Extracellular Matrix Disorders: Molecular Mechanisms and Treatment Targets

Osteoblasts produce and secrete massive amounts of procollagen I, presenting a unique challenge for protein quality control and trafficking. This procollagen is a precursor of type I collagen. We discovered that, above 35°C, the conformation of natively folded human procollagen I is less favorable than the unfolded one. To fold and stabilize the native conformation at body temperature, cells must use heat shock protein 47 (HSP47) and other specialized endoplasmic reticulum (ER) chaperones. As a result, about 15% of procollagen is misfolded and must be degraded even under normal conditions, activating cell stress–response pathways and forcing osteoblasts to always function in a high-stress mode. We found that one of the key pathophysiological mechanisms of OI and other disorders of collagen metabolism is excessive cell stress caused by excessive accumulation of misfolded procollagen in the ER.

The most common hereditary cause of excessive procollagen misfolding is a substitution of Gly in the obligatory (Gly-X-Y)n sequence that is characteristic of all collagens. Such substitutions are responsible for over 80% of severe OI cases, the majority of EDS cases, and a variety of other syndromes. Their pathophysiology is one of the key topics of our studies. For instance, our studies on OI patients with over 50 different Gly substitutions revealed several structural regions within type I collagen where such mutations might be responsible for distinct OI phenotypes. One such region is the first 85–90 amino acids at the N-terminal end of the triple helix, mutations in which prevent normal procollagen processing into collagen and cause a distinct, mixed OI/EDS pathology.

While we focus mostly on hereditary disorders affecting development, excessive procollagen misfolding may also occur upon changes in the ER associated with environmental factors, inflammation, aging, etc., and is likely to contribute to fibrosis, cancer, age-related osteoporosis, and many other common ailments. Nonetheless, the pathophysiology of misfolded procollagen accumulation in the ER remains poorly understood.

To understand this pathophysiology, we are investigating how misfolded procollagen is recognized by the cell, why it accumulates in the ER, how this accumulation affects cellular function, and how the misfolded molecules are degraded. In one approach, we visualize these processes by imaging fluorescently tagged procollagen in live cells. To facilitate these studies, we created several plasmids enabling transient expression of fluorescent procollagen in mouse and human cells. We also created several cell lines in which the fluorescent tag was inserted into the gene encoding endogenous pro-alpha2(I) chain. These new tools for live-cell imaging of procollagen have already been shared with dozens of laboratories around the world.

Imaging of fluorescent procollagen enabled us to demonstrate that normally folded procollagen is transported from the ER exit sites (ERESs) to Golgi in carriers coated by coat complex I (COPI) rather than coat complex II (COPII) proteins and that HSP47 is released before procollagen exits ERES into this secretory pathway, in contrast to widely held beliefs. We found that vesicle-like COPII–coated structures containing procollagen and HSP47, previously interpreted as secretory carriers, were instead dilated, and that disrupted ERESs containing misfolded procollagen, which were still attached to the ER, were being engulfed by lysosomes, or were being degraded inside the lysosomes. This study delineated a new ERES microautophagy pathway, which was subsequently confirmed for other secretory cargo and studied in greater details by the lab of our collaborator Jennifer Lippincott-Schwartz. In the last year, our studies of a new mouse model of the Cole-Carpenter syndrome (a very rare, severe bone fragility disorder) revealed that the ERES microautophagy of misfolded procollagen is mediated by SEC24D, an isoform of the SEC24 cargo receptor protein in the COPII coat. Together, all these observations explain otherwise puzzling distinctions in skeletal pathologies caused by mutations in HSP47, COPI, and COPII coat proteins. Rerouting of ERES–loaded cargo from the secretory to degradative pathway via ERES microautophagy may also have wide clinical and translation implications beyond skeletal disorders, as demonstrated by Lippincott-Schwartz' lab studies of secretory cargos other than procollagen is less specialized cells.

In another approach, we are investigating the cell-stress response to procollagen misfolding caused by a Gly610-to-Cys substitution in the triple-helical region of pro-alpha2(I). We helped develop a mouse model of OI with this mutation (G610C mouse), which mimics the pathology found in a large group of patients with the same mutation. Our study of this model demonstrated that misfolding and accumulation of mutant procollagen in the ER of osteoblasts causes ER disruption, resulting in cell stress and malfunction. We are therefore investigating the mechanism of this stress and its role in pathology by altering how the cells adapt to it and by examining cell-stress response pathways activated by the mutation. For instance, we reduced autophagic degradation of ERESs containing misfolded procollagen by osteoblast-specific knockout of ATG5 (autophagy-related factor 5). Increased bone pathology, caused by the resulting additional accumulation of misfolded procollagen in the ER, confirmed our hypothesis that osteoblast cell stress and the malfunction associated with such accumulation play a significant role in OI pathophysiology. More detailed analysis of the ATG5 knockout effects confirmed that ERES microautophagy is the primary pathway of misfolded procollagen degradation in G610C osteoblasts. Furthermore, more recent studies revealed that accumulation of misfolded G610C procollagen in osteoblast ER activates transcription of integrated stress-response genes (e.g., Ddit3, Eif4ebp1, and Nupr1) but not of the canonical ER stress transducers Atf4 and Hspa5, suggesting non-canonical cell stress and identifying its transducers as Atf5 and Hspa9 paralogs of Atf4 and Hspa5. We validated these findings by bulk, single-cell, and spatially resolved RNA sequencing, as well as by in situ RNA hybridization and Western blotting.

By combining the live-cell imaging, genetic, and biochemical analysis, we found that misfolding of procollagen in the ER of G610C osteoblasts activates the mitochondrial arm of the integrated cell-stress response (ISR) rather than the canonical unfolded-protein response (UPR). Misfolded G610C molecules are not recognized and retained in the ER lumen by quality-control chaperones. Instead, they are retained at ERESs, blocking the exit of all secretory proteins from the ER into the secretory pathway. The resulting ER overcrowding activates the ISR without a UPR as a result of disruption of ER–mitochondria contacts. The ISR is sufficient to stabilize G610C osteoblasts in a less efficient yet functional steady state, in which procollagen synthesis is reduced and degradation is increased. However, the same ISR may not be sufficient to prevent more severe ER disruption in other cells or other mutations, which may then cause misfolding of globular proteins and a concomitant secondary UPR. For instance, we found that hypertrophic chondrocytes in the growth plate of the same G610C animals undergo a secondary UPR, which blocks their transition into osteoblasts and thereby limits longitudinal bone growth. Our discovery of mitochondrial involvement not only identified a new response pathway to protein misfolding in the ER but also opened up a new research direction. We are currently investigating how the ER–mitochondria contacts are disrupted and how the resulting mitochondrial dysfunction affects the cells.

Our observations suggested that the pathology associated with procollagen misfolding may be partially reversed by improving cell adaptation to misfolded procollagen accumulation in the ER, thereby improving cellular function. Although this would not eliminate the detrimental effects of secreted mutant collagen, pharmacological treatment of cell malfunction is the most realistic short-term strategy for targeting the causes rather than the effect of bone pathology in OI. It is also likely to be a good long-term strategy for the treatment of cell malfunction caused by procollagen misfolding in cases that do not involve pathogenic mutations.

To pursue this strategy, we are examining the effects of enhancing the natural ability of cells to remove and degrade misfolded molecules via autophagy, which is the simplest way to prevent their pathogenic accumulation in the ER. Our studies on autophagy enhancement by a low-protein diet or intermittent fasting in G610C mice revealed improved osteoblast differentiation and function, resulting in better bone quality, but we also observed stunted animal growth. We are thus evaluating other approaches that might provide the same benefits of autophagy enhancement without long-term nutrient deficiency.

In particular, we are testing drugs known to reduce ER disruption by enhancing the secretion and autophagy of misfolded proteins (e.g., 4-phenylbutyrate [4PBA]) and drugs known to reduce the impact of accumulating misfolded proteins on overall protein synthesis (e.g., integrated stress response inhibitor [ISRIB]). We found that 4PBA reduces bone pathology in a zebrafish model of OI and in G610C mice. However, although low-dose 4PBA treatment improved the function of hypertrophic chondrocytes and their conversion into osteoblasts in such mice, it did not improve the function of osteoblasts, probably because it alleviated secondary UPR more efficiently than the primary non-canonical cell stress. At the same time, a higher-dosage treatment is challenging because 4PBA is very rapidly metabolized and therefore difficult to deliver in a sustainable fashion to bone cells. We are therefore exploring other drugs, alternative delivery methods, and other approaches.

A key issue in monitoring treatment efficiency in animal models, as well as in general diagnostic analysis of bone pathology, is the lack of reliable methods with which to characterize the function of bone cells. Traditional histopathology relies on subjective analysis of bone-cell morphology in tissue sections, which is not a reliable indicator of cell function. Over the last several years, we developed a new approach to visualizing and quantifying mRNA expression in individual cells in bone sections. The approach enables objective and reliable cell identification as well as in situ characterization of cell differentiation and function. Based on the interest of bone histomorphometry experts in learning this approach, we hope that it will soon be adapted not only for research but also for clinical practice.

The other direction of our clinical and translational studies is characterization of pathophysiology mechanisms in cells from patients with newly discovered skeletal dysplasias, as well as development and analysis of mouse models of these disorders. We investigated mutations in cartilage-associated protein (CRTAP), prolyl-3-hydrohylase (P3H1), cyclophilin B (CYPB), the collagen-binding molecular chaperone FKBP65, the signaling protein WNT1, the ER–membrane ion channel TRICB, the Golgi-membrane metalloprotease S2P, the transmembrane anterior-posterior transformation protein 1 (TAPT1), or collagen prolyl-4-hydroxylase 1 (P4H1). Our studies suggested that the CRTAP/P3H1/CYPB complex functions as a procollagen chaperone. FKBP65 affects post-translational modification of procollagen. TRICB deficiency reduces thermal stability of type I procollagen, suggesting dysregulation of collagen chaperones in the ER or direct involvement of TRICB in procollagen folding. Pathogenic effects of mutations in the transmembrane protein TAPT1 or in site-2 metalloprotease (S2P) might not be directly related to disruptions in synthesis, folding, or trafficking of procollagen chains. As expected, P4H1 deficiency causes under-hydroxylation of proline residues and reduced thermal stability of procollagen. Surprisingly, however, this deficiency does not affect the procollagen folding or secretion rates, apparently due to stronger HSP47 binding to triple helices with unhydroxylated prolines.

In the past year, we developed and investigated a mouse model of Cole-Carpenter syndrome 2 caused by deficiency in a COPII complex protein SEC24D (a severe skeletal dysplasia characterized by OI–like bone pathology and distinct facial dysmorphism). We found that the SEC24D deficiency prevents rerouting of misfolded procollagen from ERESs to degradation via ERES microautophagy, resulting in ERES blockage, accumulation of secretory proteins in the ER, ER disruption, and severe malfunction of osteoblasts, odontoblasts, chondrocytes, and other cells producing large amounts of type I and type II collagens. We explained why mutations in a protein sulfide isomerase P4HB in the ER lumen and a COPII protein complex subunit SEC24D outside the ER cause phenotypically similar Cole-Carpenter syndromes 1 and 2, respectively. In contrast, mutations in P4HA (a family of ER chaperones that form complexes with P4HB) and mutations in SEC23 and SEC31 (families of COPII proteins that form complexes with SEC24D) produce completely different phenotypes. Our observations suggest possible approaches to therapeutic targeting of the Cole-Carpenter syndrome as well as identify distinct biological functions of different SEC24 isoforms and different protein disulfide isomerases in the ER.

In the last few years, a large part of our efforts has been focused on understanding the causes of lung-tissue pathology in patients and mouse models of OI, given that cardio-pulmonary complications are the main cause of OI mortality. Using the G610C mouse model, we demonstrated severe under-development (hypoplasia) of saccular structures resulting from deficient lung inflation with amniotic fluid during fetal breathing. We identified diaphragm ECM pathology caused by malfunction of central tendon fibroblasts as a key contributing factor, and we established a collaboration with clinical researchers to translate our findings into clinical applications. Our mouse studies indicate that inexpensive, noninvasive ultrasound examination during the last trimester, e.g., as part of biophysical profiling, should reveal deficient lung inflation during fetal breathing and thereby predict perinatal lung hypoplasia in OI and other skeletal disorders. Such early, in utero diagnostics would enable early lung hypoplasia treatment.

In addition to developmental skeletal dysplasias, we also characterized collagen metabolism pathology in fibromas and tumors, e.g., abnormal collagen composition of uterine fibroids and the potential role of type I collagen homotrimers in cancer. The normal isoform of type I collagen is a heterotrimer of two alpha1(I) chains and one alpha2(I) chain, as stated in the Introduction. However, homotrimers of three alpha1(I) chains are produced in carcinomas and some fibrotic tissues. We found the homotrimers to be at least 5–10 times more resistant to cleavage by all mammalian collagenases than the heterotrimers, and we determined the molecular mechanism of this resistance. Our studies suggested that cancer cells might utilize this collagen isoform to build collagenase-resistant tracks, thus supporting invasion through stroma of lower resistance.

We also investigated bone pathology and tumors caused by defects in cAMP signaling, e.g., those associated with mutation in protein kinase A (PKA), which is a key enzyme in the cAMP signaling pathway. While we did not detect the type I collagen homotrimer synthesis we expected, we identified and characterized abnormal organization and mineralization of bone matrix as well as novel bone structures in mice with knockouts of various PKA subunits. For instance, we observed free-standing cylindrical bone spicules with an osteon-like organization of lamellae and osteocytes but an inverted mineralization pattern, a highly mineralized central core, and diminishing mineralization away from the central core. We assisted clinical researchers in characterizing abnormal osteoblast maturation, the role of an abnormal inflammatory response, and effects of anti-inflammatory drug treatments in such animals. Improved understanding of bone tumors caused by PKA deficiencies may not only clarify the role of cAMP signaling but may also suggest new approaches to therapeutic manipulation of bone formation in skeletal dysplasias.

To understand how collagen metabolism disorders cause cell stress and pathology, we are developing methods for characterizing cellular function in relation to ECM composition and structure. In particular, we developed high-definition (HD) infrared imaging and Raman micro-spectroscopic methods, improving spectral reproducibility by up to two orders of magnitude, based on thermomechanical stabilization of the light path through tissue. The technology enabled us to uncover the causes of progressive cartilage degradation in a mouse model of diastrophic dysplasia caused by mutations in the SLC26A2 sulfate transporter gene. It was essential for the analysis of abnormal collagen matrix deposition by CRTAP– and FKBP65–deficient cells. We used it in studies of bone structure and mineralization in the mouse models of the OI and PKA deficiencies described above and for establishing new approaches to characterizing bone formation in vitro. Beyond non-destructive characterization of ECM composition and structure in many of our studies, the technology proved useful for researchers from other NICHD and NIH laboratories. For instance, we assisted NIBIB scientists in characterizing a functionalized carbon-nanotube approach to the delivery of anticancer agents into cells that overexpress hyaluronate receptors.

We are presently working on integrating this HD infrared/Raman and histological ECM imaging with cellular-function imaging based on in situ mRNA sequencing and fluorescent in situ mRNA hybridization. We have combined all these modalities for demonstrating bona fide bone formation in cell culture as well as for studies of bone development and pathology both in culture and in histological tissue sections. We are currently focusing on implementing additional modalities of imaging differential gene expression based on commercial high-resolution spatial transcriptomics assays (e.g., Visium HD and Xenium from 10X Genomics). We plan to continue integrating rapidly advancing commercial technologies with our custom experimental methodologies, statistical analysis tools, and computer algorithms to improve the analytical and diagnostic power of combined structural and functional imaging of histological tissue sections.

Our most recent and important development in this area is a statistical theory and methods for analysis of single-cell and spatial transcriptomics data. We found that all commonly used methods produce significant artifacts. We have resolved the underlying problems by adapting an approach conceptually similar to the one that has been the gold standard in statistical analysis of cluster-randomized clinical trials. We derived the corresponding methods for weighted statistical averaging and testing that account for intracluster correlations in composition of transcripts within different cells and that make no assumptions about the data beyond random and independent sampling. Multiple tests show clear benefits of our approach and major differences in its results vs. commonly used methods. Considering that scRNA-Seq and spRNA-Seq assays are widely used in cell biology studies and are beginning to enter clinical diagnostics, we have publicly released ready-to-use computer algorithms for implementing our approach.