Introduction

Ca2+ store functions organized in the sarco/endoplasmic reticulum (SR/ER) are essential for cellular homeostasis. Store Ca2+ fluxes are predominantly mediated by SR/ER Ca2+-ATPase pumps and Ca2+ release channels, namely inositol trisphosphate (IP3R) and ryanodine receptors (RyR), and supposed to accompany counterion fluxes to balance charge, osmolality and pH between the SR/ER lumen and the cytoplasm [1,2,3]. Although several ionic fluxes, such as K+, Cl- and H+ currents, have been detected in SR/ER membranes [4,5,6], the molecular basis of the counterionic fluxes is largely unknown. It is reasonably proposed that counterion species and their current densities during Ca2+ uptake and release are divergent among cell types, because SR/ER-resident channels and transporters may be differentially expressed to generate various counterion fluxes. We previously identified two trimeric intracellular cation channel subtypes, namely TRIC-A and TRIC-B, both of which are distributed to the SR/ER and nuclear membranes and form ionic channels that are predominantly permeable to monovalent cations in planner lipid bilayer membranes [7,8,9]. The unique three-dimensional structures of TRIC channels have been elucidated, and each subunit possesses its own ion-conducting pore equipped with phospholipids under intracellular conditions [10,11,12]. In knockout mice lacking TRIC subtypes, cell types developing functional defects commonly exhibit impaired Ca2+ release and store Ca2+ overloading. For example, Tric-a-knockout mice exhibit impaired RyR-mediated Ca2+ release in muscle cells [13, 14]. In contrast, IP3R-mediated Ca2+ release is compromised in alveolar epithelial cells and osteoblasts from Tric-b-knockout mice [15, 16]. These observations indicate that TRIC channels generate counter-K+ fluxes at least in part to facilitate store Ca2+ release in various cells. Additionally, more recent observations suggest that TRIC-A directly interacts with and activates RyR channels in addition to providing a counterion current [17].

Long bones develop through the biological process called endochondral ossification, and the initial stage of this process is cartilage formation [18]. During the cartilage formation in early embryogenesis, chondroblasts undergo differentiation and organize morphologically distinct zones, each of which contains homogeneous chondrocytes specified by morphological characteristics. Round-shaped chondrocytes propagate in the epiphyseal end and produce type II collagen COL2A1. Then, the round chondrocytes structurally change into flat chondrocytes that proliferate to arrange characteristic columnar arrays. The columnar chondrocytes subsequently differentiate into hypertrophic chondrocytes expressing type X collagen, and finally swell and undergo apoptosis. Finally, the region scattered with the generated apoptotic bodies is gradually replaced by trabecular bone through the action of osteoclasts and osteoblasts.

Osteogenesis imperfecta (OI) is a genetic disease characterized by repeated bone fractures due to reduced bone mass [19]. The majority of OI cases result from defective type I collagen; structural mutations and altered posttranslational modifications lead to its insufficient synthesis, unfolding, mistrafficking, poor secretion and disincorporation into the bone matrix. OI-causing mutations are also found in collagen-unrelated genes, such as the osteoblast-specific transcription factor Osterix and the osteoblast-specific transmembrane protein IFITM5. Furthermore, homozygous deletion mutations in the TRIC-B (also referred to as TMEM38B) locus have been identified in several OI pedigrees [20,21,22,23]; the critical mutations are all supposed to produce defective TRIC-B channels in the reported patients. Indeed, Tric-b-knockout mice develop an OI-like phenotype, and the Tric-b deficiency induces store Ca2+ overloading due to compromised Ca2+ release in osteoblasts [16]. The experimental evidence indicates that the pro-collagen processing in the ER is likely deranged by the defective store Ca2+ handling of osteoblasts, leading to insufficient bone matrix and resulting in poor bone mineralization in OI patients with TRIC-B mutations [16, 24]. However, the impact of the Tric-b deficiency has not been examined in growth plates contributing to long-bone outgrowth. In this report, we investigated the irregular cell death and impaired extracellular matrix (ECM) synthesis observed in Tric-b-knockout growth plate chondrocytes.

Results

Dead cells in Tric-b-knockout growth plates

We first explored the impact of Tric-b deficiency on growth plate chondrocytes by histologically analyzing developing bones from the Tric-b-knockout mice just before birth (E18.5). Consistent with the previous observation that the body size of Tric-b-knockout mice is slightly decreased when compared with that of wild-type littermates [15], the femoral length was reduced in the knockout mice (Fig. S1A). In the longitudinal femoral sections prepared from the knockout mice, round, columnar and hypertrophic chondrocyte zones were regularly formed to constitute the developing growth plates, while their respective areas were decreased when compared with wild-type controls (Fig. 1A). The knockout and wild-type growth plate chondrocytes were roughly similar in morphology, and cell densities normalized to each zonal area were also similar between the genotypes. However, alcian blue staining for acidic proteoglycans detected that the ECM portion of the round chondrocyte zone was significantly decreased in Tric-b-knockout growth plates.

Fig. 1: Insufficient ECM organization and aberrant cell death in Tric-b-knockout growth plates.
figure 1

(A) Zonal area, cell density and ECM occupancy in E18.5 Tric-b-knockout growth plates. Alcian blue-stained femoral growth plate sectional images (left images; scale bar, 200 μm). Round (R), columnar (C) and hypertrophic (H) chondrocyte zones are indicated by white dashed lines. Typical round (R) and columnar (C) chondrocytes are also shown in the magnified images (scale bar, 20 µm). The zonal area size, cell density normalized to zonal size and alcian blue-positive portion in total growth plate and individual cell zones are statistically analyzed between the knockout and wild-type growth plates (right bar graphs). (B) Photo-microscopic images of dilated dead cells (left panels), electron-microscopic images of apparently normal cells (middle panels) and dilated dead cells (right panels) observed in E18.5 Tric-b-knockout round and columnar cell zones. Scale bars, 20 μm. The frequencies of dilated dead round and columnar chondrocytes detected in the knockout growth plates are shown in the bar graph. In the bar graphs, the data are presented as the mean ± SEM., and the numbers of mice examined are shown in parentheses. Statistical differences between the genotypes are indicated by asterisks (*p < 0.05 and **p < 0.01 in t-test).

Full size image

Unexpectedly, Tric-b-knockout growth plates contained severely dilated cells, which were largely negative in hematoxylin-eosin staining and randomly located in the round and columnar cell zones (Fig. 1B). In electron-microscopic observation, the dilated cells contained highly condensed nuclei and almost empty cytoplasm indicative of apoptosis. Although the dilated dead cells were very low in frequency (~0.6% of total cells in combined round and columnar cell area), such irregular cells were never detected in wild-type femoral bones. The dead cells were also detected in the growth plates of humeral and rib bones isolated from the knockout mice (Fig. S1B). TdT-mediated dUTP-biotin nick end labelling (TUNEL) has been used for detecting DNA fragmentation in apoptotic cells. Although TUNEL-positive cells could be hardly detected due to their low numbers, the positive cells were significantly increased in Tric-b-knockout round and columnar cell zones (Fig. S2A). Therefore, Tric-b deficiency seemed to occasionally cause apoptotic cell death in proliferating growth plate chondrocytes during long bone development. On the other hand, proliferating cell nuclear antigen (PCNA) is an auxiliary component of DNA polymerase during DNA duplication and repair, and has been used as a marker for proliferating cells. PCNA-positive cells were observed with similar frequencies between Tric-b-knockout and wild-type growth plates, suggesting that Tric-b deficiency does not obviously affect cell cycle in proliferating chondrocytes (Fig. S2B).

Pro-collagen accumulation in Tric-b-knockout chondrocytes

Proliferating growth plate chondrocytes mainly produce COL2A1 as a major cartilage matrix component. We next focused on collagen synthesis in Tric-b-knockout chondrocytes. In wild-type round chondrocytes from developing femoral bones, COL2A1 deposits occasionally appeared as intracellular puncta (>5 μm2) without colocalization with the ER maker KDEL sequence (Fig. 2A upper panels). However, in Tric-b-knockout chondrocytes, COL2A1 deposits were more frequently detected and colocalized with the ER marker (Fig. 2B). Furthermore, the deposits became larger and denser in Tric-b-knockout growth plates, and such severely expanded deposits sporadically covered the bulk of cytoplasm in the presumed dying cells (Fig. 2A lower panels).

Fig. 2: Pro-collagen overaccumulation in Tric-b-knockout round chondrocytes.
figure 2

(A) Confocal microscopic images of round chondrocytes, nuclear-stained with DAPI and fluorescence-stained using antibodies against COL2A1 and the ER marker KDEL. As seen in the left low-magnification images (scale bar, 200 μm), COL2A1-positive intracellular deposits (>140 μm2, arrowheads) were frequently detected in the round cell zones of Tric-b-knockout growth plates. The growth plate chondrocyte zones are indicated by the white dashed lines, and high-magnification images of the cells presenting are indicated by the white dashed boxes. In the high-magnification images (scale bar, 10 μm), wild-type growth plate panels show regular round chondrocytes (a) and one COL2A1 deposit-bearing cell (b), while Tric-b-knockout growth plate panels show dilated cells developing COL2A1 deposits (c, d). (B) Appearance ratios and cell sizes of COL2A1-positive chondrocytes in growth plates. (C) Western blot analysis of intracellular COL2A1 in growth plate chondrocytes. Representative COL2A1-immunoreactivities were shown (upper panel), and the digitalized immunoreactivities are statistically compared between wild-type and Tric-b-knockout round chondrocytes (lower bar graph). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a loading control. In the bar and scatter graphs, the data are presented as the mean ± SEM, and the numbers of mice examined are shown in parentheses. Statistical differences from the genotypes are indicated by asterisks (**p < 0.01 in t-test).

Full size image

To assess the relative amount of intracellular COL2A1 within growth plate chondrocytes, we solubilized intracellular proteins in a deoxychorate-containing solution and removed ECM by centrifugation (Fig. 2C). Tric-b-knockout lysates reproducibly exhibited dense immunostaining signals against COL2A1, although signals against the control glycolytic enzyme were similar between knockout and wild-type lysates. The results, together with the histochemical observations, indicated that the ER elements were overloaded with COL2A1 fibers in Tric-b-knockout chondrocytes. It is reasonably proposed that Tric-b deficiency deranges the ER processing or ER-Golgi trafficking of the pro-collagen fibers.

ER stress in Tric-b-knockout chondrocytes

When immature protein levels reach maximal acceptable levels in the ER lumen, the major transmembrane sensors IRE1 (transmembrane protein kinase inositol-requiring enzyme 1), ATF6 (activating transcription factor 6) and PERK (PKR-like ER kinase) activate the unfolded protein response (UPR) as a homeostatic mechanism in response to ER stresses [25]. For example, the ER chaperone BiP/GRP78 is susceptibly induced by activation of either the sensor proteins in various cell types. Under physiological conditions, UPR is moderately activated in cell types that abundantly produce secretory proteins, such as growth plate chondrocytes, pancreatic β cells and plasma cells that extensively produce collagen, insulin and antibodies, respectively. To investigate atypical UPR levels in Tric-b-knockout chondrocytes, we prepared cell lysates and total RNA from the round chondrocyte-enriched femoral epiphyses. In Western blot analysis, BiP contents were higher in the knockout lysates than those in wild-type lysates, indicating that UPR was highly activated in Tric-b-knockout chondrocytes (Fig. 3A).

Fig. 3: Altered ER stress responses in Tric-b-knockout chondrocytes.
figure 3

(A) Western blot analysis of UPR-related proteins in cell lysates prepared from round chondrocyte-enriched growth plates. Target protein contents were monitored using antibodies against BiP, phospho- and total-eIF2α, ATF4, CHOP, p90 and p50 forms of ATF6 and BBF2H7. Representative immunoblot images are shown in the upper panels, and relative immunoreactivities normalized to wild-type values are presented in the bar graph. GAPDH was used as a loading control. (B) RT-PCR analysis monitoring the unspliced (Xbp1u) and spliced (Xbp1s) Xbp1 mRNAs in round chondrocyte-enriched growth plates. Electrophoresis gel images of amplified cDNAs are shown in the upper panel, and relative cDNA intensities normalized to wild-type values are presented in the bar graph. Gapdh mRNA was also amplified as an internal control. (C) Quantitative RT-PCR analysis monitoring mRNAs transcribed from UPR-related genes in round chondrocyte-enriched growth plates. The cycle threshold (Ct) indicates the cycle number at which the amount of amplified cDNA reaches a fixed threshold in each RT-PCR reaction. In the graphs, the data are presented as the mean ± SEM, and the numbers of mice examined are shown in parentheses. Statistical differences from the genotypes are indicated by asterisks (*p < 0.05 and **p < 0.01 in t-test).

Full size image

Activated PERK phosphorylates the eukaryotic translation initiation factor eIF2α to reduce translation of newly synthesized proteins and to selectively induce the transcription factor ATF4 for the induction of stress-related genes such as Chac1 and Trb3 [25, 26]. Western blot analysis showed that both phospho-eIF2α and ATF4 contents were remarkably elevated in the knockout lysates (Fig. 3A). RT-PCR analysis indicated that Chac1 and Trib3 mRNAs were inducibly transcribed in the knockout chondrocytes (Fig. 3C). Therefore, PERK/eIF2α signaling seemed to be highly activated in the knockout chondrocytes. This conclusion was further supported by the immunohistological staining of ATF4 and COL2A1 (Fig. S3). ATF4 immunofluorescence was positive in ~30 cells per wild-type growth plate, while ~106 ATF4-positive cells were detected in the Tric-b-knockout growth plate. When focusing on the cell populations bearing large COL2A1 deposits (>10 μm2), ~5 cells were assigned as ATF4-COL2A1 deposit-double positive cells in the wild-type growth plate, while such double-positive cells were increased to ~50 cells in the knockout growth plate.

IRE1 activation stimulates Xbp1 mRNA splicing and thus promotes the transcription of XBP1-induced genes including Edem1 and Erdj4 [27]. In RT-PCR analysis, Tric-b-knockout growth plates exhibited no aberrant features in Xbp1 mRNA splicing and XBP1-induced gene expression (Fig. 3B, C). On the other hand, ATF6 p90 is cleaved under ER stress conditions, and the resulting cytoplasmic fragment ATF6 p50 translocates into the nucleus as an active transcription factor to induce the expression of the ATF6-induced genes, including Grp94 and Herp [28]. We observed similar ATF6 cleavage and ATF6-induced gene expression between Tric-b-knockout and wild-type specimens (Fig. 3B, C). Therefore, of the major UPR pathways, IRE1 and ATF6 signalings were functioning at similar intensities between the knockout and wild-type chondrocytes, while PERK/eIF2α signaling seemed to be facilitated in the knockout cells. This conclusion was further supported by the microarray data derived from round chondrocyte-enriched specimens (Fig. S4A); the heatmap data suggested that CHOP-regulated and ATF4-induced genes were preferentially activated presumably downstream of eIF2α phosphorylation in the knockout chondrocytes.

BBF2H7 (box B-binding factor 2 human homolog on chromosome 7) is an ER stress sensor essential for chondrogenesis, and its cleaved activation stimulates Sec23a and Mcl1 gene expression [29]. SEC23A protein contributes to COPII vesicle formation for ER-Golgi trafficking, while MCL1 (myeloid cell leukemia sequence 1) belongs to the anti-apoptotic BCL-2 (B-cell leukemia/lymphoma 2) family. Impaired Sec23a and Mcl1 expression promotes ER dilation and apoptosis in Bbf2h7-knockout chondrocytes [29], similar to the morphological abnormalities observed in Tric-b-knockout chondrocytes. Western blotting results showed that BBF2H7 was cleaved more than normal in the knockout chondrocytes (Fig. 3A), and RT-PCR analysis indicated that both Sec23a and Mcl1 genes were similarly activated between the knockout and wild-type chondrocytes (Fig. 3C). Therefore, the BBF2H7 pathway seemed to function normally in Tric-b-knockout chondrocytes.

Apoptosis-related alterations in Tric-b-knockout chondrocytes

It has been reported that caspase 8 (CASP8), CASP9 and CASP12 become differentially active to function as initiator caspases under severe ER stress conditions [30]. Western blotting suggested no aberrant activation of CASP8 and CASP9 in Tric-b-knockout chondrocytes (Fig. 4A). In contrast, both intact and cleaved forms of CASP12 were more abundant in the knockout growth plates than in wild-type controls. Therefore, of the ER stress-related caspase subtypes, CASP12 was most likely preferentially activated in Tric-b-knockout chondrocytes.

Fig. 4: Caspase subtypes in Tric-b-knockout chondrocytes.
figure 4

(A) Western blot analysis of CASP8, CASP9 and CASP12 in cell lysates prepared from round chondrocyte-enriched growth plates. Both pro- and cleaved forms of the caspase subtypes were detected by specific antibodies, and representative immunoreactivities are presented (upper panels). The immunoreactivities of pro- and cleaved-CASP12 were digitalized and are statistically compared between wild-type and Tric-b-knockout specimens (bar graph). GAPDH was also analyzed as an internal control. (B) Immunochemical analysis of CASP3. Western blot analysis in round chondrocyte-enriched lysates indicated similar pro-CASP3 levels in wild-type and Tric-b-knockout specimens, but failed to detect cleaved-CASP3 in both the specimens (upper panels). However, immunohistochemical analysis occasionally detected cleaved-CASP3-positive cells in Tric-b-knockout round chondrocyte zones (middle panels: scale bar, 10 μm). The appearance frequencies of cleaved-CASP3-positive cells were statistically compared between wild-type and Tric-b-knockout growth plates (bar graph). In the bargraphs, the data are presented as the mean ± SEM., and the numbers of mice examined are shown in parentheses. Statistical differences from the genotypes are indicated by asterisks (**p < 0.01 in t-test).

Full size image

CASP3 catalyzes the cleavage of many key cellular proteins and serves as a major effector caspase that executes apoptosis [30]. Western blotting failed to detect cleaved-CASP3 in both Tric-b-knockout and control growth plates (Fig. 4B). However, immunohistochemical analysis occasionally detected cleaved CASP3-positive dilated cells in Tric-b-knockout growth plates, while such immunofluorescence-positive cells were never observed in wild-type growth plates (Fig. 4B). Therefore, the cleaved CASP3-positive cells were probably assigned as dilated chondrocytes undergoing apoptosis in Tric-b-knockout growth plates. Overall, CASP12 and CASP3 likely contributed to Tric-b-knockout apoptosis as initiator and effector caspases, respectively.

Altered store Ca2+ handling in Tric-b-knockout chondrocytes

In our previous studies, impaired store Ca2+ release and store Ca2+ overloading were commonly observed in functionally defective cells prepared from Tric-a- and Tric-b-knockout mice [9, 13,14,15]. To examine store Ca2+ handling of Tric-b-knockout chondrocytes, we prepared slice specimens from embryonic femoral bones and collected Fura-2 imaging data from round chondrocytes. Using a perfusion protocol with normal, Ca2+-free, ATP-supplemented and Ca2+ ionophore ionomycin-containing bathing solutions (Fig. 5A), we examined IP3-induced Ca2+ release in response to purinergic P2Y receptor activation, ionomycin-induced Ca2+ leak and store-operated Ca2+ entry (SOCE). In Tric-b-knockout chondrocytes, Ca2+ transients evoked by P2Y receptor activation became weak, but ionomycin-induced Ca2+ leak and SOCE remained unaltered (Fig. 5A). Therefore, IP3-induced Ca2+ release was significantly impaired in the knockout chondrocytes. However, it was rather surprising that intracellular stores were not Ca2+-overloaded despite the impaired IP3R-mediated Ca2+ release in the knockout chondrocytes (Fig. 5B).

Fig. 5: Impaired store Ca2+ handling in Tric-b-knockout chondrocytes.
figure 5

Round chondrocytes in E17.5 femoral bone slices were examined by Fura-2 imaging. (A) Using the perfusion protocol indicated, representative recording traces obtained from two cells in wild-type (WT) and Tric-b knockout mice are shown (left panel). ATP-evoked Ca2+ transients (ATP), ionomycin-induced Ca2+ leak responses (IM) and store-operated Ca2+ entry responses (SOCE) are statistically analyzed (dot graph). (B) Ionomycin-induced Ca2+ responses with or without naltriben pretreatment. Representative recording traces from two cells in each genotype are shown (left panels), and the observed Ca2+ responses were statistically analyzed (dot graphs). The data are presented as the mean ± SEM, and the numbers of cells and mice examined are shown in parentheses. Statistical differences between the genotypes are marked with asterisks (**p < 0.01 in t-test). ns not significant.

Full size image

Growth plate chondrocytes generate spontaneous Ca2+ influx by intermissive gating of cell-surface TRPM7 channels, which can be pharmacologically activated with the channel agonist naltriben [31]. Thus, it is presumed that naltriben preconditioning facilitates autonomic Ca2+ entry and ensures full Ca2+ loading in intracellular stores. Immediately after naltriben pretreatment, ionomycin-induced Ca2+ release was obviously higher in Tric-b-knockout chondrocytes than in wild-type controls (Fig. 5B). Therefore, Tric-b deficiency might decelerate Ca2+ leakage from full stores, thus generating temporal Ca2+-overloaded stores. The temporal Ca2+ overloading and weakened ATP-induced Ca2+ release were consistent with the prevailing notion that TRIC channels facilitate Ca2+ release by providing counter-cationic currents [9]. It might be a reasonable speculation that IP3R gating is enhanced, thus facilitating Ca2+ leakage and preventing store Ca2+ overloading in the knockout chondrocytes, because steady-state phospholipase C (PLC) activity was likely elevated under PERK-activated conditions (see below section).

Facilitated Ca2+ entry in Tric-b-knockout chondrocytes

The elevation of resting intracellular Ca2+ concentration ([Ca2+]i) is generally associated with enhanced Ca2+ influx. In growth plate chondrocytes, TRPM7 channels are intermittently activated by intrinsic phosphoinositol turnover and predominantly responsible for resting Ca2+ influx [31]. The resting [Ca2+]i of Tric-b-knockout chondrocytes was significantly elevated in a normal bathing solution (Figs. 5A and 6A). However, under Ca2+-free, TRPM7 inhibitor FTY720-treated and PLC inhibitor U73122-supplemented conditions, Tric-b-knockout and wild-type chondrocytes exhibited similar resting [Ca2+]i levels (Fig. 6A-C). Therefore, steady-state PLC activity was probably enhanced, and thus, TRPM7-mediated Ca2+ entry was facilitated in Tric-b-knockout chondrocytes.

Fig. 6: Elevated resting [Ca2+]i in Tric-b-knockout chondrocytes.
figure 6

In the Fura-2 imaging using the perfusion protocol indicated, representative recording traces obtained from two cells in wild-type (WT) and Tric-b knockout mice are shown (left panels). The effects of Ca2+-free bathing solution (A), the TRPM7 inhibitor FTY720 (B) and the PLC inhibitor U73122 (C) on resting [Ca2+]i are summarized (dot graphs). The data are present as the mean ± SEM, and the numbers of cells and mice examined are shown in parentheses. Statistical differences between the genotypes are marked with asterisks (**p < 0.01 in t-test). ns not significant.

Full size image

To evaluate the link between activated PERK/eIF2α signaling and elevated resting [Ca2+]i in Tric-b-knockout chondrocytes, we utilized the PERK inhibitor GSK2606414 and the PERK activator CCT020312. GSK2606414 treatments (20 μM) did not significantly affect resting [Ca2+]i in wild-type chondrocytes but clearly decreased [Ca2+]i in the knockout chondrocytes (Fig. 7A). Therefore, under GSK2606414-treated conditions, the knockout and wild-type cells exhibited similar resting [Ca2+]i In contrast, CCT020312 treatments (1 μM) obviously elevated [Ca2+]i in wild-type chondrocytes but not in the knockout chondrocytes (Fig. 7B). The observations likely suggested that activated PERK/eIF2α signaling mainly contributed to TRPM7 channel facilitation by stimulating steady-state phosphoinositol turnover in the knockout chondrocytes.

Fig. 7: Effects of PERK modulators on resting [Ca2+]i in Tric-b-knockout chondrocytes.
figure 7

In the Fura-2 imaging using the perfusion protocol indicated, representative recording traces obtained from two cells in wild-type (WT) and Tric-b knockout mice are shown (upper panels). The effects of the PERK inhibitor GSK2606414 (A) and the PERK activator CCT020312 (B) on resting [Ca2+]i are summarized (dot graphs). The data are presented as the mean ± SEM, and the numbers of cells and mice examined are shown in parentheses. Statistical differences between resting Fura-2 ratios before and after drug treatments are marked with asterisks (**p < 0.01 in t-test). ns not significant.

Full size image

Discussion

OI patients with TRIC-B mutations commonly display short stature [20,21,22,23], and Tric-b-knockout mice are small in body size [15]. Long-bone outgrowth primarily depends on vital proliferation and devoted ECM synthesis in growth plate chondrocytes [18]. Based on the observations in this study, we propose the sequential mechanisms underlying bone outgrowth disturbance and apoptosis in Tric-b-knockout growth plates (see the conceptional steps (1) ~ (10) in Fig. S5A). Ca2+-dependent chaperones and processing enzymes contribute to the maturation of secretory proteins in the ER and Golgi [32], and vesicular trafficking between the ER and Golgi requires Ca2+-dependent processes [33]. Defective store Ca2+ handling generally aggravates ER stress by disturbing protein processing and vesicular trafficking; for example, the store Ca2+ pump inhibitor thapsigargin is a potent UPR inducer. (step 1) Tric-b deficiency compromises store Ca2+ handling (Fig. 5) and likely deranges the processing and trafficking machinery, leading to (step 2) the overaccumulation of unfolded proteins including pro-collagen within ER elements in growth plate chondrocytes (Fig. 2). (step 3) Pro-collagen overaccumulation likely aggravates ER stress in Tric-b-knockout chondrocytes, and thus (step 4) leads to excess PERK-mediated eIF2α phosphorylation (Fig. 3). The intracellular pro-collagen deposits suggest defective ECM maturation in the ER, while eIF2α hyperphosphorylation likely broadly attenuates cellular translation. Therefore, (step 5) ECM synthesis/secretion is probably impaired in the knockout chondrocytes, and (step 6) ECM contents are reduced in Tric-b-knockout growth plates. In particularly, the ECM portion of the round chondrocyte zone is clearly reduced (Fig. 1A), and this regression is likely to mainly contribute to the impaired outgrowth of developing bones in the knockout mice. On the other hand, (step 7) activated PERK/eIF2α signaling seems to induce ATF4 and CHOP expression then lead to excess UPR gene expression (Fig. 3) and (step 8) also may essentially contribute to PLC stimulation for facilitating Ca2+ entry and elevating resting [Ca2+]i in the knockout chondrocytes (Fig. 7). Accordingly, the knockout chondrocytes may tend to be sensitive to apoptosis, because the calpain-CASP12-CASP3 cascade has been reported in various types of apoptosis [34,35,36]. In the knockout chondrocytes, (step 9) further incidental [Ca2+]i elevation may be predisposed to activate Ca2+-dependent calpain for CASP12-cleaved stimulation, and (step 10) resulting CASP12 activation likely triggers the CASP3-mediated apoptotic cascade. However, the accidental apoptosis rarely takes place in knockout growth plates (Figs. 1 and S2) and seems unlikely to obviously contribute to the impaired bone outgrowth.

In conventional processing, the three major ER stress sensors PERK, ATF6 and IRE1 become active by the dissociation of the ER chaperone BiP/GRP78 on the luminal side because excess unfolded proteins attract the chaperone [25]. Our biochemical, gene expression and Ca2+ imaging data suggested that ATF6 and IRE1 signalings were similarly activated between Tric-b-knockout and wild-type cells, but that PERK/eIF2α signaling was excessively activated in the knockout chondrocytes. To verify the proposed PERK activation, we examined its autophosphorylation by immunoblot analysis, and unexpectedly detected comparable phospho-PERK contents between Tric-b-knockout and wild-type growth plates (Fig. S4B). In addition to PERK, three other kinases, HRI, PKR and GCN2, become active under stress conditions such as viral infection and heme depletion, and are known to catalyze eIF2α phosphorylation that initiates PERK/eIF2α signaling [37]. Again, the autophosphorylated forms of the additional kinases seemed similar in content between the genotypes (Fig. S4C). Therefore, eIF2α hyperphosphorylation may not be due to kinase activation, and it can be thus hypothesized that the dephosphorylation of phospho-eIF2α is dampened in the knockout chondrocytes. Previous studies have indicated that the eIF2α dephosphorylation is mainly catalyzed by protein phosphatase 1 (PP1) complexes in several cell types [38] and also that forceful PERK/eIF2α signaling stimulates PP1-mediated eIF2α dephosphorylation as a feedback regulation [39]. A PP1 holoenzyme consists a catalytic subunit, PP1α, PP1β or PP1γ, and one or two regulatory PP1-interacting proteins; >200 distinct PP1 complexes are thought to differentially formed depending on cell types because there are many PP1-interacting proteins so far identified [38]. Therefore, it can be reasonably presumed that PP1-mediated catalytic activity or feedback machinery may be compromised in Tric-b-knockout chondrocytes (Fig. S5B). Gene expression profiles for the PP1 catalytic subunits were not altered in the knockout growth plates (Fig. S4A), but the composition of major PP1 complexes is totally unknown in growth plate chondrocytes. To clarify the detailed pathophysiological mechanism underlying the excess activation of PERK/eIF2α signaling in Tric-b-knockout chondrocytes, it seems important to elicit the structure and function of PP1 complexes in growth plate chondrocytes.

Of the three major ER stress pathways, the PERK/eIF2α signaling has been repeatedly reported to be linked with aberrant cellular Ca2+ handling [40,41,42,43,44]. For example, the PERK/eIF2α signaling is sensitively activated in response to ER Ca2+ depletion [40, 41], and elevated [Ca2+]i activates PERK/eIF2α signaling to promote apoptosis in virus-infected cells [44]. Such observations may imply that deranged Ca2+-handling preferentially induces PERK/eIF2α signaling (see step 3 in Fig. S5A). Furthermore, in our Ca2+ imaging experiments, the PERK inhibitor seemed to immediately attenuate activated Ca2+ influx in the knockout chondrocytes, suggesting that downstream of activated PERK/eIF2α signaling, PLC is activated for enhancing TRPM7-mediated Ca2+ influx in growth plate chondrocytes [31]. Therefore, we can propose a bidirectional link between PERK/eIF2α signaling and cellular Ca2+-handling in growth plate chondrocytes; PERK/eIF2α signaling is activated by deranged store Ca2+ handling due to Tric-b deficiency, and activated PERK/eIF2α signaling stimulates Ca2+ influx in a PLC-dependent manner. However, in the PERK/eIF2α signaling cascade proposed thus far, Ca2+-dependent processes and Ca2+-handling proteins serving as PERK substrates have not been reported. From a biological point of view, the molecular mechanism underlying the proposed bidirectional link seems to be the important issue that needs to be solved in future studies.

Materials and methods

Histological analyses

Tric-b knockout mice were generated and genotyped as described previously [9].

For histological analysis, the E18.5 femur, humerus and rib bones were fixed in 4% paraformaldehyde, embedded in Super Cryoembedding Medium (Section-lab, Japan), and frozen in liquid nitrogen. Serial cryosections (~10 μm in thickness) were prepared from the fixed specimens and treated with a commercial hematoxylin and eosin solution (Wako Pure Chemical, Japan) or alcian blue solution (Merck, USA) for microscopic observation (BZ-X710, Keyence Co., Japan).

For immunohistochemical analysis, the femoral cryosections were treated with 1% bovine serum albumin to block nonspecific binding. The cryosections were incubated with primary antibodies against COL2A1, ATF4 and KDEL, and then were incubated with an AlexaFluor 488-conjugated antibody against goat anti-mouse IgG and an AlexaFluor 555-conjugated antibody against rabbit IgG (Table S1). After DAPI nuclear staining, fluorescence-labeled sections were examined under a microscope (BZ-X710, Keyence Co). For immunohistochemical analysis of PCNA, the cryosections (4 µm in thickness) were immersed with a citrate buffer and heated at 90 ˚C for 20 min. After treatments with a 3% H2O2 solution for 5 min and a blocking solution (Blocking One Histo, Nacalai Tesque, Japan) for 30 min, the cryosections were incubated with primary antibodies against PCNA, and then incubated with horseradish peroxidase-conjugated secondary antibody. PCNA reactivity was visualized using diaminobenzidine hydrochloride substrate kit (Abcam, UK) and analyzed by microscopic observation (BZ-X710, Keyence Co.). For TUNEL staining, the cryosections (6 µm in thickness) were stained using TUNEL assay kit (Abcam, ab206386) according to the manufacturer’s instructions and observed by a microscope (BZ-X710, Keyence Co.). Captured images from hematoxylin/eosin-stained and fluorescence-labeled sections were quantitatively analyzed using BZ-X analyzer (Keyence) and ImageJ (U.S. National Institutes of Health) software.

Ultrastructural analysis

For electron-microscopic analysis, the femoral bones were fixed in prefixative solution (3% paraformaldehyde, 2.5% glutaraldehyde, 0.1 M sodium cacodylate, pH 7.5) and placed in postfixative solution (0.1% OsO4, 0.1 M potassium ferricyanide, 0.1 M sodium cacodylate, pH 7.4) at room temperature. The specimens were dehydrated using ethanol and acetone, and embedded in Epon to prepare thin sections (100 ~ 150 nm in thickness) for analysis under a transmission electron microscope (JEM-200CX, JEOL, Japan).

Gene expression analysis

Total RNA was prepared from mouse tissues using a commercial kit (Isogen, Nippon Gene). RNA preparations from femoral cartilage plate sections enriched with round chondrocytes were reverse-transcribed and analyzed using the GeneChip Mouse Genome 430 2.0 (Affymetrix) according to the manufacturer’s instructions by an outsourcing company (Takara Bio Co., Japan). The array probe intensities were analyzed with the robust multiarray analysis expression algorithm, which represents the log transformation of intensities (background corrected and normalized) from the gene chips [45], and were visualized in the heatmaps.

To further analyze gene expression, mRNA contents were determined by quantitative RT-PCR as described previously [46]. Total RNA was reverse-transcribed using the ReverTra ACE qPCR-RT Kit (Toyobo), and the resulting cDNA was examined by real-time PCR (LightCycler 480 II, Roche). The cycle threshold (Ct) was determined from the amplification curve as an index for relative mRNA level in each reaction. The RT-PCR primer sets used in this study are listed in Table S2.

Immunoblot analysis

Round chondrocyte-enriched growth plate parts were isolated from the E18.5 femoral bones and homogenized in lysis buffer containing 4% sodium deoxycholate, 20 mM Tris-HCl (pH 8.8), 100 mM NaF, 10 mM Na3PO4, 1 mM Na3VO4, and 20 mM β-glycerophosphate. After sonication (Astrason Ultrasonic Processor XL, Misonix), the homogenates were centrifuged (17,000 x g, 30 min) to remove tissue debris. After measurement of total protein concentration (BCA Protein Assay Kit, Pierce), the resulting growth plate lysates were subjected to 8–16.5% SDS–polyacrylamide gel electrophoresis, and separated proteins were then transferred to PVDF membranes (polyvinylidene difluoride, Merck Millipore). After treatments with a blocking reagent (Blocking One solution, Nacalai Tesque, Japan), the membranes were incubated with primary antibodies and then incubated with secondary antibodies; antibodies were diluted in Can Get Signal (Toyobo, Japan) before the immunoreaction. Immunoreactivity was visualized using a chemiluminescence reagent (GE Healthcare Life Sciences) and an image analyzer (Amersham Imager 600, GE Healthcare Life Sciences) and were quantitatively analyzed using ImageJ software. The antibodies used in this study are listed in Table S1.

Fura-2 Ca2+ imaging

Bone slices were prepared and Ca2+ imaging was performed as described previously [31]. The bathing solution used was HEPES-buffered saline (150 mM NaCl, 4 mM KCl, 1 mM MgCl2, 2 mM CaCl2, 5.6 mM glucose and 5 mM HEPES, pH 7.4). For indicator loading, bone slices prepared using a vibratome slicer were placed on glass-bottom dishes (Matsunami, Japan) and incubated in HEPES-buffered saline containing 15 μM Fura-2 AM (Dojindo, Japan) for 60 min at 37 °C. For ratiometric imaging, excitation wavelengths of 340 and 380 nm were alternately delivered, and an emission wavelength of >510 nm was detected by a cooled electron multiplying charge-coupled device camera (model C9100-13; Hamamatsu Photonics, Japan). The bone slices were mounted on an upright fluorescence microscope (DM6 FS, Leica) carrying a water immersion objective (HCX APO L 40×, Leica).

Quantification and statistical analysis

All data obtained are presented as the means ± SEM. with n values indicating the number of examined mice or cells. Student t-test was used for two-group (Prism 7, GraphPad Software Inc.): p < 0.05 was considered to be statistically significant.