Protein kinase D3 modulates MMP1 and MMP13 expression in human chondrocytes

Many catabolic stimuli, including interleukin-1 (IL-1) in combination with oncostatin M (OSM), promote cartilage breakdown via the induction of collagen-degrading collagenases such as matrix metalloproteinase 1 (MMP1) and MMP13 in human articular chondrocytes. Indeed, joint diseases with an inflammatory component are characterised by excessive extracellular matrix (ECM) catabolism. Importantly, protein kinase C (PKC) signalling has a primary role in cytokine-induced MMP1/13 expression, and is known to regulate cellular functions associated with pathologies involving ECM remodelling. At present, substrates downstream of PKC remain undefined. Herein, we show that both IL-1- and OSM-induced phosphorylation of protein kinase D (PKD) in human chondrocytes is strongly associated with signalling via the atypical PKCι isoform. Consequently, inhibiting PKD activation with a pan-PKD inhibitor significantly reduced the expression of MMP1/13. Specific gene silencing of the PKD isoforms revealed that only PKD3 (PRKD3) depletion mirrored the observed MMP repression, indicative of the pharmacological inhibitor specifically affecting only this isoform. PRKD3 silencing was also shown to reduce serine phosphorylation of signal transducer and activator of transcription 3 (STAT3) as well as phosphorylation of all three mitogen-activated protein kinase groups. This altered signalling following PRKD3 silencing led to a significant reduction in the expression of the activator protein-1 (AP-1) genes FOS and JUN, critical for the induction of many MMPs including MMP1/13. Furthermore, the AP-1 factor activating transcription factor 3 (ATF3) was also reduced concomitant with the observed reduction in MMP13 expression. Taken together, we highlight an important role for PKD3 in the pro-inflammatory signalling that promotes cartilage destruction.


Introduction
The progression of joint diseases such as rheumatoid arthritis (RA) and osteoarthritis (OA) is associated with inflammation, whereby inflammatory mediators released by infiltrating PLOS  immune cells as well as resident joint cells induce altered gene expression that promotes extracellular matrix (ECM) degradation [1]. Cytokines such as interleukin (IL-)1 and tumour necrosis factor α (TNFα) are key mediators in the inflammatory responses in destructive joint disease [2,3], and we have previously demonstrated enhanced cartilage ECM catabolism by these mediators when combined with IL-6-family cytokines such as oncostatin M (OSM) in vitro and in vivo [4][5][6][7][8]. This co-operative exacerbation in catabolic potential is likely to occur in many inflammatory milieu, such as the arthritides, to promote cartilage ECM catabolism. Normal articular cartilage is maintained by a single resident cell type, the chondrocyte, which preserve homeostasis by regulating the expression of ECM components and catabolic factors such as the matrix metalloproteinases (MMPs). Collectively, the MMPs can degrade all ECM macromolecules, and during inflammatory joint diseases stimulated chondrocytes secrete elevated levels of MMPs which, once activated, mediate proteolysis of tendon, bone and cartilage [9,10]. MMP-1 and MMP-13 are the collagenolytic MMPs most strongly associated with cartilage collagenolysis, a key proteolytic event in joint diseases since it is essentially irreversible [11]. Indeed, MMP-13 is the most potent collagenase with respect to type II collagen [12], the major structural collagen in articular cartilage. Potent pro-inflammatory stimuli such as IL-1+OSM activate a complex array of signal transduction pathways, shown to be common to a wide variety of pro-inflammatory stimuli [13], which together markedly enhance the induction of MMP1/13 in human chondrocytes [7,[13][14][15]. The cFos/cJun activator protein-1 (AP-1) transcription factor complex is critical for MMP gene expression (see [16] and references therein), although we recently reported a role for the AP-1-binding factor activating transcription factor 3 (ATF3) in selectively mediating MMP13 expression which was nevertheless AP-1-dependent [13]. Furthermore, we have also previously demonstrated the importance of signal transducer and activator of transcription (STAT)3, phosphatidylinositol-3' kinase (PI3K/Akt) and protein kinase C (PKC) signalling pathways [17][18][19]. The atypical PKC isoform PKCι has been shown to modulate MMP1/13 induction following IL-1+OSM stimulation [19], but the identity of the downstream signalling components it affects remains unknown, as do potential points of cross-talk.
Protein kinase D (PKD) comprises a family of serine/threonine protein kinases that belong to the Ca 2+ /calmodulin-dependent kinase superfamily. Three isoforms exist in humans: PKD1, PKD2 and PKD3 (reviewed in [20]). Diverse cellular functions, including cell survival/proliferation and apoptosis, plasma membrane-directed transport, metastasis and inflammation have been reported [20][21][22][23][24] for PKD isoforms. With respect to cartilage remodelling, PKD is reported to regulate both the expression and activity of several MMPs. Studies have shown that PKD1 down-regulates the expression and activity of multiple MMPs in breast cancer cells lines, particularly the gelatinase MMP-9 [22]. Conversely, gene silencing of PKD2 and PKD3 leads to a decrease in the activity of MMP-9 in prostate cancer, suggesting MMP gene regulation by distinct PKD family members is cell-type specific. PKD3 has been reported to regulate nuclear localization and activity of histone deacetylases 5 and 7 (HDAC5/7) [25]. OA cartilage exhibits elevated HDAC7 levels, whilst HDAC7 depletion in SW1353 human chondrosarcoma cells strongly suppressed IL-1-dependent induction of MMP13 [26]. PKD3 can be activated independently of PKC via phosphorylation of tyrosine residues within the pleckstrin homology domain following oxidative stress, which is a feature of OA [27]. Moreover, PKD isoforms have also been shown to be a downstream target of G-protein-coupled receptors, via increases in phospholipase C activity and auto-phosphorylation [28]. Such receptors play roles in innate and adaptive immunity, and have been implicated in the pathology associated with arthritis [29][30][31].
PKD3 is ubiquitously expressed in adult human tissues [32] and has been shown to be expressed during skeletogenesis in mice, notably in the cartilage primordia of bones [33], which are areas of active ECM turnover (via MMP-13) as cartilage is remodelled prior to mineralization. Interestingly, phenotypic analysis of a mutant mouse strain harbouring a gene-trap deletion of the PKD3 gene (prkd3) revealed a mild skeletal abnormality including decreased mean trabecular bone volume and thickness [34], suggestive of reduced remodelling in cartilage primordia. In this context, downstream PKD3 signalling may also be involved in cytokine-induced MMP13 expression in human chondrocytes. Herein, we show that PKD3 is indeed involved in pro-inflammatory signalling via modulating the expression of the AP-1 genes FOS and JUN, and enhancing MMP expression. PKD3 activity also stimulates the induction of ATF3 which is specifically involved in cytokine-induced MMP13 expression in human chondrocytes.

Materials
All chemicals were obtained from Sigma Chemical Co (Poole, UK) unless otherwise stated and of the highest purity available. All cytokines were recombinant human. IL-1α was a generous gift from Dr Keith Ray (GlaxoSmithKline, Stevenage, UK). OSM was prepared in-house using expression vectors kindly provided by Prof. J. Heath (University of Birmingham, UK) and methods described [35]. Gö6983 was from Merck Chemicals (Nottingham, UK). PKD inhibitor kb NB 142-70 was from Tocris Bioscience (Bristol, UK). The STAT3 inhibitor S31-201 was from Selleckchem (TX, USA). Kinase inhibitors and small interfering (siRNA) reagents were screened for toxicity using the Toxilight assay of adenylate kinase release (Lonza, Wokingham, UK), and always used at concentrations that did not affect cell viability over the assay period.

Chondrocytes
Human chondrocytes were obtained by the enzymatic digestion of macroscopically normal articular cartilage from OA patients undergoing joint replacement surgery as described [36]. All subjects gave informed consent and the study was approved by the Newcastle and North Tyneside Joint Ethics Committee (REC 14/NE/1212). Chondrocytes were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% foetal bovine serum (FBS), 100 IU penicillin, 100 μg/ml streptomycin, 40 U/ml nystatin.

Cell fractionation and immunoblotting
Chondrocyte lysates were prepared as described previously (18) and stimulated with IL-1 (0.2 ng/ml) ± OSM (10 ng/ml) for up to 60 min. In some experiments, chondrocytes were subjected to subcellular fractionation using NE-PER Nuclear and Cytoplasmic Protein Extraction Kit or Subcellular Protein Fractionation Kit (both from ThermoFisher Scientific, Loughborough, UK). GAPDH was used as a loading control for whole cell lysates whilst MEK2 and lamin A/C were used for cytoplasmic and nuclear extracts, respectively. Lysates or fractions were resolved by SDS-PAGE, transferred to PVDF membranes and subsequently probed using the following anti

Gene silencing
PRKCZ and PRKCI siRNAs were Dharmacon ON-TARGETplus™ (ThermoFisher Scientific), comprising 4 specific siRNA complexes [16], whilst validated MISSION™ siRNA reagents (Sigma) were used to silence: Primary human chondrocytes were prepared and cultured as above, and transfected as described previously (18). After transfection, cells were serum-starved for 24 h prior to stimulation as indicated. Depletion of gene-specific mRNA levels was calculated by comparison of expression levels with cells transfected with 100 nM siCONTROL (siCon: non-targeting siRNA #2, Cat. 001210-02; Dharmacon). GIPZ lentiviral vector containing non-targeting control shRNA (shCon) or PRKD1 shRNA (shPRKD1: V2LHS-170466) were purchased from OpenBiosystems (Thermo Fisher Scientific). Packaging and envelope plasmids pCMV-dR8.91 and pMD2.G were a kind gift from Prof. Nick Reynolds (Newcastle University). The HEK293T producer cell line (SBI, California, USA), maintained in DMEM plus 10% FBS, was employed for packaging. On the day of transfection, medium was changed to DMEM containing 10% heat-inactivated FBS. Packaging, envelope and shRNA plasmids were co-transfected using JetPEI transfection reagent (Polyplus, Illkirch, France), according to the manufacturer's instructions. Supernatant was harvested 72 h post-transfection and filtered (0.45 μm). Viral particles were then concentrated 50-fold using a Lenti-X concentrator (Takara Bio Europe/ Clontech, Saint-Germain-en-Laye, France), and viral titre calculated using the Lenti-X qRT-PCR Titration Kit (Takara). Multiplicity of infection (MOI) was then calculated via serial dilution. Primary chondrocytes were prepared and cultured as above, and exposed to lentiviral infection at a MOI of 30 for 72 h. Transduction was assessed using confocal microscopy for green fluorescent protein (GFP). After transduction, cells were serum-starved for 24 h then stimulated as indicated.

Real-time PCR of relative mRNA levels
Primary human chondrocytes were stimulated with IL-1 (0.05 ng/ml) ± OSM (10 ng/ml) for 1 h to measure FOS, JUN and ATF3, or 24 h for MMP mRNAs. RNA was stabilised in cell lysates in a 96-well format, cDNA synthesised and real-time PCR assays conducted using conditions described previously [18]. Primer and probe sequences are as previously detailed [13,19]

Statistical analyses
Statistical differences between sample groups were assessed using one-way analysis of variance (ANOVA) with a post-hoc Bonferroni's multiple comparison test or Student's 2-tailed unpaired

Inhibition of PKD phosphorylation curtails IL-1+OSM induction of MMP1 and MMP13 in human articular chondrocytes
We previously demonstrated roles for PKCι and PKCz in IL-1+OSM-mediated expression of the key collagenases MMP1 and MMP13 in human chondrocytes [19]. Moreover, we have also shown both IL-1 and OSM induce PKD phosphorylation such that we sought to confirm that PKD was indeed a downstream substrate of PKC signalling. Cytokine-induced PKD phosphorylation was assessed in the presence of the Gӧ6983 (20 μM), a PKC inhibitor that does not inhibit PKD, which completely abolished both IL-1-and OSM-stimulated PKD phosphorylation (Figs 1A and S2). Assessment of the effect of gene silencing of PRKCI and PRKCZ (coding for PKCι and PKCz, respectively; S3A Fig) indicated that, of the atypical PKC isoforms, only PKCι played a role in this phosphorylation (Fig 1B and 1C).
Since combined PKC/PKD inhibition (using Gӧ6976) suppressed MMP expression and reduced cytokine-induced collagen release from cartilage [16], we confirmed that a specific PKD inhibitor (NB142-70; PKDi) also significantly reduced IL-1+OSM-induced expression of MMP1 and MMP13 expression ( Fig 1D). Together, these findings implicate a role for PKD in cytokine-induced MMP expression in human chondrocytes.

PKD3 depletion correlates with reduced MMP1 and MMP13 expression in human chondrocytes
To determine which of the three individual PKD isoforms accounted for the PKDi-mediated reduction in MMP expression, selective 'knockdown' of the PKD isoforms (S3B Fig) indicated that only PRKD3 silencing led to a significant decrease in the expression of both MMP1 and MMP13 (Fig 2). Expression of these MMPs was markedly enhanced with PRKD1 silencing but unaltered following PRKD2 depletion (S4A and S4B Figs, respectively). Since these data strongly implicated PKD3 as the only 'pro-inflammatory' moiety of PKD signalling for cytokine-induced MMP expression, specificity assessment of PRKD3 silencing for the other collagenolytic MMP species indicated that MMP8, but not MMP14, was also significantly suppressed (S5 Fig).

PKD3 depletion alters STAT signalling in IL-1+OSM-stimulated human chondrocytes
Analyses of the signalling pathways activated in human chondrocytes following PRKD3 silencing with IL-1+OSM stimulation revealed reductions in phosphorylation of all three MAPK groups but no apparent effect on the phosphorylation of Akt Ã S473 (Fig 3A and 3B). As dual phosphorylation of STATs occurs for full activity [37,38], we also observed reductions in Ser 727 phosphorylation of STAT1 and STAT3 (Fig 3A and 3B), important for transcriptional activity [37], whilst a more detailed analysis using subcellular fractionation indicated that although cytoplasmic (and indeed total) levels of Tyr 705 phosphorylation of STAT3 appeared to be relatively unaffected, reduced nuclear levels were observed (Fig 4A) following PRKD3 depletion. In contrast, total and nuclear STAT1 Ã Y701 was unaffected by PRKD3 silencing (Fig 4A).
To further confirm a role for STAT3 in IL-1+OSM-mediated MMP expression [17,19] the STAT3 inhibitor S31-201 significantly reduced both MMP1 and MMP13 expression in human chondrocytes (Fig 4B). Since PRKD3 silencing curtailed IL-1+OSM-mediated JNK activation, and in light of the previously reported PKD-dependent activation of JNK [39] and our previous report indicating no dependency for ERK [19], we determined whether the observed  effects on STAT phosphorylation could be mediated by JNK moieties. Gene silencing of JNK1 and JNK2 revealed that JNK2 appeared to have a role in the Ser 727 phosphorylation of both STAT1 and STAT3 following IL-1+OSM stimulation (Fig 4C and 4D). Protein kinase D3 modulates collagenase expression

PKD3 indirectly regulates AP-1 factors in IL-1+OSM-mediated MMP13 expression
Previous data have shown a role for STAT3 in the induction of the AP-1 factor cFos [17], and we show here that the effect of PRKD3 silencing on STAT3 phosphorylation is concomitant with a reduction in the expression of both FOS and JUN (Fig 5A) as well as cFos protein and phosphorylated cJun (Fig 5B and 5C). We have recently reported the involvement of ATF3 in cytokine-induced MMP13 expression [13], and assessment of ATF3 expression following PRKD3 silencing confirmed that it was indeed significantly reduced ( Fig 6A). This was further reflected in markedly reduced nuclear ATF3 protein in cytokine-stimulated human chondrocytes following PRKD3 gene silencing (Fig 6B and 6C).

Discussion
The signal transduction associated with potent pro-inflammatory stimuli is complex although it is emerging that many of the activated pathways are common to a multitude of inflammatory mediators [13]. The interactions and cross-talk serve to markedly enhance MMP induction in human chondrocytes which ultimately drive ECM cataoblism [6,7,13,14,18,19]. We have used the catabolic stimulus of IL-1+OSM as a model inflammatory stimulus to investigate the signalling that underpins the expression of collagenolytic MMPs, in particular MMP13 due to its high capacity to degrade the major structural component of cartilage, type II collagen [12]. Our own data and those of others [13,17,40] have shown a strong dependence on AP-1 complexes of cFos/cJun in order for MMP gene expression to occur, and this has been linked to AP-1 consensus motifs in the proximal promoters of both MMP1 and MMP13 (see [16]). Phosphorylation of PKC isoforms in human chondrocytes occurs following stimulation with IL-1 or OSM, whilst their silencing has been shown to reduce both MMP1 and MMP13 expression as well as the induction of FOS [19]. These stimuli also promote phosphorylation of the downstream PKC substrate, PKD [19]. Herein we extend these observations through the identification of PKCι as the atypical PKC isoform that mediates downstream PKD phosphorylation. Moreover, pharmacological inhibition of PKD species significantly reduced IL-1+OSM-induced MMP1 and MMP13 expression. However, the PKD inhibitor kb NB142-70 does not discriminate between PKD isoforms, and indeed many early studies on PKD reported findings relating to PKD as a single 'entity' rather than considering the three individual protein kinase isoforms. In this context, only silencing of PKD3 (PRKD3) mimicked the inhibition of PKDi on MMP1/13 expression thus highlighting its role in pro-inflammatory signalling in chondrocytes. Moreover, these data also indicate that stimulation of two specific cytokine receptor pathways (IL-1 and OSM, respectively) converges on PKD3 activation via PKCι (Fig 7).
Earlier findings indicated that STAT3 is a potent mediator of cytokine-induced MMP expression in chondrocytes [17,19]. Evidence is provided here too indicate that PRKD3 knockdown also Protein kinase D3 modulates collagenase expression modulates the activation of STATs 1 and 3 via reducing Ser 727 phosphorylation, a requirement for maximal transcriptional activation of target genes [37]. Notably, STAT1/3 tyrosine phosphorylation was unaffected and, since nuclear import of STAT3 is independent of Tyr 705 phosphorylation [38], the reduced nuclear STAT3 was presumably due to a PKD3-dependent alteration in the importin-α/importin-β1-Ran pathway [38]. Furthermore, the catalytic activity of PKD3 has been shown to regulate its own nuclear import through auto-phosphorylation and/or interaction with other proteins [41], such that PRKD3 depletion prior to stimulation could markedly restrict subsequent nuclear trafficking to limit STAT3-dependent gene transcription. Indeed, PRKD3 silencing led to a reduction in the expression of both cFOS and cJUN at both the mRNA and protein levels, as previously shown to be critical for MMP1 and MMP13 expression [19]; the induction of FOS has already been attributed to STAT3 activation [19]. Moreover, phosphorylation of cJun by JNK (Ser 63/93 ) as well as by PKD isotypes (Ser 58 ) has been reported [42], thus generating the transcriptionally active AP-1 (cFos/cJun) complexes associated with MMP1/13 transcription [16]. The bone abnormalities in a gene-trap prkd3 deletion mouse [34] also support a key role for PKD3 in the regulation of MMP1/13 expression in chondrocytes and cartilage.
Of the other collagenolytic MMPs, MMP14 does not possess a promoter-proximal AP-1 element [43,44] and was unaffected following PRKD3 silencing as we would predict. Bioinformatics has identified a potential proximal AP-1 for MMP8 [44] and, in line with our findings of reduced AP-1 expression, this MMP was also supressed. However, although many non-canonical AP-1 sequences have been shown to be functional, other reports do not support the presence of a proximal AP-1 element for MMP8 [43,45] which would indicate other regulatory mechanisms being impacted by PKD3 activity following cytokine stimulation.
We provide further evidence for a critical role for cFos in the cytokine-mediated upregulation of MMP1 and MMP13 since the effect of PRKD3 silencing on FOS expression was specific. Importantly, silencing of either PRKD1 or PRKD2 had no modulatory effects on MMP1/13, FOS or JUN (S4 and S6 Figs).
Despite a demonstrable role for the immediate early gene FOS and AP-1(cFos/cJun)-dependent induction of MMP1 and MMP13 (reviewed in [46]), we have also reported an absence of cFos binding to the MMP13 promoter indicative of an indirect role for such cFos/cJun complexes [13]. Indeed, cytokine-induced MMP13 expression involved the bZIP transcriptional regulator ATF3, the expression of which was AP-1(cFos/cJun)-dependent [13]. The obvious corollary therefore was that ATF3 expression would be supressed following PRKD3 silencing thus reducing its ability to transcriptionally regulate MMP13. Our findings herein confirm this mechanism and further highlight that although AP-1 (cFos/cJun) complexes are critical for cytokine-induced expression of genes such as MMP13, their role is indeed indirect: this is also most likely the case for MMP1 [13] although no such regulatory factors have yet been reported. Furthermore, earlier findings indicate that the magnitude of FOS expression following proinflammatory stimulation appears to directly influence the subsequent levels of MMP expression [17] which requires the expression of AP-1-dependent transcriptional regulators [13]. This is probably the case for other examples of complex pro-inflammatory stimuli that markedly Protein kinase D3 modulates collagenase expression enhance MMP expression (eg. [5,6,8,14,17,18]. For cytokine-induced MMP13 we further confirm the involvement of ATF3, the expression of which is critically reliant upon phosphorylation of PKD3.  Proposed mechanism for the involvement of PKD3 in cytokine-stimulated MMP13 induction in human chondrocytes. Stimulation by IL-1 and OSM mediates the phosphorylation and activation of PKCι, whilst OSM stimulation leads to activation of the JAK/STAT pathway including Tyr 705 phosphorylation of STAT3. PKCι phosphorylates/activates ERK as well as PKD3 at Ser 731/735 , leading to PKD3 Ser 916 autophosphorylation. In turn, this leads to the phosphorylation of JNK which induces Ser phosphorylation of STAT3, priming it for transcription, as well as stabilising and phosphorylating cJun (PKD3 may also directly phosphorylate cJun). Both STAT3 and ERK (via Elk1 activation) lead to the induction of FOS, and when combined with phospho-cJUN drives the expression of various transcriptional regulators which regulate the resulting MMP induction including ATF3-dependent MMP13 expression. The scheme is based on the current study and input from Refs. (13,(17)(18)(19)22,(39)(40)(41)(42)(43)(44)(45)(46). https://doi.org/10.1371/journal.pone.0195864.g007

Conclusion
Our data indicate that PKD3 functions downstream of PKCι to affect the cytokine-mediated induction of MMP1 and MMP13 in human chondrocytes. Thus, PKD3 may represent a novel therapeutic target for consideration in the better management of inflammatory joint diseases, as well as other pathologies associated with aberrant collagenase expression and ECM catabolism.
Supporting information S1 Fig. Confirmation of antibody specificity using human chondrocyte lysates. The specificity of each antibody used in the study was confirmed using whole cell lysates, except for MEK2 and lamin A/C (cytoplasmic and nuclear extracts, respectively), prepared as described in the Methods from primary human articular chondrocytes either unstimulated or stimulated with IL-1 (0.2 ng/ml) ± OSM (10 ng/ml). Following SDS-PAGE, proteins were transferred to PVDF membranes and probed with the indicated antibodies. Full-length blots are presented to highlight the specific immuno-reactivity of each antibody (the arrow indicates the expected molecular mass). (PDF) Following transfection with siRNA specific for PRKD3 or a non-targeting siCon (100 nM), primary human articular chondrocytes were stimulated with IL-1+OSM as described in the Methods for 24 h. Cell lysates were then subjected to real-time PCR (n = 6; mean ± S.E.) for MMP8 and MMP14 (upper and lower panels, respectively) as described in Methods. Data are presented as relative expression levels normalised to 18S rRNA housekeeping gene, where ÃÃ , p 0.01, versus siCon (ns = not significant). All data are representative of at least three separate chondrocyte populations. (PDF) S6 Fig. Gene silencing of PRKD1 or PRKD2 does not supress cytokine-induced FOS and JUN expression. Following transduction with lentiviral shRNA (MOI = 30) specific to PRKD1 or shCon (A), or transfection with siRNA specific for PRKD2 or a non-targeting siCon (100 nM) (B), primary human articular chondrocytes were stimulated with IL-1+OSM as described in the Methods for 1 h. Cell lysates were then subjected to real-time PCR (n = 6; mean ± S.E.) for FOS or JUN (upper and lower panels, respectively) as described in Methods. Data are presented as relative expression levels normalised to 18S rRNA housekeeping gene, where ns = not significant versus the relevant Control. All data are representative of at least three separate chondrocyte populations. (PDF)