A comparative analysis of TonEBP conditional knockout mouse models reveals inter-dependency between compartments of the intervertebral disc

The notochord is a rod-like, transient and essential embryonic component of the forming axial skeleton. It acts as a structural tissue and as an instructional center for the dorsoventral patterning of the paraxial mesoderm, which will give rise to the sclerotome and sclerotome-derived compartments of the intervertebral discs (IVDs). Those include the annulus fibrosus (AF), the cartilage end plates (CEPs), the vertebral growth plates (GPs), and the bony vertebral bodies (Chiang et al., 1996; Placzek et al., 1991; Sivakamasundari and Lufkin, 2012; Williams et al., 2019; Yamada et al., 1991). Embryonic notochord-derived cells persist postnatally in each IVD of the postnatal spine, forming the gelatinous-like nucleus pulposus (NP) in the center of each IVD. Lineage-tracing studies have demonstrated that NP cells derive entirely from the notochord (Choi et al., 2008; McCann et al., 2012). However, there are still important gaps of knowledge about the biology of the IVD. Those include the fate of IVD cells during the growth, maintenance and aging of the IVD, and whether they maintain peri- and postnatally an instructional function toward adjacent tissues that is necessary for the maintenance of IVD structure and function. Studies on ex vivo IVD tissues by Dahia and collaborators, focused on Hh signaling, support the notion of an instructional NP postnatally (Dahia et al., 2012).

In mouse, starting at embryonic day (E) 13.5, sclerotomal condensations compress and segment the notochord along the anteroposterior body axis to form each IVD, leading by E15.5 to the three tissue compartments typical of the IVD: the central NP, the AF, which encases the NP, and the CEPs and GPs (hereafter CEP/GP) at the rostrocaudal sides of the NP (Aszodi et al., 1998; Choi et al., 2008; Haga et al., 2009; Lawson and Harfe, 2015; McCann et al., 2012). The gelatinous nature of the NP extracellular matrix (ECM) gives this tissue its properties as a shock absorber to withstand the compressive forces acting on the spine. The fibrous nature of the AF enables the IVD to endure tension and to hold the central NP in place during compression. The CEP/GP, forming a thin layer of hyaline cartilage, is involved in the diffusion of oxygen, nutrients and metabolites into the avascular NP (Giers et al., 2017; Urban et al., 2004). To what extent abnormalities in the CEP/GP compartment alter the homeostasis of the NP and contribute to IVD degeneration remains a matter of debate. Recent studies indicated that CEP/GP compositional abnormalities, including high amounts of type II collagen (COL2) and aggrecan (ACAN), high mineral content, and fewer mature cross-links, hinder nutrient diffusion through the CEP (Wong et al., 2019), which may contribute to NP pathologies. The experimental digestion of human CEP ECM constituents by matrix metalloproteinase 8 (MMP8), a metalloproteinase with affinity for COL2 and ACAN, by contrast, has been shown to enhance CEP transport properties (Dolor et al., 2019). Therefore, signals through and/or from the CEP/GP, in the form of small diffusible molecules secreted by CEP/GP cells or release from their ECM, may contribute to NP homeostasis, pathological innervation (Fields et al., 2014) and degeneration during aging (Adams et al., 2000).

IVDs are characterized by their high proteoglycan content. The negatively charged glycosaminoglycan chains of these proteoglycans attract cations that subject resident cells to a high osmotic pressure, in a similar manner to cells in the kidney medulla that are exposed to urea and NaCl during diuresis. Without counteracting mechanisms to balance extracellular osmotic pressure, this environment would lead to cellular water loss and cell death. Such counteracting mechanisms include first the import of Na⁺, K⁺ and Cl⁻ via electrolyte transporters. These intracellular electrolytes, however, can negatively alter protein and DNA structure, protein translation, cytoskeleton architecture and mitochondrial function; thus, this first response is followed by a gradual replacement of electrolytes by non-charged organic osmolytes, such as myo-inositol, betaine, taurine or sorbitol (Yancey et al., 1982). These molecules accumulate within cells as a result of induction of osmoprotective genes allowing their synthesis or uptake (Burg et al., 1997; Ito et al., 2004; Nakayama et al., 2000). TonE-binding protein (TonEBP), also called ORE-binding protein (OREBP) or nuclear factor of activated T cells 5 (Nfat5), is a transcription factor that binds to the promoter of these osmoprotective genes upon a rise in extracellular osmotic pressure and stimulates their transcription (Ko et al., 2000; Miyakawa et al., 1999). TonEBP is expressed in cells of the notochord, NP and GP/CEP, which are exposed to high extracellular osmolarity generated by ions associated with the negative charge of their proteoglycan-rich ECM (Adams et al., 1977; Cs-Szabo et al., 2002; Hayes et al., 2001; Ishihara et al., 1997), and thus could have an osmoprotective role in the axial skeleton, as in kidney cells (Caron et al., 2013; Tessier et al., 2019; Tsai et al., 2006; van der Windt et al., 2010). In vitro evidence based on the use of promoter-luciferase constructs and shRNA-based TonEBP loss-of-function approaches showed that TonEBP in rat NP cells stimulates the transcription of Aqp2 (Gajghate et al., 2009), TauT (Slc6a6) (Gajghate et al., 2009), Hsp70 (Hspa1a/1b) (Tsai et al., 2006) and GlcAT-I (B3gat3) (Hiyama et al., 2009), which are all involved in osmoresponse mechanisms. TonEBP was also shown to stimulate the expression of aggrecan (Acan), a major ECM proteoglycan of the NP (Tsai et al., 2006) and to induce the expression of pro-inflammatory genes such as Il6, Tnfa (Tnf), Nos2 and Ccl2 in NP cells in vitro (Johnson et al., 2016). In addition, overexpression of a dominant-negative form of TonEBP in NP cells impaired their survival under hypertonicity (Tsai et al., 2006). Despite these in vitro data, the in vivo requirement of TonEBP for notochord and IVD formation remains less established. TonEBP loss-of-function or expression of a dominant-negative form of TonEBP in the chordate Ciona impaired notochord development and vacuolization of notochordal cells (He et al., 2022). However, mouse embryos lacking the DNA-binding domain (DBD) of TonEBP globally (TonEBP^Δ6-7 homozygous) showed no detectable defect in notochord inflation and vacuolization, nor a reduction of Acan expression at E12.5 and E17.5, although a mild delay in NP formation and vertebral column development was reported (Go et al., 2004; Tessier et al., 2019). Postnatally, TonEBP^Δ6-7/+ heterozygous adult mice displayed a mild age-related degeneration of the IVD at 12 months of age (Tessier et al., 2020). The global and TonEBP heterozygous nature of this mouse model, and the lethality of knockout (KO) adult mice (and likely humans with TONEBP loss-of-function variants), however, were a limitation to understanding the role of this gene in the growth and maintenance of the IVD. In addition, two studies reported distinct TonEBP mutant alleles that both generate near full-length TonEBP RNAs and proteins lacking only the DBD (Go et al., 2004; Lee et al., 2016), thus limiting possibilities to identify or confirm the full spectrum of the functions of TonEBP supported by in vitro shRNA approaches (Kang et al., 2019; Lee et al., 2016).

To overcome the limitations of previous TonEBP loss-of-function models, we generated a new TonEBP floxed allele to determine the effect of complete TonEBP loss-of-function on IVD formation, maintenance and degeneration. In this study, we report the phenotypes of mice in which TonEBP is ablated in different compartments of the IVD embryonically or postnatally, and the knowledge derived from these analyses in term of opportunities to characterize the crosstalk between IVD compartments necessary for optimal IVD growth and maintenance.

The production of stable and shorter TonEBP proteins that lack the N-terminal portion of the TonEBP DBD is common to all available mutant TonEBP alleles (TonEBP^Δ6-7/+ and TonEBP^ex6 alleles) (Go et al., 2004; Lee et al., 2016). We confirmed this phenomenon in the TonEBP^ex6 floxed mice generated by Kuper et al. (2014) upon Col2-Cre-mediated recombination in primary chondrocytes, as observed by the presence of TonEBP RNAs and proteins of shorter size compared with wild-type (WT) counterparts (Fig. S1A,B). We also observed that recombination of the TonEBP^ex6 floxed allele generated aberrant alternative splicing isoforms that were not observed in WT chondrocytes (Fig. S1C). These TonEBP alleles may thus not allow the identification of DBD-independent functions of TonEBP and may create non-physiological isoforms for which an impact on cell behavior cannot be ruled out. Therefore, we inserted via CRISPR/Cas9 an additional loxP site after the last exon of the TonEBP^ex6 floxed allele to generate a new TonEBP^{ex6-15/ex6-15} allele, which can be targeted by Cre-recombinase activity to produce a total loss of TonEBP (Fig. 1A; Materials and Methods). Following verification of CRISPR-based genomic sequence modification by Sanger sequencing, we then crossed TonEBP^{ex6-15/ex6-15} mice with Col2-Cre^Tg/+ transgenic mice (Ovchinnikov et al., 2000) to assess recombination efficiency of this new TonEBP floxed allele. We observed that the expression of TonEBP was downregulated by 90% in conditional KO (cKO) primary chondrocytes prepared from Col2-Cre^Tg/+; TonEBP^{ex6-15/ex6-15} (hereafter called Col2^cKO) compared with WT (Col2-Cre^+/+; TonEBP^{ex6-15/ex6-15}) chondrocytes (Fig. 1B), resulting in a total absence of detectable TonEBP protein (Fig. 1C). The expression of known TonEBP osmotic stress response target genes, including Ar and Bgt1 (Slc6a12) (Miyakawa et al., 1998), was also blunted in Col2^cKO chondrocytes following hypertonic stress induction (75 mM NaCl, 8 h; Fig. 1D). These data support complete TonEBP loss of function with this new TonEBP^{ex6-15/ex6-15} allele.

To validate further the loss of TonEBP function, we backcrossed TonEBP^{ex6-15/ex6-15} mice with Cmv-Cre transgenic mice (Cmv-Cre^Tg/+; TonEBP^{ex6-15/ex6-15}) for three generations, thereby recombining the floxed sequences ubiquitously. Then, mice that inherited the recombined TonEBP allele (TonEBP^+/Δ) were used as breeders for embryonic development studies. We did not detect any morphological difference between WT (TonEBP^+/+), heterozygous (TonEBP^+/Δ) and homozygous (TonEBP^Δ/Δ) embryos at E12.5 (Fig. 1E, top row). At E13.5, we observed the formation of an oedema in the dorsal part of the spine in all TonEBP^Δ/Δ embryos (Fig. S1D), and the severity of the edema increased at the E15.5 time point (Fig. 1E, bottom row, red arrowheads). These results are similar to what was reported by Mak et al. in TonEBP^−/− embryos (Mak et al., 2011). No TonEBP^Δ/Δ embryos were identified within the born litters (Fig. 1F), as observed upon global homologous recombination of TonEBP, which induced in utero lethality at the embryonic stage (∼E15.5) (Go et al., 2004). This new TonEBP^{ex6-15/ex6-15} allele thus allows complete in vivo TonEBP loss of function upon Cre-mediated recombination, which alleviates the limitations associated with previously generated TonEBP alleles.

To determine the role of TonEBP in the formation of the mouse IVD, we analyzed the spine of Col2^cKO mice, in which the Cre-recombinase is active in all compartments of the IVD at E9.5 (Ovchinnikov et al., 2000). Col2^cKO mice were all viable, but were lighter and shorter than WT littermates at 8 months of age and in both sexes, with the size reduction affecting both axial (Fig. 2A) and appendicular bone elements (Fig. S1E). On lumbar spine X-rays, vertebral height of mutant mice was 33% lower than age and sex-matched WT controls, and the disc height index (DHI) was 123% higher (Fig. 2B). On histological Safranin O-stained sections, we observed obvious alterations of NP morphology in Col2^cKO mice, including lower NP width and NP aspect ratio (width/height), and an increase in cell density (Fig. 2C,,Table 1). We also observed a loss of GP structural continuity in all Col2^cKO mice in the form of marrow cavities between CEP and GP regions of the caudal IVD mainly, which were not observed in WT controls. The degeneration score calculated from Hematoxylin and Eosin-stained sections (Melgoza et al., 2021) reflected morphological and cellular changes in all IVD compartments, with strongest alterations in AF and interface areas (Fig. 2D,E).

To determine the early events leading to the severe IVD phenotypes observed in 8-month-old adult Col2^cKO mice, we analyzed the IVDs of Col2^cKO mice during IVD growth, 21 days after birth [postnatal day (P) 21]. We observed in mutant mice a ruffled appearance of the osteochondral surface on both rostral and caudal IVD sides, and a loss of typical proliferative chondrocyte column organization (Fig. 3A), reflected by a reduction in the column index value (Fig. 3B) (Killion et al., 2017). CEP and GP height, as well as cell density were not affected in Col2^cKO mice at that age (Table 2). As observed in the 8-month-old group, NP width and aspect ratio were lower in P21 Col2^cKO mice, although only the latter reached statistical difference (Table 2). Although NP area and NP cell density were not different between genotypes (Table 2), NP cells had a larger cytoplasmic area, as shown by a higher proportion of cells with an area above 50 µm² in Col2^cKO mice compared with WT littermates (Fig. 3C). The degeneration score calculated in P21 WT and Col2^cKO mice revealed milder alterations compared with 8-month-old mice with cumulative scores not reaching significance in any IVD compartment (Fig. 3D,E).

We detected in Col2^cKO mice a 33% reduction in signal intensity for COL2 in the GP and at the border between CEP and NP by immunofluorescence. COLX expression was also lower (48%) in the GP hypertrophic zone of Col2^cKO mice (Fig. 3F). The loss of TonEBP in these mice did not lead to a decrease in the ACAN signal in the CEP/GP or in the NP (Fig. 3F,G).

As previous in vitro studies suggested that TonEBP is required for NP cell survival under hypertonicity (Tsai et al., 2006), we measured the effect of TonEBP deletion on IVD cell death by terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) analysis, with the hypothesis that the loss of TonEBP would lead to cell death at all developmental stages investigated. Although a clear TUNEL signal was observed in the GP of both WT and Col2^cKO mice at P21, we did not detect any cell death in the NP of cKO mice at that age (Fig. S2), or at P10, 4 days prior to the start of the physiological NP cell apoptosis and loss reported in WT tissues (Dahia et al., 2009) (Fig. 4A,B). However, we detected a fivefold increase in cell death in the mutant GP compartment at P10, prior to the GP abnormalities detected at P21 (Fig. 4A,B). By contrast, we did not detect any difference in cell proliferation between WT and Col2^cKO mice in any IVD compartment by immunohistochemistry for the proliferation marker Ki67 (Mki67) at P10 (Fig. 4C,D). These data indicate that TonEBP is required for the formation of the IVDs and for the survival of vertebral GP chondrocytes, but they do not support a cell-autonomous role of TonEBP in the control of NP cell survival or ACAN expression.

The IVD-wide extent of TonEBP gene deletion in Col2^cKO mice did not allow us to decipher in which IVD compartment TonEBP is required for IVD formation. To determine the tissue of origin of the phenotypes observed in Col2^cKO mice, we selected the inducible Krt19-CreERT^Tg/+ Cre transgenic line, which targets specifically the NP (Mohanty et al., 2020), and the Col2-CreERT^Tg/+ line, which targets all IVD compartments except the NP and outer AF (Wei et al., 2019). We first confirmed specific targeting of each IVD compartment in Cre^Tg/+; TdTomato^Tg/+ transgenic lines by TdTomato reporter activity tracing, following Cre induction by tamoxifen (Tam) injections at P3 and P5. We detected TdTomato activity exclusively in the NP of Krt19-CreERT^Tg/+ reporter mice, whereas in Col2-CreERT^Tg/+ reporter mice, all compartments except the NP and outer AF displayed TdTomato activity (Fig. 5A). We next generated two inducible cKO mouse lines that lacked TonEBP in NP cells or in cells of CEP/GP/inner AF by crossing these two lines to TonEBP^{ex6-15/ex6-15} mice. Cre-recombinase induction and subsequent TonEBP loss of function was induced by Tam injections at P3 and P5, and histological analyses were performed at two time points: 5 days (P10) and/or 16 days (P21) later.

Deletion of TonEBP specifically in NP cells of the IVD in Krt19-CreERT^Tg/+; TonEBP^{ex6-15/ex6-15} mice (hereafter called iK19^cKO) led to NPs that were on average 24% wider than WT controls at P21. Most of them had a flatter appearance than WT counterparts, although NP aspect ratio, NP area and NP cell density did not change (Fig. 5B, Table 2). NP cell surface in iK19^cKO mice was normal, as shown by frequency distribution analysis (Fig. 5C). Unexpectedly, the loss of TonEBP in NP cells also led to a mild but significant increase in the thickness (height) of both CEP (35%) and GP (22%) compartments relative to WT controls (Fig. 5B, Table 2), despite the CEP/GP compartments not being sites of Cre activity in this model (Fig. 5A). Consistent with this result, cell density was 50% higher in the GP proliferative zone of iK19^cKO mice compared with WT controls (Table 2). Calculation of the cumulative degeneration score did not reveal significant abnormalities in the NP or surrounding compartments (Fig. 5D,E).

Similar to observations in IVD-global Col2^cKO mice, deletion of TonEBP specifically in the CEP/GP/inner AF compartments of the IVD in Col2-CreERT^Tg/+; TonEBP^{ex6-15/ex6-15} mice (hereafter called iCol2^cKO) led to a 23% increase in the height of the NPs, and to a 19% reduction in the NP aspect ratio at P21 (Fig. 5B, Table 2), despite absence of Cre activity in NP cells (Fig. 5A). Frequency distribution analysis for NP cell size confirmed in this postnatal inducible model the presence of larger NP cells in iCol2^cKO mice (Fig. 5C), although NP cell density was not affected (Table 2). The cumulative degeneration score at P21 did not reveal significant phenotypes in any IVD compartment, but, as observed in Col2^cKO mice, showed trends for alterations in AF and boundary between compartments (Fig. 5D,E).

We detected similar COL2 and COL10 levels in the NPs and CEPs/GPs of WT and iK19^cKO mice (Fig. 6A,B) by immunofluorescence. The signal for ACAN was also not affected in any IVD compartment following the loss of TonEBP in NP cells of iK19^cKO mice, consistent with observations in Col2^cKO mice and with the normal mRNA expression level of Acan in the NP cells of iK19^cKO mice [despite efficient downregulation of TonEBP and TonEBP target gene expression (Aqp1) in this compartment; Fig. 6C]. By contrast, we detected a significant 60% reduction in COL2 and COL10 levels in the CEP/GP (but not NP) of iCol2^cKO mice (Fig. 6A,B), confirming results obtained in Col2^cKO mice, whereas the overall signal intensity for ACAN was similar between the CEP/GP of WT and iCol2^cKO mice (Fig. 6B). The expression of ACAN, COL2 and COLX was unaffected in the NPs of iCol2^cKO mice (Fig. 6A).

We next analyzed cell death by TUNEL staining in WT and cKO mice at P10, a stage preceding the first main changes in IVD structure observed at P21, following Tam injection at P3 and P5. Although minimal cell death was detectable in the NP of WT, iK19^cKO and iCol2^cKO mice, we measured a twofold increase in the proportion of TUNEL⁺ chondrocytes in the GP of iCol2^cKO mice (Fig. 7A,B), consistent with what was observed in Col2^cKO mice. These data confirm the role of TonEBP in the survival of chondrocytes, but not NP cells.

To assess putative changes in cell proliferation induced by the loss of TonEBP in specific IVD compartments, we measured bromodeoxyuridine (BrdU) cell incorporation at P10 as well. No change in the number of BrdU⁺ cells was detected in the NP and CEP/GP compartments of iCol2^cKO mice (Fig. 7C,D). Unexpectedly, cell proliferation was threefold higher in both the NP and GPs of iK19^cKO mice compared with WT mice, with the GP containing the highest number of proliferating cells (Fig. 7C,D), and despite the fact that Cre-recombinase is not expressed in the GP compartment in this Cre line. This increase of cell proliferation in the GP of iK19^cKO mice was consistent with the thickened CEP/GP observed in these mice at P21 (Table 2). These results support the existence of a crosstalk between NP and CEP/GP that is necessary for normal IVD growth.

The consistent enlargement of NP cells observed following embryonic or postnatal deletion of TonEBP in Col2-Cre+ cells suggested that the loss of TonEBP in cells of the CEP/GP/inner AF had indirect repercussions on NP cells. Although it remains to be determined how mutant CEP/GP cells alter NP cells (possibly by changes in secreted products, ECM composition, structure or diffusion properties; see Discussion), we could detect a reduction in the expression level of the NP markers brachyury (T) and keratin 19 (KRT19) in the NPs of iCol2^cKO mice versus WT control mice (Fig. 8A), whereas the expression of these two markers was not affected by the loss of TonEBP in NP cells of iK19^cKO mice. These results suggest that TonEBP in cells of the CEP/GP is required to maintain the NP phenotype. An increase in DHI was also observed in adult Acan-CreERT2^Tg/+; TonEBP^{ex6-15/ex6-15} cKO mice (iAcan^cKO; Fig. 8B), in which TonEBP was ablated postnatally at P3 in all IVD compartments (Henry et al., 2009; Zheng et al., 2019). These results in independent mutant mouse lines indicate that the yet-be-identified CEP/GP-derived and TonEBP-dependent mechanism underlying the NP phenotype in these cKO mice has a dominant effect over the NP cell-intrinsic role of TonEBP. The similar IVD phenotypes between Col2^cKO (IVD-global embryonic deletion) and iAcan^cKO mice (IVD-global postnatal deletion) at 4 months of age, which included short stature (Fig. S3), short vertebrae and high DHI (Fig. 8B), low vertebral trabecular bone mass, number and higher spacing (Table 3), morphological NP, CEP/GP and AF parameters (Table 4) and trends in degeneration scores (Fig. 8D,E), also indicate that TonEBP had a predominant role in the growth of the IVD postnatally. We nevertheless observed some differences between these two mutant mouse lines, such as formation of a marrow cavity in the middle of the caudal growth plate in Col2^cKO mice (consistent with the phenotype observed in 8-month-old Col2^cKO mice), which was not observed in iAcan^cKO mice at this stage (Fig. 8C).

We have generated four cKO lines characterized by complete TonEBP loss of function in specific compartments of the IVD to explore the role of this gene in NP versus CEP/GP cells, at embryonic versus postnatal stages. This comparative analysis, summarized in Fig. 9, revealed that in the physiological in vivo environment, TonEBP is necessary for chondrocyte survival and to maintain the structure of the CEP/GP postnatally. The effect of TonEBP loss on the vertebral growth plates and end plates also indirectly affected the shape of the NP and vertebral trabecular and cortical bone mass, thus emphasizing the importance of CEP/GP integrity for normal IVD/spine growth and structure. Lastly, these analyses uncovered that TonEBP in NP cells in vivo is not required for their survival or for the expression of ACAN, but controls the proliferation of both NP cells and vertebral GP chondrocytes. Such a compartment-specific loss-of-function approach allows for a better understanding of the complex interdependency and crosstalk between IVD compartments in vivo, which could help the design of new or more efficacious preventive and regenerative strategies to improve the management of low back pain and spine degeneration.

Upon spine loading, NP and CEP/GP cells are subjected to shifts in extracellular tonicity, caused by the high proteoglycan content of their matrix and their associated ions and water. NP cells overexpressing a dominant-negative form of TonEBP in vitro were found to be more sensitive to cell death when exposed to an acute rise of hypertonicity (300 mOsM versus 500 mOsM) (Tsai et al., 2006), suggesting that TonEBP promotes NP cell survival. Unexpectedly, ablation of TonEBP in NP cells, either postnatally in iK19^cKO mice or embryonically in the notochord/NP of Col2^cKO mice, did not support a pro-survival role of TonEBP in NP cells in vivo. These results are consistent with the lack of description of higher NP cell death in previously reported TonEBP loss-of-function mouse models (Tessier et al., 2019, 2020). The observation that the loss of TonEBP did not lead to a significant decrease in Acan expression in mutant NPs or ACAN protein level (determined by immunofluorescence) in the CEP/GP and NP of all cKO mutant mice analyzed is also at odds with in vitro data that portrayed TonEBP as a direct, positive regulator of Acan transcription in rat NP cells and in the ATDC5 chondrogenic cell line (Caron et al., 2013; Tsai et al., 2006). We surmise that the differences observed between this study and previous ones related to the role of TonEBP in NP cells stem mainly from limitations of the in vitro experimental setting, or from incomplete TonEBP deletion in prior studies. Compared with the in vivo physiological environment, the 2D nature of cell culture conditions, its abnormal ECM content/structure, its extra-physiological oxygen and nutrient levels, along with the propensity for NP cells to dedifferentiate upon ex vivo culture in 2D, are likely to alter cell behavior and responsiveness to stimuli. This in vitro setting does not account either for the influence of factors from tissues adjacent to NP cells in vivo. The in vivo nature of the analysis performed with the new TonEBP^{ex6-15/ex6-15} allele, and the complete inactivation of TonEBP achieved in cKO mice derived from this mouse line, alleviates these limitations. Possible alternative interpretations of these results are that hypertonicity in the NP may not reach a level high enough to require a pro-survival and ACAN-producing function of TonEBP in NP cells.

The increase in cell proliferation detected in the NP and GP of iK19^cKO mice was unexpected, as TonEBP loss of function has mostly been associated with the inhibition of cell proliferation in multiple cell lineages, including aortic cardiac smooth muscle cells (Su et al., 2020), cardiomyocytes (Mak et al., 2011), lymphocytes (Go et al., 2004; Xin et al., 2016) or cancer cells (Kim et al., 2018; Qin et al., 2017). In hepatocellular carcinoma cells, knockdown of TONEBP accelerated S-phase entry (Qin et al., 2017), suggesting that TonEBP can in some cases inhibit cell proliferation. The stimulatory effect of TonEBP-deleted NP cells on adjacent GP chondrocytes, which maintained normal (WT) expression of TonEBP in iK19^cKO mice, however, suggests that in NP cells, TonEBP either represses the expression or secretion of a pro-proliferative factor, or stimulates the expression or secretion of an anti-proliferative factor that acts in an autocrine manner in the NPs, and in a paracrine manner to affect chondrocytes of the vertebral GPs. Dahia et al. previously reported ex vivo evidence for a role of NP cells in maintaining AF structure via Hedgehog signaling (Dahia et al., 2012). These are important findings, as the mechanism and the mediators involved have potential to be targeted to stimulate postnatal NP and GP cell proliferation in the context of strategies aimed at regenerating diseased IVDs. We conclude from these results that TonEBP in NP cells during postnatal IVD growth is not required for their survival or fate maintenance, but governs the proliferation of NP and GP cells in a likely autocrine and paracrine manner, respectively.

Results from our set of TonEBP cKO mice indicated that this gene plays a major role in vertebral GP chondrocytes, and through this role not only affects the structure of the CEP/GP/AF, but also that of the NP and vertebral bony compartments. In contrast to NP cells, the loss of TonEBP in chondrocytes did indeed lead to cell death in the GP of both Col2^cKO and iCol2^cKO mice, indicating that TonEBP is required in this lineage, in a cell-autonomous manner, for cell survival. The short height of vertebral GPs makes it challenging to investigate how the loss of TonEBP alters the behavior of chondrocytes in this tissue. Although their developmental origin is different, the long bone epiphyseal growth plates in cKO mice will help define the role of TonEBP in the chondrocytic lineage. Nevertheless, the abnormal vertebral GP, NP, AF and vertebral trabecular and cortical bone compartments observed in all cKO mutants characterized by the loss of TonEBP in Col2-Cre+ or Acan-Cre+ cells indicate that TonEBP in cells of the CEP/GP/inner AF is central to the growth and structure of all compartments of the IVD-vertebral units.

Interestingly, the change in NP shape observed in 8- and 4-month-old adult Col2^cKO mice was detectable as early as 18 days (P21) following postnatal TonEBP deletion solely in the CEP/GP/inner AF of iCol2^cKO mice, and prior to major disruption of the CEP/GP/AF structure. These observations support the importance of the CEP/GP/inner AF to maintain NP structure, and the possible existence of soluble factors involved in a crosstalk between CEP/GP/AF and NPs. The cell of origin of the putative factors involved (cells from the CEP versus GP versus inner AF) is challenging to address because of the lack of Cre lines able to dissociate these three compartments, and the nature of the mediators involved also remains to be determined. It is unlikely that the NP shape and cell density alterations detected in Col2^cKO and iCol2^cKO mice stem from migrating cells originating from surrounding inner AF or CEP/GP compartments, as we did not observe any labeled cells within the NP using lineage tracing in Col2-CreERT^Tg/+; TdTomato^Tg/+ mice. It is possible that the death of GP KO chondrocytes or a change in their maturation through the vertebral GP alter the nature or amount of factors produced by these cells that are necessary for the maintenance of the NP phenotype or the composition of their ECM (which also could contribute to the increased proportion of large NP cells observed in Col2^cKO and iCol2^cKO mice). The postnatal deletion of Ihh in vertebral GP chondrocytes in Col2-CreER^Tg/+; Ihh^f/f cKO mice, similarly to TonEBP Col2^cKO, iCol2^cKO and iAcan^cKO mice, leads to enlarged NPs (Maeda et al., 2007), and, in that case again, whether GP-derived IHH or the loss of GP cells and subsequent changes in ECM mechanical or diffusion properties in cKO mice contribute to their NP phenotype remains to be determined.

The mechanical properties of COL2-depleted cartilages are thought to be insufficient to counteract NP proteoglycan swelling pressure, which may prevent the proper ‘squeezing’ of NP cells into the intervertebral segments to form the NPs during embryonic development in Col2a1-null mice (Aszodi et al., 1998). The low COL2 content in the GP ECM of Col2^cKO and iCol2^cKO mice may thus also be mechanically suboptimal for the circumferential constraint of NP cells and for maintaining the typical elliptical shape of the postnatal NPs. The loss of TonEBP in the CEP/GP and/or inner AF also disrupts AF structure, and thereby the mechanical compression of the NP, leading to its abnormal shape in Col2^cKO, iCol2^cKO and iAcan^cKO mice.

Another mechanism to be explored to explain the impact of the GP/CEP on the NP is that the loss of TonEBP in the CEP/GP induces a change in the permeability of this barrier tissue, thus altering the diffusion of water and/or nutrients/metabolites. This model is supported by the effects of experimental MMP8 treatment of human CEPs, which led to an increase in solute intake in the CEP (Dolor et al., 2019). Such changes in CEP solute intake could theoretically impact NP cells by altering NP cell extracellular tonicity or metabolism. Wong and collaborators also reported that a greater amount of collagen and ACAN (as well as more mineral and less cross-link maturity) was associated with low CEP diffusivity (Wong et al., 2019), so the reduction in COL2 ECM content in both Col2^cKO and iCol2^cKO mice may increase CEP diffusivity.

Although TonEBP cKO mice were not analyzed at a time point relevant to aging, these results have potential relevance to IVD degeneration. The vast majority of studies related to IVD degeneration and low back pain have been, for good reasons, NP-centric. Our results in this comparative analysis of TonEBP cKO mice, along with pioneering data by others (Adams and Dolan, 2012; Adams et al., 2000; Bonnheim et al., 2022; Habib et al., 2023; Vergroesen et al., 2015), support the notion that disc degeneration and low back pain may stem from abnormalities in any of the IVD compartments (not only the NP), and that variability in clinical presentations may stem from which compartment the condition arises from. This implies, acknowledging possible differences between mice and humans, that CEP/GP/AF and not only NPs, could be targeted for a more personalized management of IVD degeneration.

Generation of the TonEBP^{ex6-15/ex6-15} floxed allele was performed by inserting a new loxP site in the existing TonEBP^exon6/exon6 floxed allele (Kuper et al., 2014). Briefly, a gRNA (5′-GAATATCAGGCTAGCATATA-3′, Synthego) was used to induce a Cas9-mediated cleavage after the stop codon of TonEBP (intron 15) for insertion of a LoxP sequence. The gRNA was selected using the Wellcome Trust Sanger Institute Genome Editing website (https://www.sanger.ac.uk/htgt/wge/crispr/536662274) as previously described (Lanza et al., 2018). To repair the double-strand break and insert the LoxP sequence, we used a single-stranded DNA oligonucleotide (ssODN) (5′-TCTTTCATTTAGAGTGACAGAATATCAGGCTAGCATAATAACTTCGTATAGCATACATTATACGAAGTTATTAAGGCCCTAGATCCAATCATAGTAGAAAAGGAGAGCATGAAAGAAGAAAAAATTTTACAGAATAACAAATCTTATAATCTGGGGCTATAT-3′). Donor ssOligos were purchased as custom Ultramer^® oligonucleotides (Integrated DNA Technologies). The Cas9 mRNA was purchased from Thermo Fisher Scientific (A25640). All gRNAs were re-analyzed by Nanodrop prior to assembling the microinjection mixtures. The CRISPR reagent mixture consisted of Cas9 mRNA (100 ng/μl), sgRNA (20 ng/μl) and ssOligo (100 ng/μl) in a final volume of 60 μl 1×PBS (RNAse-free).

For the microinjection of CRISPR/Cas9 reagents, TonEBP^exon6/exon6 female mice, 24-32 days old, were injected with 5 IU/mouse of pregnant mare serum, followed 46.5 h later with 5 IU/mouse of human chorionic gonadotropin and mating to TonEBP^exon6/exon6 males. Fertilized oocytes were collected at 0.5 days post-coitum for microinjection. The sgRNA/Cas9/ssOligo mixture was microinjected into the cytoplasm of at least 200 pronuclear-stage zygotes. Injected zygotes were transferred into pseudopregnant ICR females on the afternoon of the injection (approximately 25-32 zygotes per recipient female).

To genotype the F0 litters and assess the correct insertion of the new LoxP site, we used the following genotyping primers: Forward: 5′-GTCACTCTGAATGACAGCTTAGA-3′, and Reverse 5′- ACTGAACCTAGGGCGTCAAA-3′, with expected size 236 bp for the WT allele and 270 bp for the mutated allele. Sanger sequencing was performed on genomic DNA from the purified PCR to verify correct sequence insertion. Homozygous founders were viable, fertile, and bred normally.

We crossed TonEBP^{ex6-15/ex6-15} with multiple Cre transgenic mice, including Cmv-Cre (Schwenk et al., 1995) (here called TonEBP ^Δ/Δ), Col2-Cre (Ovchinnikov et al., 2000) (here called Col2^cKO), Col2-CreERT (Nakamura et al., 2006) (here called iCol2^cKO), Krt19-CreERT (Means et al., 2008) (here called iK19^cKO) and Acan-CreERT2 (Henry et al., 2009) (here called iAcan^cKO). For Cre transgene induction in inducible Cre lines, Tam (Sigma-Aldrich, T5648) was first dissolved at a concentration of 100 mg/ml in 100% ethanol and then diluted to a final concentration of 10 mg/ml into sterile vegetable oil. Tam was injected subcutaneously at P3 and P5 (100 µg/gram of body weight) to both WT and cKO mice. This dose and regimen induced efficient recombination in expected tissues (Fig. 5A).

To assess Cre activity and tissue specificity, Cre^Tg/+; TonEBP^ex6-15/^ex6-15 mice were crossed with the Rosa26Tomato reporter transgenic line (Madisen et al., 2010).

Genomic DNA from ear clips was used for PCR genotyping. Primers used to assess allelic recombination by PCR and to verify Cre specific transgenic mice (Couasnay et al., 2019) are listed in Table S1. For determination of embryonic ages, noon on the day of the postcoital plug was taken to be E0.5. All mouse procedures were approved by the institutional BCM IACUC board (protocol AN-7143).

Radiographs were obtained by using a digital cabinet X-ray system (Kubtec X-pert 80). The height of vertebral bodies was measured at the dorsal, midline and ventral regions along the coronal plane and averaged. DHI was calculated as previously described (Tajerian et al., 2011). For µCT analyses, each spine was immobilized with sponges in a μCT specimen tube (Scanco Medical AG) filled with 70% ethanol, and the L3/L4 region was imaged using a μCT40 scanner (Scanco Medical AG) according to the ASBMR guidelines (Bouxsein et al., 2010). Scan settings were the same for each specimen (55 kVp, 145 μA, isotropic voxel size of 12 μm), and a Gaussian image filter with sigma=0.8 and support=1.0 to segment bone from nonmineralized tissue was applied. Standard trabecular vertebral parameters were determined using the manufacturer's built-in software (μCT Evaluation Program V6.6; Scanco Medical AG). Cortical thickness and densitometric parameters were determined based on the approach described by Harris et al. (2020). Briefly, the 3D segmented image of the trabecular bone was subtracted from the 3D segmented image of the whole vertebrae between endplates to produce a 3D segmented image of the thin cortex. This cortex was evaluated using the Scanco mid-shaft script to determine shell thickness and shell tissue mineral density.

Lumbar spines were fixed in 4% paraformaldehyde in 1×PBS for 16 h at 4°C and decalcified in 0.5 M EDTA pH 8.0 for 3 weeks before being embedded for paraffin or cryosection preparation. Following decalcification, all samples were washed twice for 30 min each in distillated water. All sections were performed in the coronal plane of the lumbar spine. For paraffin, 4-µm-thick serial sections were obtained using a Leica RM2255 microtome. Specific antigen retrieval conditions and a list of antibodies used for immunofluorescence are given in Table S2. To detect apoptotic cells, the DeadEndTM fluorometric TUNEL system (Promega, G3250) was used, according to the manufacturer's instructions. TUNEL-positive cells were counted per selected region of interest (ROI) of the IVD using the Freehand Tool in ImageJ software and normalized to the number of DAPI-positive nuclei in this ROI.

For BrdU incorporation, mice were injected intraperitoneally with 100 µg BrdU (Sigma-Aldrich, 19-160) per gram of body weight and sacrificed 2 h later. Following fixation in 4% paraformaldehyde and embedding, paraffin sections were used for detection of BrdU by immunohistochemistry following manufacturer's instructions (Cell Signaling Technology, 5292S). BrdU-positive cells were counted per selected ROI of the IVD using the Freehand Tool from ImageJ software and normalized to the number of DAPI-positive nuclei in this ROI.

Staining of sections with Safranin O was carried out using standard histological procedures. Briefly, re-hydrated paraffin sections were incubated in Weigert iron Hematoxylin (VWR, 100504-410) for 5 min, rinsed under running tap water for 5 min, stained with a 0.02% Fast Green (Sigma-Aldrich, F7258) solution in water for 20 min, dip-rinsed for 10 s in 1% acetic acid solution, followed by 10 min incubation in a solution of 0.5% Safranin O (Sigma-Aldrich, S8884) and three washes in 95% ethanol for 10 s each wash. Slides were dehydrated, mounted with Permount (Fisher Scientific, SP15-100) and scanned using an IX PICO device (Molecular Devices) at 10× magnification.

For cryosectioning, spines were incubated for 16 h in 15% and then 30% sucrose (Tissue–Tek^®, Fisher Scientific 4583) overnight at 4°C. Cryosections were prepared at a thickness of 10 µm and stored at −80°C. Cryosections were washed twice in PBS and counterstained with 50 µg/ml Hoechst 33258 (Thermo Fisher Scientific, H3569) for 15 min. The slides were mounted using FluorSave (Fisher Scientific, 34-578-920ML) and imaged using an IX PICO device (Molecular Devices) at 10× magnification.

Lumbar intervertebral disc (L4-L5) degeneration scoring was performed by two trained experimenters who were unaware of the genotypes on Hematoxylin and Eosin-stained coronal sections, as described by Melgoza et al. (2021). Results are represented in a heat map scoring system for ease of visualization of results, and as the sum of each score per IVD compartment, namely NP, AF, CEP and interface boundaries (Int.) for statistical comparison between genotypes.

Immunohistochemistry sections were analyzed using ImageJ software. After converting each channel to grayscale, the boundaries of the different IVD compartments (NP, CEP/GP, AF) were digitally traced using the Freehand tool. These images were then thresholded to create binary images. Staining intensities were analyzed using integrated density level for each channel (sum of all the pixels within each ROI) and normalized to the DAPI signal in this ROI. Geometric measurements for each IVD compartment were performed with ImageJ software and the Freehand Line tool. CEP, GP and NP height was measured on plane-matched IVD coronal sections, through three distance measurements positioned on the left, central and right region of each compartment that were averaged and compared between each genotypes. NP area included the entire NP space, including NP cells and the void area caused by histological processing. Measurements were performed by experimenters unaware of the genotype. The NP aspect ratio was determined as width divided by height of the NP area measured on Safranin O-stained images of coronal sections of lumbar IVD (L5-L4), as described by Choi et al. (2018). Cell surface distribution was determined by measuring the surface of the cytoplasm for each cell in two averaged NPs (L5-L4 and L4-L3) per mouse, and frequency distribution was calculated with a bin width of 50 µm², separating surfaces below or above 50 µm². Column index was calculated from a modified method as described by Killion et al. (2017). False color conversion for people with color blindness can be performed using free software, such as Visolve (https://www.ryobi.co.jp/products/visolve/en/visolve.html).

Primary chondrocytes were isolated from the ribs of P3 WT and Col2^cKO mice as previously described (Gosset et al., 2008) and cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. For hypertonic stress stimulation, growth medium was supplemented with 75 mM of NaCl for 8 h to bring osmolarity to 450 mOsM. NP isolation for RNA extraction was performed as previously described (Bratsman et al., 2019) following Tam injection at P3 and P5 and pup euthanasia at P21.

Total RNA was isolated from cells or NP tissues using the TRIzol reagent (Thermo Fisher Scientific, 15596-018) and RNeasy Mini Kit (QIAGEN, 74104), respectively, according to manufacturer's instructions. RNAs were reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, 4368813) following the manufacturer's instructions. Real-time PCR (qPCR) was performed on a Bio-Rad CFX96 using SsoAdvanced™ Universal SYBR^® Green SuperMix (Bio-Rad, 1725272). Primer efficiency was determined using a standard curve with 1:4 dilution and specificity of amplification was verified from the melting curve analysis. All primer sets used had an amplification efficiency between 95 and 100%. Expression of target genes was normalized to Tbp (Tata-box binding protein) or Gapdh (glyceraldehyde 3-phosphate dehydrogenase) expression levels, two genes for which expression was not affected by genotype. Relative gene expression levels were calculated with the 2ΔΔCt method (Schmittgen and Livak, 2008). The sequences of the primers used in this study are listed in Table S1.

M-MLV Reverse Transcriptase PCR (Thermo Fisher Scientific, 28025013) for isoform detection was performed as per manufacturer's instructions. Briefly, 1 μg of total RNA from WT and Col2-CreTg/+; TonEBP^exon6/exon6 primary chondrocytes were reverse transcribed into single-stranded cDNA and used to perform PCR using the primers listed in Fig. S1C.

Cells were lysed for 30 min in ice-cold lysis buffer (150 mM NaCl, 10 mM Tris HCl, pH 8, 5 mM EDTA, 1% NP-40, 0.1% SDS, 0.5% sodium deoxycholate) containing a protease inhibitor cocktail (Sigma-Aldrich, P8340-1ML). Protein extracts were resolved on a 4-15% Mini-PROTEAN^® TGX™ gel (Bio-Rad, 4561084) and electroblotted onto polyvinylidene fluoride membranes. After blocking with 5% non-fat dry milk in Tris-buffered saline with 0.05% Tween 20 (TBST), blots were probed with primary antibodies against TONEBP (Abcam, ab3446, 1/1000), GAPDH (Cell Signaling Technology, 2118S, 1/5000) overnight at 4°C. Secondary rabbit and mouse horseradish peroxidase-conjugated antibodies were used (Cell Signaling Technology, 7074S and 7076S; 1/1000) and revealed with Clarity™ Western ECL Substrate (Bio-Rad, 1705060) and imaged with a ChemiDoc Imaging SystemTM (Bio-Rad)

Data are expressed as mean±s.d. Unless otherwise stated, experiments were repeated at least three times. Depending on the number of samples and their distribution, results were analyzed using a one-way ANOVA, unpaired Student's t-test or a nonparametric Mann–Whitney test, as indicated in legends. P<0.05 was considered significant. Statistical analysis was performed with the GraphPad Prism 5.0 software (GraphPad Software, Inc.).

We thank Drs J. Haney and D. Lanza (BCM Genetically Engineered Rodent Model Core) for their help in the generation of the TonEBP^{ex6-15/ex6-15} floxed allele. We also thank Dr J. Titze for sharing the TonEBP^ex6 floxed mice, Mr S. Uppuganti, O. E. Ruiz and Dr J. Nyman for their help with vertebral μCT analyses, Dr C. Dahia for her training in IVD histopathology scoring, and S. Beck, L. Beck, X. Yang and E. Tsouko for their critical feedback on the manuscript. Resources accessed through Genetically Engineered Rodent Models Core were supported by a National Institutes of Health-National Cancer Institute grant (P30CA125123) to the Dan L. Duncan Comprehensive Cancer Center.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.202354.reviewer-comments.pdf