indicated. Scale bar in confocal microscopy images: 5 μm. Experiments were triplicated for biological and technical replicates.

Having observed that MESD could be functioning as a general chaperone, we hypothesized that collagen proteins could be direct substrates of MESD. In order to investigate whether MESD has chaperone function for pro-α1(I), we tested if MESD directly binds and modulates the stability of folded pro-α1(I) in control fibroblasts. pro-α1(I) was enriched with the nondenaturing immunoprecipitated fraction of MESD, suggesting that MESD interacts with pro-α1 intercellularly (Fig. 5E). Bacterially purified recombinant MESD showed chaperone activity for the heat-denatured region (300^th-744^th amino acid region) of recombinant pro-α1(I) purified from HEK293 cells. In a turbidity assay, heat-denatured pro-α1(I) exhibited aggregation represented by high absorbance at a light wavelength of 600 nm (Fig. 5F). However, co-incubation of pro-α1(I) with MESD or MESD^D233N reduced the aggregation of heat-denatured pro-α1(I) in solution (Fig. 5F). This suggests that MESD exerts a direct chaperone effect on pro-α1(I) by preventing aggregation of pro-α1(I). We excluded the possibility that the effect of molecular crowding acted to reduce the aggregation of pro-α1(I) in the presence of MESD, because coincubation of pro-α1(I) with GFP failed to reduce the aggregation of heat-unfolded pro-α1(I) (Fig. 5F). It was also noted that the endogenous expression of COL1A1 and COL1A2 were reduced in the proband fibroblasts compared to control fibroblasts (Fig. 5G).

We confirmed the chaperone activity of MESD for type I collagen in fibroblast cells. Knockdown of endogenous MESD in control fibroblasts resulted in the formation of type I collagen aggregates (Fig. 5H, and I), similar to as observed in proband fibroblasts. On the other hand, ectopically expressed wild-type MESD reduced type I collagen aggregates in proband fibroblasts to levels comparable to those in control fibroblasts (Fig. 5H, and I). Because MESD suppresses type I collagen aggregation in vitro, our data suggest that MESD plays a role in maintaining the folded form of type I collagen in ER, and this phenomenon may contribute to the stepwise assembly of collagen helices.

Nonsecreted cytosolic aggregates of type I collagen perturb cell attachment, disrupt endocytosis and block intercellular nanotubes to delay cell cycle progression of proband fibroblasts

The proband showed a typical phenotype of skin laxity. In several reports of dermal aging-related skin laxity, collagen production and secretion were shown to be decreased [47]. Aberrant collagen microfibril assembly and transport are also caused by mutations in collagen genes in OI [48]. The phenomenon of proband skin laxity and the finding that MESD is a direct chaperone of type I collagen prompted us to investigate the pathogenicity of type I collagen aggregates in proband fibroblasts in which the spatial function of MESD^D233N is impaired due to its mislocalization in the cytoplasm. We observed lower deposition of type I collagen at the edges of proband fibroblast cells (Fig. 6A, and B), implying that type I collagen was not adequately secreted from the proband fibroblasts.

Fig. 6. Type I collagen aggregates perturb cell attachment, intercellular nanotube dynamics and endocytosis to retard the growth and proliferation of proband fibroblasts.

(A) Control and proband fibroblasts were immunostained for pro-α1(I). Representative confocal immunofluorescence microscopy images showing high deposition of type I collagen at the edges of plasma membrane of control fibroblasts and type I collagen aggregates in proband fibroblasts. (B) Quantification of the percent of control and proband fibroblasts showing pro-α1(I) at the edges of cells (100 cells in each group). (C) Control and proband fibroblasts were immunostained for pro-α1(I) and ITGB1. Representative confocal immunofluorescence microscopy images showing high colocalization of type I collagen and ITGB1 at the edges of only control fibroblasts. (D) Correlation coefficient of colocalization of type I collagen and ITGB1 in control and proband fibroblasts (25 cells in each group). (E) Control and proband fibroblasts were stained for PAX1. Representative confocal immunofluorescence micros-copy images showing many focal adhesion sites in control fibroblasts. Many proband fibroblasts did not show good focal adhesion by PAX1. (F) Quantification of number of PAX1 foci in control and proband fibroblasts (100 cells in each group), (G) Control and proband fibroblasts were immunostained for pro-α1(I). Representative confocal immunofluorescence microscopy images showing higher number of ICNs and type I collagen aggregates in ICNs of proband fibroblasts. (H) Up: quantification of number of ICNs in control and proband fibroblasts (100 cells in each group); Down: quantification of the percent of ICNs with type I collagen aggregates in control and proband fibroblasts (100 cells in each group). (I) Up: control and proband fibroblasts were immunostained for EEA1. Representative confocal immunofluorescence microscopy images showing decreased EEA1 puncta, representing lower number of early endosomes, in proband fibroblasts. Down: control and proband fibroblasts were incubated with Dextran-AF-488. Representative confocal fluorescence microscopy images showed lower number of Dextran-AF-488-positive puncta, representing lower rate of endocytosis, in proband fibroblasts. (J) Quantification of number of EEA1 puncta and Dextran-AF-488-puncta in control and proband fibroblast cells (100 cells in each group). (K) Growth and proliferation assay of control and proband fibroblasts using trypan blue staining. Number of viable control and proband fibroblasts over a period of six days from cell seeding. (L) Proliferation assay of control and proband fibroblasts by crystal violet staining over a period of six days from cell seeding. (M) Analysis of cell cycle stages of control and proband fibroblasts by measuring propidium iodide stained DNA content in FACS (50,000 cells in each group). Quantifications are shown as mean ± S.D.; P values are indicated. Scale bar in confocal microscopy images: 5 μm. Experiments were triplicated for biological and technical replicates.

Because binding of integrin beta 1 (ITGB1) to type I collagen facilitates adhesion of cells to extracellular matrix (ECM) [49], we expected that unsecreted type I collagen might deregulate the attachment of proband fibroblasts to surface. Not only we observed low colocalization of pro-α1(I) with ITGB1 in proband fibroblasts (Fig. 6C, and D), type I collagen aggregates also sequestered ITGB1 in the cytosol of proband fibroblasts (Fig. 6C), whereas ITGB1 showed normal distribution and colocalization with pro-α1(I) at the cell periphery of control fibroblasts (Fig. 6C, and D). The weak attachment of proband fibroblasts with surface was also evident from the observed deficiency of paxillin-positive foci in proband cells (Fig. 6E, and F).

Proper cell attachment not only help in cellular communication [50] but also regulate dynamic membrane events, such as formation of intercellular nanotubes (ICNs) [51]. Therefore, we investigated whether loss of interaction of ITGB1 with type I collagen at membrane of proband fibroblasts has an effect on the destabilization of ICNs. Large aggregates of type I collagen were observed in a significant number of ICNs of proband fibroblasts (Fig. 6G, and H). Although we did not test molecular flux in control and proband fibroblasts, we speculate that the aggregates of type I collagen may interfere with the transport of molecules and vesicles through the ICNs of proband fibroblasts. It is also striking that the proband fibroblasts had a higher number of ICNs compared with the control fibroblasts (Fig. 6G, and H). Because the formation and structure of ICNs are dynamic [52], they form and dissolve in relation to the demands of cell-to-cell communication. The higher number of ICNs in the proband fibroblasts suggests that the modulated attachment of proband fibroblasts with surface and aggregates of type I collagen may disrupt nanotube organization and signaling in a manner that may limit the resolution of ICNs between fibroblasts.

Interestingly, we observed that endocytosis was also significantly reduced in the proband fibroblasts. EEA1-positive vesicles were significantly less present in proband fibroblasts compared with control fibroblasts (Fig. 6I, and J), suggesting that the formation of early endosomes was downregulated in proband. In addition, uptake of dextran-AF-488 from the culture medium was significantly lower in proband fibroblasts (Fig. 6I, and J). Thus, impaired endocytosis was responsible for the formation of fewer early endosomes in proband fibroblasts compared with controls.

We tested whether the proteotoxcity and compromised cell adhesion caused physiological stress in proband fibroblasts. Cell growth assays over a sixday period showed that growth and proliferation of proband fibroblasts were temporally slower than those of control fibroblasts (Fig. 6K, and L). To understand whether the proband fibroblasts had also undergone higher cell death over time, we performed normalized cell cycle analysis of control and proband fibroblasts. It was observed that a significant number of proband fibroblasts remained in a quiescent state (G₀ stage) of the cell cycle (Fig. 6M), which was not otherwise the case for control fibroblasts (Fig. 6M). In contrast, the proband fibroblasts had a lower number of cells in the G₁ and G₂/M phases compared with the control fibroblasts (Fig. 6M). However, the number of dead cells did not differ significantly between the control and fibro-blast cells (Fig. 6M). These observations led to the conclusion that ER stress and aggregation of proteins (including type I collagen) due to the mislocalization of MESD^D233N in the cytoplasm did not induce significant cell death but induced cellular stress that delayed cell cycle progression.

Cytosolic type I collagen aggregates are degraded by autophagy in proband fibroblasts

Proteotoxic stress enhances cellular autophagy to facilitate degradation of protein aggregates [53]. Because mislocalization of MESD^D233N leads to acute ER stress resulting in aggregation of type I collagen and other proteins, we investigated whether autophagic mechanisms contribute to reducing the burden of aggregated proteins, including type I collagen aggregates, in proband fibroblasts. A higher level of autophagy induction in proband fibroblasts was evident by an increased number of LC3B-positive autophagosomes compared with control fibroblasts (Fig. 7A, and B). Similarly, proband fibroblasts showed a significantly greater number of SQSTM1 (an autophagy receptor protein for the polyubiquitinated protein aggregates) puncta (Fig. 7A, and B), implying that the higher autophagy induction in proband fibroblasts correlated with higher aggrephagy of SQSTM1-positive ubiquitinated protein aggregates. Not only induction but also autophagy flux in autolysosome formation from autophagosomes was increased in proband fibroblasts (Fig. 7A, and B). We used the standard traffic light assay [54] using tf-LC3B (GFP RFP LC3B) transfected into control and proband fibroblasts to monitor the autophagy flux. GFP⁺ RFP⁺ LC3B puncta (autophagosome) and GFP⁻ RFP⁺ LC3B puncta (autolysosome) were also significantly increased in proband fibroblasts compared with control fibroblasts (Fig. 7A, and B). These results suggest that enhanced induction and flux of autophagy degrade aggregated proteins in proband cells.

Fig. 7. Type I collagen aggregates are cleared by autophagy in proband fibroblasts.

(A) Left and middle: control and proband fibroblast cells were immunostained for LC3B and SQSTM1. Representative confocal immunofluorescence microscopy images showing higher number of LC3B and SQSTM1 puncta, representing higher autophagy/aggrephagy, in proband fibroblasts. Right: control and proband fibroblasts were transiently transfected with RFP-GFP-LC3B (tf-LC3B). Representative confocal fluorescence microscopy images showing higher number of RFP⁺ GFP⁺ puncta (autophago-some) and RFP⁺ GFP⁻ puncta (autolysosomes), representing higher autophagy flux, in proband fibroblasts. (B) Up: quantification of LC3B and SQSTM1 puncta in control and proband fibroblasts (100 cells in each group). Down: quantification of number of RFP⁺ GFP⁺ puncta and RFP⁺ GFP⁻ puncta in control and proband fibroblasts (50 cells in each group). (C) Control and proband fibroblasts were immunostained for pro-α1(I) and SQSTM1. Representative confocal immunofluorescence microscopy images showing higher number of SQSTM1 puncta colocalized with type I collagen aggregates in proband fibroblasts. (D) Correlation coefficient of colocalization of SQSTM1 and type I collagen in control and proband fibroblasts (50 cells in each group). (E) Control and proband fibroblasts were immunostained for pro-α1(I) and LC3B. Representative confocal immunofluorescence microscopy images showing higher number of LC3B puncta colocalized with type I collagen aggregates in proband fibroblasts. (F) Correlation coefficient of colocalization of LC3B and type I collagen in control and proband fibroblasts (50 cells in each group). (G) Control and proband fibroblasts were treated with 10 μM MG132 or 1 μM bafilomycin A₁ for 48 h. Cells were immunostained for pro-α1(I). Representative confocal immunofluorescence microscopy images showing excessive accumulation of type I collagen aggregates in bafilomycin A₁ treated proband fibroblasts. (H) Quantification of percentage of MG132 and bafilomycin-A₁ treated control and proband fibroblasts showing type I collagen aggregates. (I) Control and proband fibroblasts were treated with MG132 and bafilomycin A₁ as mentioned in 7G. Representative immunoblots of LC3B, SQSTM1 and pro-α1(I) from lysate of MG132 and bafilomycin A₁ treated control and proband fibroblasts. (J) Densitometric quantification of the LC3B, SQSTM1 and pro-α1(I) bands relative to TUBA of the blots of 7I. Quantifications are shown as mean ± S.D.; P values are indicated. TUBA is the loading control in the immunoblots. Scale bar in confocal microscopy images: 5 μm. Experiments were triplicated for biological and technical replicates.

We next examined whether cytosolic type I collagen aggregates undergo autophagic clearance in proband fibroblasts. Indeed, we observed strong colocalization of SQSTM1 and LC3B with type I collagen aggregates (Fig. 7C-F), suggesting that type I collagen aggregates in proband fibroblasts were undergoing aggrephagy. In contrast, control fibroblasts did not show type I collagen aggregates, and colocalization of SQSTM1 and LC3B with pro-α1(I) was negligible (Fig. 7C-F). To confirm autophagic degradation of cytosolic type I collagen aggregates, we treated control and proband fibroblasts with inhibitors of the proteasome (MG132) and autophagy (bafilomycin A₁). The degradation of pro-α1(I) was significantly impaired in proband fibroblasts treated with bafilomycin A₁ but not in proband fibroblasts treated with MG132 compared with untreated proband fibroblasts (Fig. 7G-J). Treatment with bafilomycin A₁ and MG132 showed no significant effect on pro-α1(I) degradation in control fibroblasts compared with untreated control fibroblasts (Fig. 7G-J). These results confirm that type I collagen aggregates are degraded by autophagy [aggrephagy] in proband fibroblasts.

Discussion

Protein quality control is essential for maintaining homeostatic responses under various cellular conditions [55]. A robust network of diverse machineries and mechanisms maintains a steady state cellular proteome. Impaired expression, localization, and function of quality control maintenance proteins, such as chaperones, could lead to specific or global failure of proteostasis in diseased cells. For example, an imbalance of proteostasis is observed in many forms of OI [56], resulting in underdeveloped and brittle bones in humans [56]. While mutations in COL1A1 and COL1A2 are the main cause of certain types of OI, mutations in the ER-resident regulatory proteins, such as CRTAP, P3H1, PPIB, MESD, cause other types of OI. Essentially, mutations in the regulatory proteins eventually lead to abnormal folding and post-translational modifications of collagen and other proteins. While deregulated proteostasis can be phenotypically characterized in most OI cells, the intricate mechanisms underlying these phenomena are less understood. In this study, we have shown how the D233N mutation in the ER-retention motif of an ER chaperone, MESD, leads to its mislocalization to the cytoplasm, resulting in ER stress-associated protein aggregation in fibroblasts of an individual with OI type XX. We also show that MESD is a direct chaperone of pro-α1(I), and loss of ER localization of MESD^D233N leads to aggregated and nonsecreted type I collagen. Reduced secretion of type I collagen decreases the strength of attachment of proband fibroblasts to the matrix, a phenomenon that also leads to impaired intercellular nanotube dynamics and endocytosis. Although aggregated type I collagen is degraded by autophagy, the overall proteotoxicity and perturbed WNT signaling significantly delays the cell cycle in proband fibroblasts (Fig. 8).

Fig. 8.

Molecular mechanisms of MESD^D233N-mediated ER proteotoxicity, aggregation of type I collagen, disruption of WNT signaling, downregulated expression of bone morphogenic proteins in individual affected with osteogenesis imperfecta type XX. Mislocalization of MESD^D233N from ER to cytosol leads to loss of its ER specific chaperone function, leading to aggregation of several proteins, including type I collagen and LRP5. Aggregated type I collagen causes proteotoxic stress, whereas aggregated LRP5 (and possibly other LRPs) loses its coreceptor function for WNT. Overall, OI type XX cells manifest loss of matrix attachment, perturbed cell-to-cell communication and membrane associated events, such as endocytosis, leading to growth arrest, albeit aggregated type I collagen are cleared by autophagy to reduce the proteotoxicity in diseased cells.

To date, 12 affected individuals have been found with truncating variants in MESD. The reported pathogenic MESD variants are mainly present in exon 3 in homozygous state (9/12) and in exon 2 and 3 (3/ 12) in heterozygous state. All twelve affected individuals had fractures as a predominant feature of OI. Vertebral fractures were noted in 11 patients and, in particular, vertebral compression was noted in an 11-year-old patient [13] similar to our patient. Multiple rib fractures were noted in some of the patients examined, which we did not observe in our patient. Remarkably, prenatal fractures were observed in seven patients. Chondrolysis of the hip was noted in our patient, which was not observed in any of the reported patients with MESD mutation. Blue sclerae were noted in five affected individuals, and only one patient reported hearing impairment. None of the patients exhibited dentinogenesis imperfecta; instead, oligodontia was observed in three individuals. We did not observe any of these features in our patient. In addition, our patient has lax skin and hypermobile joints. The lethal phenotype was observed in four stillbirths that had compound heterozygous mutations in exons 2 and 3, which are thought to have completely abrogated the function of the MESD protein. In this study, the patient has a missense variant in exon 3 that causes a milder phenotype as compared with other reported affected individuals with MESD variants.

The D233N mutation does not alter the stability of MESD. Because D233 is localized at the extreme C-terminal loop of MESD, this residue is not part of a region involved in intra-molecular tertiary contacts. Therefore, the D233N mutation does not affect the structured chaperone domain of the protein. The mutations in the unstructured regions, especially in the terminal segments of the proteins, are mostly tolerable and buffered for stability and functionality, although such mutations could be disruptive in localization and intermolecular interactions. Previous reports of OI type XX -related MESD mutations are mostly clustered in the exon 3 of the gene [13, 32], resulting in truncated versions of the protein that lack the ER retention signal. As with MESD^D233N, we speculate that such MESD mutants also retain stability but lose their localization to ER.

The simulation result showed that the stability of the wild-type MESD fluctuated at times. We suggest that the transient destabilization of the MESD structure is due to its unbound conformation. While MESD is held on the ER membrane by binding to the KDEL receptor, we simulated the unbound protein in an aqueous environment. We believe that binding of MESD to the KDEL receptor could cause floating and rigidity at one end of the protein, thereby reducing the fluctuation of several residues of the protein. However, the transient destabilization of MESD fits its chaperone function. To support and maintain the folding of the substrate protein, MESD could partially open its structural region of the chaperone domain to accommodate the substrate protein. Indeed, the structural fluctuations are clustered in the helix and loop regions of the chaperone domains. This suggests that the intrinsically unstable properties of several residues of MESD are required to enable its chaperone function.

On the contrary, MESD^D233N is a more rigid structure than MESD. Although this rigidity confers more stability to the MESD^D233N structure, it could also partially impair the chaperone function of the mutant protein. Although MESD^D233N exhibits chaperone function under in vitro conditions, its activity is slightly lower than that of wild-type MESD. The D233N mutation affected the chaperone domain of MESD^D233N such that multiple residues of a helix and loops of the chaperone domain continuously interact with each other, a phenomenon that was not otherwise observed in wild-type MESD. These interactions lead to faster stabilization of MESD^D233N compared with MESD. Despite the continuous non-covalent binding of several residues of the chaperone domain of MESD^D233N, the protein still has the ability to increase the volume of its globular core. Thus, the MESD^D233N mutant retains its chaperone activity. Because MESD^D233N is mislocalized to the cytoplasm, it remains to be tested whether the protein functions as a cytosolic chaperone. It would also be interesting to find the chaperone functions of other MESD mutants reported in OI type XX.

MESD^D233N was mislocalized in the cytoplasm, with a small amount of the protein retained in ER of proband fibroblasts. Although we believe that the mutant MESD is retrotranslocated to the cytoplasm because of the loss of the ER retention signal in the protein, the amount of mutant MESD in ER could be the residual amount of protein that is newly translated and undergoing the ER translocation. In addition, the RENL sequence of MESD^D233N might still have binding ability, albeit very low, to KDEL receptors. These two phenomena could explain the very low retention of MESD^D233N in ER. It was noted that many of the previous reports of OI type XX associated truncated forms of MESD resulted in very severe and progressive disease [13, 32]. However, the individual reported in our study showed normal physiological growth and development, except for the severe OI phenotypes of vertebral compression and fractures, skin laxity and generalized osteopenia. Whereas the other reported truncated forms of MESD completely lacked ER localization, a residual amount of MESD^D233N in ER could potentially maintain a minimal chaperone function in ER. The latter event could account for a lower severity in terms of proband survival.

We note that MESD is a direct chaperone of type I collagen. The chaperone function of MESD might be particularly important in supporting the helical folding of type I collagen, or it might also maintain the folded state of type I collagen. MESD has not been shown to bind ATP or use energy generated by nucleotide hydrolysis to exert its chaperone activity. Therefore, MESD might be a weak chaperone that augments the primary folding of its substrates. On the contrary, MESD could be a chaperone that helps to maintain the properly folded state of its substrates. Although MESD contains the C-terminal ER retention signal, it also contains an N-terminal signal peptide sequence that could control the transport (secretion) of MESD from ER to the extracellular space through the Golgi body. Such a phenomenon fits with the ‘maintenance chaperone’ function of MESD for substrates such as LRPs and type I collagen that are required to be transported in properly folded form to the cell membrane and into the extracellular space. Although we have demonstrated that pro-α1(I) is a substrate of MESD, we do not rule out the possibility that other collagens could also be substrates of MESD. In addition, the exact region of type I collagen that binds to the chaperone domain of MESD remains to be identified.

We observed aggregates of LRP5 and type I collagen in the cytosol of proband fibroblasts. Although MESD^D233N was also localized in the cytoplasm of the proband cells, it is interesting that mutant MESD did not contribute to the reduction of the aggregates. Because MESD is possibly a maintenance chaperone lacking disaggregase activity, it is unable to disentangle the preformed aggregates of LRP5 and type I collagen in the cytoplasm. This apparently explains why MESD^D233N does not function as a chaperone for cytosolic protein aggregates, although it has intrinsic chaperone activity.

The chaperone function of MESD for pro-α1(I) illustrate why mutations in MESD cause severe forms of OI. The unfolded and nonmembranous LRPs in the MESD mutation conditions in previous studies and in our study suppress canonical WNT signaling, resulting in decreased formation of bone morphogenic proteins. However, previous reports of MESD mutations do not directly describe the mechanism of collagen loss in the bones of individuals affected by OI type XX. Because COL1A1 transcription is directly under the control of β-catenin [57], attenuated WNT signaling also reduces the level of cellular type I collagen. In addition to transcriptional downregulation of the COL1A1 gene, aggregated type I collagen is not secreted, and the aggregates are degraded by autophagy (aggrephagy) in the fibroblasts of the proband. Overall, the lower expression and rapid degradation of type I collagen result in a lower supply of extracellular type I collagen for bone formation. However, autophagic clearance of aggregated type I collagen could be beneficial for the survival of affected cells, as accumulation of aggregates of type I collagen and other proteins could be extremely proteotoxic to cell survival [58].

LRP-mediated WNT signaling is considered as one of the most important triggers for bone formation and development. Previous studies and our observations indicate that loss of WNT signaling coupled with LRP dysfunction may be responsible for defects in bone development in OI type XX. Because bone morphogenic proteins (BMPs) are subject to direct transcriptional regulation by WNT, deregulation of WNT signaling due to aberrant expression, stability, and localization of LRPs is considered a major contributor to pathogenicity in OI type XX. Again, we identify proteotoxicity due to aggregation of type I collagen, LRP5 (possibly other LRPs), and other proteins in OI type XX. We suggest that both mechanisms, i.e., proteotoxicity (ER stress, aggregation of type I collagen, LRP5, and other proteins) and deregulated WNT/BMP signaling, contribute simultaneously to the pathogenicity of OI type XX.

Lower extracellular type I collagen has several effects on the cell. Because the interaction of type I collagen with integrin contributes to cell attachment and dynamic membrane events [59], a significant proportion of proband fibroblasts were observed to be loosely attached to the growth surface. This phenomenon also partially explains the skin laxity of the proband. Type I collagen aggregation and loss of attachment also disrupt endocytosis and cell-to-cell communication by ICNs. However, we do not rule out the possibility that dysfunctional LRP/WNT signaling could also be contributing to these pathological phenotypes in the proband fibroblasts. The mechanistic features of these events remain to be determined at the physiological level.

Taken together, our study describes a functional network that highlights the importance of MESD chaperone function in maintaining ER proteostasis. Deregulated crosstalk of MESD with type I collagen and LRP5 due to mutations in MESD generates a proteotoxic stress and impaired WNT signaling that impairs bone development and generates other pathophysiological effects in OI type XX.

Methods and materials

Family ascertainment and diagnosis

A girl from the southern part of India was diagnosed with osteogenesis imperfecta on the basis of radiographs of vertebrae and other bones. The prenatal, birth, and developmental history of the individual was obtained at the Department of Medical Genetics, Kasturba Medical College (KMC), Manipal. Anthropometry, biochemical analysis and radiological investigations (MRI scan, PET scan and DEXA of spine) were also done at KMC, Manipal. Written informed consent was obtained from the parents of the patient for medical photography, skin biopsy, genetic analysis, and participation in the study. The study was prospectively reviewed and approved by the ethics committee of Kasturba Medical College and Kasturba Hospital (Ethical clearance number: 363/2020).

Exome sequencing and data analysis

Genomic DNA was extracted (Qiagen, 51104) from the blood cells of the proband and parents. Singleton exome sequencing of the proband was performed by massively parallel sequencing using the NovaSeq platform (Illumina Inc., USA) with a targeted average coverage of 100x. Exonic and flanking exonic-intronic regions were captured using the Agilent SureSelect CREv3 Capture Kit. The FastQC toolkit was used for initial quality assessment, followed by alignment of the raw reads on the human reference genome (GRCh38) using BWA-MEM (v0.7.15) [60]. Using the exome sequences, variant calling was performed using the established methods of GATK (v3.6) [61] for germline SNV and INDEL discovery. Allele frequencies and states were determined from the variant call format (VCF) file using BCFTOOLS (v1.3.1) and custom Perl scripts. Data were annotated using Annotate Variation (ANNOVAR) [62], with allele frequencies and states obtained from gnomAD [63], GenomeAsia [64], Singapore Genome Project [65], and our internal database of 1909 exomes. Rare variants (allele frequency < 1%) were obtained from annotated variants, and exonic and splice variants were considered for further analysis (Table S5). A detailed protocol of data processing, variant calling, derivation of allele frequencies and allele states, and annotation of variants was described in our previous study [66].

Pathogenicity scores predicted by CADD [67], REVEL [68], MCAP [69], and ClinPred were mainly used to evaluate disease-causing variants from the coding region. The variant was classified according to the ACMG criteria [70].

Cloning

Total RNA from control and proband fibroblasts was extracted (ThermoFischer Scientific [TFS], AM1924), followed by preparation of the cDNA pool by reverse transcription (TFS, 4368814) using oligo-dT primer (Bioserve). The list of clones and plasmids used in this study is given in Table 1, and specific primers were used for PCR (New England Biolabs Inc [NEB], M0531S) amplification of the cDNA of MESD and MESD^D233N, as listed in Table 2. Overall, the cloning procedure was similar to that described in our previous studies [71, 72]. Briefly, the PCR products and empty plasmid were double digested with restriction enzymes (NEB), followed by ligation of the restriction-digested PCR products and plas-mids by T4 DNA ligase (TFS, K1422). The ligated products were transformed into the ultracompetent DH5α strain of Escherichia coli. Positive colonies were selected by colony PCR (Takara Bio, RR350A), and the sequence of the clones was verified in the CDFD Research Support and Service group.

Table 1. Clones and plasmids.

Clone/Plasmid	Source	Identifier
Bacterial expression
MESD_pET21b	Present study
MESDD233N_pET21b	Present study
GFP_pET21b	[73, 74]
Luc2_pET21b	[74]
FLAG_MESD_pcDNA3.1	Present study
pcDNA3.1	ThermoFischer Scientific	V790-20
pET21b	Novagen	69741-3

Table 2. Oligonucleotides for gene cloning.

No.	Clone Cloning primers	Primer	Sequence
1	MESD_pET21b	Forward	CGAATTCCATATGCACCACCACCACCACCACGCGGCTTCCAGGTGGGCGCG CAAGGCCGTG
		Reverse	CCGCTCGAGTCACAGGTCTTCTCTTTTATTCCCAGCTCGATT
2.	MESDE262K_pET21b	Forward	CGAATTCCATATGCACCACCACCACCACCACGCGGCTTCCAGGTGGGCGCGCA AGGCCGTG
		Reverse	CCGCTCGAGTCACAGGTTTTCTCTTTTATTCCCAGCTCGATT
3.	FLAG_MESD_pcDNA3.1	Forward	CGAATTCCATATGGACTACAAAGACGATGACGACAAGGCGGCTTCCAG GTGGGCGCGCAAGGCCGTG
		Reverse	CCGCTCGAGTCACAGGTCTTCTCTTTTATTCCCAGCTCGATT

The clone for recombinant pro-α1(I) [COL1A1] was DNA sequence from TrueORF clone (Origene, RC206234), encoding the region (N-terminal Histagged) Gln300-Asp744 of COL1A1.

Recombinant protein production and purification

Recombinant MESD, MESD^D233N, Luc2 and GFP were purified separately from bacterial expression system according to the protocol described in our previous studies [71, 74]. Briefly, bacterial expression clones were transformed into the competent BL21D3 strain of Escherichia coli. The bacteria from a single colony were inoculated into the LB+ampicillin medium to produce the primary and secondary cultures. Protein production in the secondary culture was induced by maintaining 1 mM isopropyl-β-D-1-thiogalactopyranoside (IPTG, Sigma Aldrich [SA], I6758) for 12 h at 37°C. Bacterial cells were harvested (5000 g, 37°C, 15 min), followed by lysis of cells by sonication (Diagenode S. A., condition: 40 mA, 10 min, 30 s on / 40 s off cycle) in chilled lysis buffer (50 mM Tris-HCl, pH 8.0, 300 mM NaCl, 10 mM imidazole, 1 mM PMSF (SA, P7626). The precleared (12000 g, 4°C, 40 min) cell lysate was passed through the column containing Ni⁺²-NTA agarose beads (Qiagen [QA], 30250). Protein-bound beads were repeatedly washed with wash buffer (50 mM Tris-HCl, pH 8.0, 300 mM NaCl, 40 mM imidazole), followed by elution of recombinant proteins with elution buffer (50 mM Tris-HCl, pH 8.0, 300 mM NaCl, 300 mM imidazole). Proteins were dialyzed in different dialysis buffers depending on the requirements of subsequent experiments. The quality of the proteins was checked by SDS-PAGE. >98% pure proteins were used in different assays.

According to the manufacturer’s datasheet, the recombinant pro-α1(I) [COL1A1] (Origene, TP790158) was produced as secretory expressed in HEK293 cells.

Circular dichroism spectroscopy

Circular dichroism spectroscopy for far UV (190-260 nm) wavelength scanning and slow thermal unfolding (20°C-90°C) of 5 μm MESD and MESD^D233N (in 20mM NaH₂PO₄, pH 7.4, 50 mM NaF) were performed in a JASCO-810 spectropo-larimeter with a Peltier temperature controller. The Δε-values (changes in ellipticity values) of proteins during wavelength scanning were recorded under the following conditions: light path length of 0.1 cm, 100 nm/min wavelength scan speed, 1 sec response time, 1 nm data pitch, 2 nm bandwidth. In slow thermal unfolding experiments, proteins were heated at a rate of 1°C/min and Δε₂₂₂ values (changes in ellipticity values at 222 nm) were recorded simultaneously. Secondary structure components were analyzed using the CDSSTR program in the Dichro-Web platform. Melting temperatures (T_m) of proteins were calculated using the Spectra Manager denaturation analysis program.

Chaperone assays

Luciferase reactivation assay

100 nM recombinant Luc2 was maintained alone or mixed with 1μM MESD or MESD^D233N or GFP in a reaction buffer (20 mM Tris-Cl (pH: 8), 50 mM NaCl, 10 mM MgCl₂, 10 mM DTT). The mixtures were either held at 37°C for 30 min or heated to 60°C for 30 min and then cooled on ice for 10 min. Luminescence in the reaction mixtures was generated using the Luciferase Assay System (Promega Inc., E1500) and luminescence signals were recorded using a Perkin Elmer LS-55 luminescence spectrometer with an integration time of 10 sec. Luc2 activity was expressed in relative light units (RLU).

Turbidity assay

Recombinant 500 ng/ml pro-α1(I) [COL1A1] (Q300-D744 region) (Origene, TP790158) was maintained alone or mixed with 1 mM MESD or MESD^D233N or GFP in a reaction buffer (20 mM Tris-Cl (pH: 8), 50 mM NaCl, 10 mM MgCl₂, 10 mM DTT). The mixtures were either held at 37°C for 30 min or heated to 65°C for 30 min and then cooled on ice for 10 min. The turbidity of 30 μl reactions was measured in a 96-well plate using 600 nm light in the Varioskan LUX Multimode Microplate Reader (TFS).

Cell culture

Skin fibroblasts were obtained from the proband and three age-matched control individuals (named as Control-1, Control-2, and Control-3) with informed consent and approval from the ethics committee. Cells were cultured in advanced DMEM (TFS, 12491023) supplemented with 2 mM L-glutamine (TFS, 25030164), 10% fetal bovine serum (TFS, 26140095), and 1x penicillin-streptomycin (TFS, 11548876). Cells were maintained optimally in a humidified static incubator at 37°C and 5% CO₂[75, 76].

Transfection of DNA clones was performed with lipofectamine2000 (TFS, 11668019) and opti- MEM (TFS, 31985062) as described in our earlier study [77]. Transfection of siRNA (listed in Table 3) was performed with lipofectamine 3000 (TFS, L3000001) according to the manufacturer’s protocol.

Table 3. siRNAs.

Target of siRNA	Source	Identifier
Control	Sigma Aldrich	SIC001
MESD	Santa Cruz Biotechnology	sc-90278

Cells were incubated with 1 mg/ml Dextran-Alexa fluor 488 (MW: 3000) for 6 hours at optimal growth conditions.

Cells were treated with 10 μm MG132 or 1 μm bafilomycin A₁ for 48 h.

RNA isolation and quantitative PCR

Total RNA was extracted from cultured fibroblast cells using the RiboPure™ RNA Purification Kit (TFS, AM1924) according to the manufacturer’s protocol. The cDNA pool was prepared from RNA by reverse transcription (TFS, 11756050) using random hexamers, and quantitative PCR (qPCR) was performed using SYBR green PCR master mix (TFS, 4309155) with forward and reverse primers as indicated in Table 4. The qPCR data were analyzed using QuantStudio™ Design and Analysis software. Relative quantification of target genes was normalized to the expression of the endogenous control gene (GAPDH).

Table 4. Oligonucleotides for qPCR.

No.	Clone	Primer	Sequence
4	MESD	Forward Reverse	AATTACGAGCCTCTGGCAGG TCAGCACACCTGTCTTGACC
5	MESD-ORF	Forward   Reverse	GCGGCTTCCAGGTGGGCGC GCAAGGCCGTG CAGGTCTTCTCTTTTATTCCC AGCTCGATT
	LRP4	Forward Reverse	TGACTCGGATGAGCAGGACT ACTCCTTGTCGGAGCACTTG
6	LRP5	Forward Reverse	AGTGCCCTCTACTCACCCAT TGTAGAAAGGCTCGCTTGGG
7	LRP6	Forward Reverse	TTGGGCTCAACCGTGAAGTT AATGATGGTGCGGTTTTGGC
8	COL1A1	Forward Reverse	GGAGAGGAAGGAAAGCGAGG AGGAGAACCACGTTCACCAG
9	COL1A2	Forward Reverse	CTGCTGGAAGTCGTGGTGAT ACGAAGCCCTTCTTTCCCAG
10	BMP2	Forward Reverse	GCTAGACCTGTATCGCAGGC CTCCGGGTTGTTTTCCCACT
11	BMP4	Forward Reverse	GGAGCTTCCACCACGAAGAA GGAAGCCCCTTTCCCAATCA
12	EIF2AK3	Forward Reverse	TCGCCAATGGGATAGTGACG AAATCCGGCTCTCGTTTCCA
13	P4HB	Forward Reverse	CTGCAGAGTCCTTGGTGGAG ACCCCATCTTTGTCGAGCTG
14	HSPA5	Forward Reverse	CCCGTGGCATAAACCCAGAT GGTCATGACACCTCCCACAG
15	CANX	Forward Reverse	CGTACCTGATCCAGACGCAG AGGAGGCTTCCATTTGCCTT
16	CALR	Forward Reverse	AGGAGCAGTTTCTGGACGGA GCACCACCAGCGTCTGGCCTT
17	GAPDH	Forward Reverse	ACCTGCCAAATATGATGAC TCATACCAGGAAATGAGCTT

Enrichment of endoplasmic reticulum

The endoplasmic reticulum from the lysate of control and proband fibroblast cells was enriched by using Minute™ ER Enrichment Kit (Invent Biotechnologies Inc., ER-036) by following manufacturer’s protocol.

Immunoprecipitation and immunoblotting

Immunoprecipitation

MESD was immunoprecipitated from the lysate of control fibroblast cells. Cells were lysed in non-dena-turing lysis buffer (20 mM Tris HCl pH: 8, 137 mM NaCl, 10% glycerol, 1% Nonidet P-40 (SA, NP40S), 2 mM EDTA) at 4°C for 30 min. 2 mg of total protein in the cell lysate was incubated with the anti-MESD antibody or control IgG, followed by non-denaturing immunoprecipitation using the Crosslink Magnetic IP/co-IP kit (TFS, 88805).

Immunoblotting

60-80 μg of protein of the cell lysate was loaded onto a 12% gel (SDS-PAGE), and proteins were then transferred from the gel to a polyvinylidene fluoride (PVDF) membrane (Amersham Hybond P 0.45; GE Healthcare Life Sciences, 10600023). The membrane was blocked with 5% skim milk (in Tris buffer saline [TBS], TFS, 28358), followed by sequential application of primary and secondary antibodies as indicated in Table 5. The membrane was intermittently washed with TBST (TBS containing 0.1% Tween20) solution. Chemiluminescence signals were generated using SuperSignal west femto maximum sensitivity substrate (TFS, 34094) and signals were measured using ChemiDoc XRS+ Imaging System (Bio-Rad). The density of protein bands in the blots was quantified using ImageJ2 software.

Table 5. Antibodies.

Primary antibodies Target protein	Host	Supplier	Catalogue number
BMP2	Rabbit	Abcam	ab214821
CANX	Mouse	ThermoFischer Scientific	MA3-027
COL1A1 [pro-α1(I)]	Rabbit	Cloud-Cone Corp.	PAA350Hu02
COL1A2 [pro-α2(I)]	Rabbit	ThermoFischer Scientific	PA5-106555
DNAJB11	Rabbit	Sigma Aldrich	HPA010814
EEA1	Rabbit	Abcam	ab50313
FZD1	Rat	R&D systems	MAB11201-100
HSPA5	Rabbit	ThermoFischer Scientific	PA1-014A
HSP90B	Rabbit	Abcam	Ab32568
IgG (serum)	Rabbit	Sigma Aldrich	12-370
ITGB1	Mouse	ThermoFischer Scientific	14-0299-82
LRP5	Rabbit	Abcam	ab203201
LRP5	Mouse	ThermoFischer Scientific	MA5-17113
LC3B	Mouse	ThermoFischer Scientific	MA5-37852
MESD	Rabbit	Sigma-Aldrich	HPA039414
PAX1	Rabbit	ThermoFischer Scientific	BS-1199R
PDI	Mouse	ThermoFischer Scientific	MA3-019
SQSTM1	Mouse	Sigma Aldrich	WH0008878M1
TUBA	Mouse	Sigma Aldrich	T6199
Ubiquitin	Rabbit	ThermoFischer Scientific	PA3-16717
Secondary antibodies (Immunoblotting)
anti-rabbit IgG (whole molecule) peroxidase antibody	Goat	Sigma-Aldrich	A0545
anti-mouse IgG (whole molecule) peroxidase antibody	Rabbit	Sigma-Aldrich	A9044
Secondary antibodies (Immunocytochemistry)
anti-Mouse IgG (H+L)-Alexa Fluor Plus 488	Goat	ThermoFischer Scientific	A-11029
anti-Mouse IgG (H+L)-Alexa Fluor Plus 555	Goat	ThermoFischer Scientific	A32727
anti-Rabbit IgG (H+L)-Alexa Fluor Plus 488	Goat	ThermoFischer Scientific	A-11034
anti-Rabbit IgG (H+L)-Alexa Fluor Plus 555	Goat	ThermoFischer Scientific	A32732
anti-Rat IgG (H+L)-Alexa Fluor Plus 555	Goat	ThermoFischer Scientific	A48263

Immunocytochemistry and fluorescence

Fibroblast cells were grown on glass coverslips (Bluestar). Cells were washed with phosphate buffer saline (PBS, pH 7.4; TFS, 10010023) and fixed with 4% paraformaldehyde (SA, 158127) (in PBS) solution. Cells were permeabilized with 0.2% Triton X-100 (SA, T8787) (in PBS) and then blocked with 1% bovine serum albumin solution (BSA, in PBS) (SA, B6917). The primary and secondary antibodies, listed in Table 5, were applied sequentially to the cells, with intermediate washing with PBS and blocking with 1% BSA. Cells were mounted on glass slides with prolong antifade gold with or without DAPI (TFS, P36931; P10144).

Adherent cells were stained with Proteostat (Enzo lifesciences, ENZ-51035-0025) according to the manufacturer’s protocol.

Fluorescence images were acquired using LSM700 confocal laser scanning microscope (Carl Zeiss) with a 63x Plan Apochromat/1.4 NA oil/DIC M27 objective. Images were processed using Zen-Lite 2010 software (Carl Zeiss).

The Coloc 2 program in FIJI was used to measure the Pearson correlation coefficient of colocalization of different proteins.

The number of protein aggregates, LC3B puncta (autophagosome), RFP-GFP-LC3B puncta (auto-phagosomes and autolysosomes), and SQSTM1 puncta (ubiquitinated protein aggregates) were counted manually [78].

Cell growth assay

For the growth and proliferation assay, 20,000 control or proband fibroblast cells were seeded in each well of a six-well plate. Cells were fixed with 4% formaldehyde (in PBS) on day 6 and stained with 0.1% crystal violet (SA, C0775). Cells were also counted every other day using standard trypan blue staining.

Cell cycle analysis by fluorescence activated cell sorting

Control and proband fibroblast cells were grown for 10 days. Cells were detached from the growth surface with trypsin-EDTA solution (TFS, 15400054) and then washed with PBS. Cells were resuspended in Flow Cytometry Staining Buffer (R&D systems, FC001) containing 1 μ g/ml propidium iodide (SA, P4170). Cells were kept in this solution for 30 min at 37°C in the dark with constant mixing. The DNA content of 50,000 control and proband cells was counted using the BD-FACSARIA-III instrument (Becton Dickenson Biosciences), and the results were analyzed using the BD-FACSDIVA software.

Molecular dynamics simulation

Molecular dynamics simulations (MDS) of MESD and MESD^D233N were done in GROMACS/2018.8 [79]. GROMACS readable coordinate files of proteins were generated using OPLS-AA/L [80] force-field, and the simulation environment was created by TIP4P [81] water model. Proteins were placed in a 40 Å virtual cuboidal box with a minimum distance of 1.0 nm between the edges of the protein and the unit cell. A three-dimensional periodic boundary condition was used. Charges of solvated proteins were neutralized by 0.15 M Na⁺Cl^-.

Simulation system was relaxed, and steric clashes were removed using steepest descent in conjugate gradient energy minimization. The system was equilibrated in a constant NVT (particle number-volume-temperature) simulation at 300 K using V-rescale temperature coupling method [82], followed by an NPT (particle number-pressure-temperature) simulation at 1.0 bar using the isotropic Parrinello-Rahman pressure coupling scheme [83]. These equilibration simulations were done using the Leap-Frog integrator with a Molecular timestep of 2 fs for 1 ns, LINCS constraints for bound parameters [84], Verlet cut-off for unbound parameters [85], and particle mesh Ewald for long-range electrostatics [86]. Unconstrained full simulation of the equilibrated system was performed for 50 ns with a time step of 2 fs, trajectory values recorded every 10 ps.

Post-processing of the simulation system was done using ‘gmx trjconv’ to correct the diffusion protein coordinates and for centering in the unit cell.

Trajectories were analyzed using ‘gmx rms’, ‘gmx rmsf’, ‘gmx hbond’ and ‘gmx gyrate’ algorithms to measure the RMSD, RMSF, hydrogen bonds and radius of gyration of protein respectively. Secondary structure of the proteins at different time points were analyzed using ‘gmx do_dssp’ with Dictionary Secondary Structure of Protein (DSSP) program [87]. Structures were extracted from the trajectory files using ‘gmx trjconv’.

Bioinformatics analysis

Multiple sequence alignment of protein sequences was done in Clustal Omega [88].

Position specific sequence conservation was analyzed in Gremlin [89].

The probable pathogenicity associated with the D233N mutation of MESD was predicted in Poly-phen-2 [90].

Changes in difference of free energy of folding (ΔΔG) due to substitution of D233 with other amino acids were calculated by energy functions in FoldX [91].

Statistical analysis

Data are presented as mean ± standard deviation. Statistically significant difference between means of groups were analyzed by two-way ANOVA or two-tailed, homoscedastic student’s t-test. P value less than 0.05 was considered as statistically significant different.

Graphics

The graphics were made in Adobe Illustrator.

Data availability

All materials, data, and protocols of this study are available from the corresponding authors upon request.

Data Availability

All the data related to the experiments mentioned here are available from the corresponding authors on request.