Lin28a overexpression reveals the role of Erk signaling in articular cartilage development

Osteoarthritis (OA) is a highly prevalent skeletal condition among the elderly (Dieppe and Lohmander, 2005). Although OA is considered a degenerative disease caused primarily by accumulated mechanical stresses, evidence suggests that other factors, such as genetic components, are also involved in OA pathogenesis. Recent genome-wide association studies (GWAS) have demonstrated associations between OA and genetic polymorphisms in several genes that regulate skeletal development, such as GDF5 and PTHLH (Chapman et al., 2008; Day-Williams et al., 2011; Evangelou et al., 2009, 2011; Miyamoto et al., 2007; Valdes et al., 2011; Zeggini et al., 2012). These data suggest that subtle changes in joint morphogenesis during cartilage development can influence joint homeostasis and development of OA later in life. However, compared with the catabolic process of the articular cartilage of synovial joints in OA, understanding of articular cartilage development and growth is still limited. Particularly, mechanisms regulating the cartilage matrix volume and chondrocyte number in the articular cartilage are largely unknown.

Chondrocytes that form the articular cartilage arise from a distinct progenitor population from those of the growth plate (Hyde et al., 2007; Koyama et al., 2008); articular chondrocytes differentiate from interzone cells of the joint primordium that express Gdf5 but not the matrilin gene (Matn1), whereas growth plate chondrocytes derive from Gdf5-negative, Matn1-positive cells. During articular cartilage development, cells expressing Gdf5 and/or the lubricin gene (Prg4) at the articular surface serve as progenitors for chondrocytes in the mature articular cartilage. Using a lineage-tracing technique, a recent study demonstrated that articular cartilage increases its volume through appositional growth (Kozhemyakina et al., 2015). In this process, chondrocytes in the articular surface (superficial zone chondrocytes; SZCs) slowly divide and produce progeny in deeper layers of the articular cartilage to increase the cartilage thickness. This study suggests that the number and function of SZCs are major determinants of the cartilage volume during development and growth stages. However, the mechanism that regulates the function and/or number of SZCs during synovial joint development remains elusive.

Small regulatory microRNAs (miRNAs) play crucial roles in skeletal development and homeostasis by suppressing target gene expression (Mirzamohammadi et al., 2014; Papaioannou et al., 2014). We have shown that miRNAs play important roles in chondrocyte proliferation, differentiation and survival, both in the growth plate and in the articular cartilage (Kobayashi et al., 2008, 2015). We have also shown that overexpression of Lin28a, an RNA-binding protein that inhibits the biogenesis of let-7 (Mirlet7 – Mouse Genome Informatics) miRNAs (Viswanathan et al., 2008), reduces chondrocyte proliferation, whereas it causes limb overgrowth when expressed in early limb mesenchymal progenitor cells (Papaioannou et al., 2013). Thus, Lin28a overexpression differentially regulates the proliferation of chondrocytes and their mesenchymal progenitor cells in a cell type-dependent manner. Based on these findings, we hypothesized that Lin28a overexpression regulates articular cartilage development by controlling the cell number of articular chondroprogenitor cells through let-7 miRNA-dependent and -independent pathways.

In this study, we show that overexpression of Lin28a increases the articular cartilage volume via upregulation of extracellular receptor-regulated kinase (Erk) signaling. We also show that expression of a constitutively active form of Kras mimics the effect of Lin28 overexpression. Inhibition of Erk signaling during embryogenesis normalized the articular cartilage in these mice, thus suggesting that Erk signaling during articular cartilage development primarily determines the articular cartilage volume in postnatal stages.

We have previously demonstrated that let-7 miRNAs regulate skeletal growth using mouse models overexpressing Lin28a, an inhibitor of let-7 miRNA biogenesis. Lin28a overexpression in limb mesenchymal cells resulted in overgrowth of the limb skeleton, whereas Lin28a overexpression in chondrocytes decreased their proliferation, causing mild growth impairment (Papaioannou et al., 2013). These results suggest that let-7 miRNAs differentially regulate tissue development and growth, dependent on cell types.

In order to test the effect of Lin28a overexpression, we used a Cre-mediated binary system (Papaioannou et al., 2013). To drive Lin28a expression, we first used Col2 (Col2a1)-Cre transgenic mice, in which Cre is expressed in early chondro-osteo progenitor cells in the mesenchymal condensation, as well as in mature articular and growth plate chondrocytes. Using RNA isolated from embryonic day (E) 13.5 Col2-Cre:Lin28a mouse limbs, and also from primary limb bud cells of Prx1 (Prdx1)-Cre:Lin28a mice, we confirmed Lin28a overexpression and concomitant suppression of let-7 miRNAs by quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis (Fig. S1). The magnitude of Lin28a expression and let-7 miRNA reduction was not as robust as when Lin28a expression was driven by Prx1-Cre transgenic mice. This is likely because Col2-Cre expression starts later during limb morphogenesis and is more restricted to chondro-osteo progenitors compared with Prx1-Cre expression. Nevertheless, Lin28a overexpression in Col2-postive cells increased the thickness of the articular cartilage of the mid-tibial plateau when assessed at the age of 3 weeks (Fig. 1A,B, Fig. S2). This phenotype was replicated when Lin28a was overexpressed using Gdf5-Cre transgenic mice, in which Cre is expressed in progenitor cells of articular chondrocytes and synovial cells, but not those of growth plate chondrocytes (Rountree et al., 2004) (Fig. 1C,D, Fig. S2). Immunostaining revealed normal distribution of type II collagen in Lin28a-expressing cartilage, suggesting that the quality of the cartilage matrix is maintained (Fig. S3).

It was previously shown that let-7 miRNAs suppress the Ras signaling pathway by direct binding to transcripts encoding Ras isoforms (Johnson et al., 2005). Ras activation mainly signals through the mitogen-activated protein kinase (MAPK) and phosphatidylinositol 3-kinase (PI3K; PIK3R1) pathways (Castellano and Downward, 2011; Papaioannou et al., 2016). Therefore, we determined the activities of Erk1/2 (also known as Mapk3/1) and PI3K using primary mesenchymal progenitor cells (Fig. 2A). Serum stimulation induced greater responses in Lin28a-overexpressing cells than in control cells in Erk signaling, assessed by phosphorylation of Erk1/2, and modestly in PI3K signaling, assessed by the level of phosphorylated Akt (Akt1) (p-Akt). We did not find significant alteration in the expression of Ras, a predicted target of let-7 miRNAs, upon Lin28a overexpression (Fig. 2B). Immunohistochemical analysis also confirmed upregulation of Erk phosphorylation in articular chondrocytes in Lin28a transgenic mice (Fig. S4).

Because Lin28a overexpression increases Erk signaling, we tested whether overactivation of Ras signaling could reproduce the cartilage phenotype of Lin28a transgenic mice. We overactivated Kras signaling in articular chondrocyte progenitor cells using Cre-inducible oncogenic mutant Kras (Kras^G12D) transgenic mice (Johnson et al., 2001). Kras overactivation increased the cartilage thickness to a similar extent to that induced by Lin28a overexpression (Fig. 3A,B). Erk phosphorylation was upregulated, whereas the p-Akt level was low, in both control and Kras^G12D transgenic mice (Fig. 3C). We also tested whether PI3K activation via genetic deletion of Pten, a negative regulator of PI3K signaling, would mimic the cartilage phenotype of Lin28a transgenic mice. Although Pten deletion in growth plate chondrocytes causes overgrowth (Ford-Hutchinson et al., 2007; Yang et al., 2008), we did not find a significant increase in articular cartilage thickness in Pten conditional knockout mice (Fig. S5). These data suggest that Erk signaling, but not PI3K signaling, positively regulated the articular cartilage thickness.

Based on these results, we hypothesized that the upregulation of Erk signaling is primarily responsible for the increased cartilage thickness caused by Kras activation. To test this, we injected U0126, a small chemical inhibitor of the Erk signaling pathway, into pregnant mothers during synovial joint development. Erk inhibition significantly reduced the articular cartilage thickness in mice with Kras overactivation (Fig. 4A,B). To determine whether Erk overactivation is also responsible for the cartilage phenotype of Lin28a transgenic mice, we treated pregnant mothers with U0126, using the same treatment protocol as that used on Kras transgenic mice. U0126 injection significantly reduced the cartilage thickness of Lin28a transgenic mice (Fig. 5A,B). These results suggest that overactivation of Erk signaling during cartilage morphogenesis increases cartilage thickness.

It has been shown that genetic components significantly contribute to the pathogenesis of OA. Recent GWAS indicate that OA is associated with genes involved in limb and cartilage morphogenesis, suggesting that proper development of joint tissues influences long-term cartilage health. However, understanding of the mechanisms that regulate development and growth of articular cartilage is still limited.

In this paper, we present evidence that Erk signaling during embryonic stages positively regulates the articular cartilage thickness. Overexpression of Lin28a in articular chondrocyte progenitors reduced let-7 miRNA expression, upregulated Erk signaling and increased the cartilage thickness, without affecting the cellular density. We show that this cartilage phenotype is mimicked by overexpression of a constitutively active oncogenic Kras, and that this cartilage phenotype is suppressed by inhibition of Erk signaling.

The cartilage phenotype caused by Lin28a or Kras activation is somewhat reminiscent of that of mice missing Errfi1 (also called Mig-6) (Pest et al., 2014; Shepard et al., 2013; Staal et al., 2014; Zhang et al., 2005). Errfi1 is expressed in the superficial zone of the articular cartilage and meniscus during cartilage development (Staal et al., 2014). Errfi1 encodes a negative regulator of epidermal growth factor receptor (Egfr) signaling that is mediated mainly through the MAPK and PI3K signaling pathways (Avraham and Yarden, 2011). Therefore, loss of Errfi1 increases Egfr and thus Erk activation in the articular cartilage (Staal et al., 2014). However, unlike Errfi1-deficient mice, mice with Lin28a overexpression or Kras overactivation do not exhibit an overt increase in cell density or cell proliferation in the 3-week-old articular cartilage or embryonic knee joint primordium (Fig. S6). It is possible that Ras-independent pathways of Egfr signaling are responsible for the continuous postnatal cell proliferation in Errfi1-deficient articular chondrocytes.

The finding that Lin28a overexpression increased Erk and PI3K signaling yet did not change Ras protein expression was unexpected, considering that RNAs encoding Ras proteins are conserved let-7 miRNA targets (Johnson et al., 2005), and that Lin28a overexpression significantly reduced let-7 expression. Let-7 miRNAs are also predicted to target other molecules associated with the Erk and PI3K signaling pathways, including Igf1 and its receptor Igf1r (Ghazal et al., 2015; Shen et al., 2014; Zhu et al., 2011). We found upregulation in Igf1r expression upon Lin28a overexpression in limb bud cells (Fig. S7A). However, genetic ablation of Igf1r was not able to mitigate the articular cartilage phenotype of Lin28a transgenic mice (Fig. S7B,C), suggesting that it is unlikely that Igf1r upregulation is responsible for the increased cartilage thickness. At the moment, the precise molecular mechanism by which Lin28a overexpression stimulates Erk and PI3K signaling is unclear. It is possible that modest upregulation of multiple let-7 miRNA targets cooperatively deregulates these signaling pathways. Because let-7 miRNAs are known to regulate oncogenic transcription factors, such as Mycn and Hmga2, they might indirectly regulate expression of genes that potentiate these signaling pathways (Balzeau et al., 2017; Nguyen and Zhu, 2015). It is also possible that let-7 miRNA-independent action of Lin28a contributes to the altered signaling (Jiang and Baltimore, 2016).

The observation that Erk inhibition was able to rescue the articular cartilage phenotype caused by Kras overactivation or Lin28a overexpression strongly suggests that the overactivation of Erk, but not PI3K, signaling increases the articular cartilage thickness. This is consistent with the finding that Pten deletion, a negative regulator of PI3K signaling, did not increase the cartilage thickness. It is noteworthy that Erk inhibition only during embryonic stages is enough to prevent the effect of Kras overactivation or Lin28a overexpression. Considering that Lin28a overexpression in chondrocytes in the growth plate reduces cell proliferation, whereas Lin28a overexpression in mesenchymal progenitor cells causes limb overgrowth (Papaioannou et al., 2013), the major cell population responsible for the articular cartilage phenotype of Lin28a transgenic mice is likely to be chondroprogenitor cells of articular cartilage, but not differentiated articular chondrocytes. Consistent with this notion, when we overexpressed Lin28a in postnatal stages using tamoxifen-inducible Cre lines, we did not find increases in the articular cartilage thickness (Fig. S8A,B).

It is also an interesting question as to whether endogenous Lin28a contributes to articular cartilage morphogenesis, because germ-line Lin28a deficiency causes growth and skeletal defects in embryos (Shinoda et al., 2013). Endogenous Lin28a is strongly expressed in developing limb buds, although its expression wanes around the time when chondroprogenitor cells first start to appear in the limb bud (Yokoyama et al., 2008). In this regard, experiments to conditionally delete Lin28a will provide insight into the role of the endogenous Lin28/let-7 axis in articular cartilage morphogenesis.

Together, these results suggest that Erk activation in chondroprogenitor cells during cartilage morphogenesis increases their number, and subsequently increases the articular cartilage thickness (Fig. 5C).

Cre-inducible Lin28 transgenic mice (CAG-lsl-Lin28) (Papaioannou et al., 2013), Cre-inducible oncogenic Kras knock-in mice (Johnson et al., 2001), Col2-Cre transgenic mice (Ovchinnikov et al., 2000), Gdf5-Cre transgenic mice (Rountree et al., 2004) and Prx1-Cre transgenic mice (Logan et al., 2002) were previously described. Floxed Igf1r mice, floxed Pten mice (Lesche et al., 2002), floxed Col2-CreER mice (Nakamura et al., 2006) and Prg4-CreER mice (Kozhemyakina et al., 2015) were previously described. Mice were kept in a C57/B6-dominant mixed background. At the age of 3 weeks, no significant morphological differences were noted in the articular cartilage between males and females; therefore, sexes were not discriminated.

To suppress Erk signaling in vivo, the Mek inhibitor, U0126 [Cell Signaling Technology; 5 mg/kg body weight diluted in 100 µl phosphate buffered saline (PBS)/mouse], was injected daily intraperitoneally from E11.5 to E18.5.

The animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of Massachusetts General Hospital.

Formalin-fixed mouse hindlimbs were paraffin-processed, cut, de-waxed and rehydrated, and subjected to Hematoxylin-Eosin and Safranin-O staining, and other histological analyses according to standard procedures. Midsagittal sections of the medial tibial plateau were selected for measurements.

Immunohistochemistry was performed using anti-phosphoErk1/2 antibody (4370, Cell Signaling Technology; diluted 1:400), anti-phospho-Akt antibody (4060, Cell Signaling Technology; diluted 1:100), anti-type II collagen alpha 1 antibody (15943-1-AP, Proteintech; diluted 1:50) and anti-Ki67 (Mki67) antibody (bs-2130R, Bioss; diluted 1:250) using an ImmunoCruz rabbit ABC Staining System (sc-2018, Santa Cruz Biotechnology) according to the manufacturer's instructions.

Pregnant mice were injected with 50 μg/g 5-bromo-2'-deoxyuridine (BrdU) 2 h before sacrifice. E13.5 embryonic hind limbs were dissected, fixed in 10% formalin/PBS overnight, paraffin processed and sectioned at 5 μm thickness. BrdU detection was performed using a BrdU Staining Kit (Invitrogen) according to the manufacturer's instructions.

qRT-PCR analysis was performed using a Sybr-green based method with a StepOne Realtime PCR System (Applied Biosystems). Quantification of miRNAs was performed using either a mirVana qRT-PCR miRNA detection kit (Invitrogen) or a qScript microRNA cDNA synthesis kit (Quantabio). Values were internally normalized by the level of Actb or U6.

Mesenchymal progenitor cells were isolated from E11.5 limb buds or postnatal day (P) 0.5 skull. Cells were trypsinized to disperse, and cultured in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% FCS and antibiotics. Cells were serum starved for 2 h before FCS stimulation.

Experiments were performed according to the standard western blot procedure. Anti-phospho-Erk1/2 antibody (4370; diluted 1:2000), anti-phospho-Akt antibody (4060; diluted 1:2000), anti-histone3 antibody (9717; diluted 1:2000) and anti-pan Akt antibody (4691; diluted 1:1000) were purchased from Cell Signaling Technology. Anti-total Erk1/2 antibody was purchased from eBioscience (14-9108-80; diluted 1:1000). Anti-pan Ras antibody (61001; diluted 1:1000) was purchased from BD Biosciences. Anti-Igf1r antibody (bs-4985R) and anti-Actb antibody (845) were purchased from Bioss and Cell Signaling Technology, respectively. Fluorescent and horseradish peroxidase (HRP)-linked secondary antibodies were purchased from Azure Biosystems (goat anti-mouse IgG IR700, and goat anti-rabbit IgG IR800) and Cell Signaling Technology (goat anti-rabbit IgG HRP, 7074), respectively.

Whole-mount X-gal staining using R26R reporter mice (Soriano, 1999) was performed as previously described (Kobayashi et al., 2002).

Nonparametric, unpaired, two-tailed Student's t-test was used for analysis of statistical significance between two groups. P<0.05 was considered statistically significant. The sample size was determined based on the effect size and variability obtained during preliminary experiments.

We thank the Center for Skeletal Research (P30 AR066261) for assistance with histological analysis, and Jennifer Inloes for technical assistance.