Phlpp1 alters the murine chondrocyte phospho-proteome during endochondral bone formation

Abstract

Appendicular skeletal growth and bone mass acquisition are controlled by a variety of growth factors, hormones, and mechanical forces in a dynamic process called endochondral ossification. In long bones, chondrocytes in the growth plate proliferate and undergo hypertrophy to drive bone lengthening and mineralization. Pleckstrin homology (PH) domain and leucine rich repeat phosphatase 1 and 2 (Phlpp1 and Phlpp2) are serine/threonine protein phosphatases that regulate cell proliferation, survival, and maturation via Akt, PKC, Raf1, S6k, and other intracellular signaling cascades. Germline deletion of Phlpp1 suppresses bone lengthening in growth plate chondrocytes. Here, we demonstrate that Phlpp2 does not regulate endochondral ossification, and we define the molecular differences between Phlpp1 and Phlpp2 in chondrocytes. Phlpp2^−/− mice are phenotypically indistinguishable from their wildtype (WT) littermates, with similar bone length, bone mass, and growth plate dynamics. Deletion of Phlpp2 had moderate effects on the chondrocyte transcriptome and proteome compared to WT cells. By contrast, Phlpp1/2^−/− (double knockout) mice resembled Phlpp1^−/− mice phenotypically and molecularly, as the chondrocyte phospho-proteomes of Phlpp1^−/− and Phlpp1/2^−/− chondrocytes had similarities and were significantly different from WT and Phlpp2^−/− chondrocyte phospho-proteomes. Data integration via multiparametric analysis showed that the transcriptome explained less variation in the data as a result of Phlpp1 or Phlpp2 deletion than proteome or phospho-proteome. Alterations in cell proliferation, collagen fibril organization, and Pdpk1 and Pak1/2 signaling pathways were identified in chondrocytes lacking Phlpp1, while cell cycle processes and Akt1 and Aurka signaling pathways were altered in chondrocytes lacking Phlpp2. These data demonstrate that Phlpp1, and to a lesser extent Phlpp2, regulate multiple and complex signaling cascades across the chondrocyte transcriptome, proteome, and phospho-proteome and that multi-omic data integration can reveal novel putative kinase targets that regulate endochondral ossification.

1. Introduction

Endochondral ossification is a complex, dynamic process controlled by numerous signaling factors, hormones, and mechanical forces. Intracellular and extracellular molecules, including Sox9, PTHrP, Ihh, and Fgf18, orchestrate the transition of chondrocytes from a proliferative to post-proliferative state, and then to hypertrophy. In the epiphyseal growth plate, this process drives appendicular growth of long bones [1]. Within the cell, kinases such as PI3K/Akt, Erk/Mapk, and mTorc1 regulate chondrocytic transcriptomes and proteomes to drive cell fate and matrix production [2–4]. Kinase signaling cascades are counteracted and limited by phosphatases to prevent uncontrolled proliferation and facilitate terminal differentiation.

Pleckstrin homology (PH) domain and leucine rich repeat phosphatases 1 and 2 (Phlpp1 and Phlpp2) modulate signaling cascades by dephosphorylating serine and threonine residues in several molecules that control cell proliferation, survival, and matrix production [5–15]. We previously reported that Akt2, S6k, and Erk are hyperphosphorylated in Phlpp1^−/− chondrocytes and that this is associated with enhanced Fgfr3 signaling [7]. We also showed that Phlpp1 regulates expansion of the growth plate [7] in part through a Pth1r-dependent mechanism [16]. In addition, Phlpp1^−/− mice are protected from articular cartilage loss in a model of post-traumatic osteoarthritis [17, 18] and age-related disc degeneration [19].

Phlpp1 and Phlpp2 are highly conserved in humans and mice [15]. Phlpp1 is structurally distinct from Phlpp2 by an extended N-terminus that contains a nuclear localization signal, facilitating the nuclear entry of Phlpp1 [14]. Phlpp1 and Phlpp2 appear to have many common substrates, although a few distinct targets have been described. For example, Phlpp1 specifically dephosphorylates Akt2, while Phlpp2 dephosphorylates Akt1 [5]. Phlpp2 also dephosphorylates N-myc [20]. There is a paucity of research on the role of Phlpp2 in musculoskeletal development and disease. Moreover, it is unclear if Phlpp2 can compensate for Phlpp1 in its absence.

Here, we show that Phlpp2 deletion does not affect endochondral ossification and that Phlpp2^−/− mice have normal long bone lengths. Germline deletion of both Phlpp1 and Phlpp2 (Phlpp1/2^−/−) delays appositional growth to a similar extent as Phlpp1 deletion. Transcriptional and proteomic analyses indicate that more genes and proteins are differentially expressed in Phlpp1^−/− chondrocytes than Phlpp2^−/− chondrocytes. Similarly, phospho-proteomic analysis identified more common substrates in Phlpp1^−/− and Phlpp1/2^−/− chondrocytes compared to WT and Phlpp2^−/− chondrocytes. Multiparametric data integration identified alterations in 3-phosphoinositide-dependent kinase 1 (Pdk1; Pdpk1) and p21-activated kinase (Pak1/2) signaling pathways in Phlpp1^−/− chondrocytes, while Akt1 was altered in Phlpp2^−/− chondrocytes. Taken together, our results demonstrate that Phlpp1, but not Phlpp2, is a crucial regulator of endochondral bone formation, and identify Pdpk1 and Pak1/2-Raf signaling as potential mediators of bone lengthening.

2. Materials and Methods

2.1. Animals

All mice were maintained in an accredited facility with a 12-hour light/dark cycle and supplied with food (PicoLab^® Rodent Diet 20 5058; LabDiet, St. Louis, MO, USA) ad libitum. Animal research was performed according to the NIH and the Institute of Laboratory Animal Resources, National Research Council guidelines. The Mayo Clinic Institutional Animal Care and Use Committee approved all procedures. Phlpp2^−/− mice were re-derived by crossing Phlpp2^fl/fl mice on the C57Bl/6 background (kindly provided by Dr. Tianyan Gao) with mice expressing Cre-recombinase under the control of the Hprt promoter (Hprt-Cre) as described previously [21, 22]. Recombination of the Phlpp2 allele was confirmed by PCR with primers listed in Supp Fig 1. Phlpp1^−/− mice were generated by crossing heterozygous Phlpp1^+/− males and females and genotyped as described [23]. Phlpp1/2^−/− mice were created by crossing Phlpp1^+/−;Phlpp2^+/− males and females. The sample size for each experiment is listed in the respective table or figure legend.

2.2. X-ray imaging

Radiographs of hind limbs were collected using a Faxitron X-ray imaging cabinet (Faxitron Bioptics, Tuscon, AZ, USA). Limb length was measured on radiographs using ImageJ (1.52a) software (NIH, Bethesda, MD, USA; https://imagej.nih.gov/ij/).

2.3. Micro-computed tomography

Micro–computed tomography (μCT) imaging of femora and tibiae was performed with a SkyScan 1276 scanner (Bruker, Kontich, Belgium). Bones were fixed in 10% neutral buffered formalin (NBF) for 24 hours, then stored in 70% ethanol. Scans were performed at 55 kV, 200 μA, 10 μm pixel resolution, 0.2-degree rotation steps for 360 degrees, and four frames average imaging with a 0.25-mm A1 filter. The acquired scans were reconstructed using the Skyscan NRecon software with beam hardening and post-alignment correction. Trabecular analyses of femora were performed using Bruker CtAN software. The datasets were oriented in 3D to vertically align the longitudinal axis of each femur. The region of interest (ROI) was defined as 5% of the length of each bone. Using the distal growth plate as a landmark, the first slice of the ROI was set at 10% of the total length of each bone away from the growth plate. A gray-value threshold of 50 was applied to trabecular segmentations. Quantified outcomes were bone volume/total volume (BV/TV) and bone mineral density (BMD) [24].

2.4. Chondrocyte isolation

Chondrocytes were isolated from the epiphyses of 4- to 6-day-old mice as previously described [7, 16, 25]. Briefly, epiphyses were dissected and digested for one hour in 3 mg/mL collagenase type II, and then overnight in 0.5 mg/mL collagenase type II in serum-free DMEM. The following day, cell suspensions were filtered through a 0.2 μm filter. For experiments validating Phlpp2 deletion in Phlpp2^−/− mice, chondrocytes were pelleted at 1200 × rpm for 5 minutes, washed 3 times with PBS, and immediately processed for qRT-PCR or western blotting as described below. For all other western blotting experiments, chondrocytes were plated at 1 × 10⁵ cells/cm² following collagenase digestion and grown for 48 hours in monolayer, then washed 3 times with ice-cold PBS and processed as described below. For transcriptomic, proteomic, and phospho-proteomic analyses, chondrocytes were cryopreserved following the overnight digestion until a sufficient sample size was achieved for each genotype. Prior to analysis, chondrocytes were thawed and plated in DMEM containing 10% FBS and 1% Penicillin / Streptomycin overnight. The following morning, cells were washed in PBS and lysed in the appropriate buffer for unbiased screening as described in each specific methods section.

2.7. RNA extraction and qRT-PCR

Following cell lysis in TRIzol (Invitrogen), total RNA was extracted using an RNA purification kit (Zymo Research) and 1.5 to 2 μg (depending on the experiment) was reverse transcribed to cDNA with the iScript cDNA synthesis kit (BioRad). RT-PCR was completed using gene-specific primers to Phlpp2 (5’-GGGCTGAGCGCCTCGTTGTT-3’, 5’-ACGCCTGCCGTTGCCATCTC-3’). Transcript levels were normalized to the reference gene Ywhaz (5’-GCCCTAAATGGTCTGTCACC-3’, 5’-GCTTTGGGTGTGACTTAGCC-3’). Abundance and relative fold changes in transcript gene expression were quantified using the 2^−ΔΔCt method relative to WT chondrocytes [26].

2.8. Protein extraction and western blotting

Cell lysates were collected in RIPA lysis buffer (150mM NaCl, 50mM Tris-HCl, 1% NP-40, 0.1% sodium deoxycholate, 0.1% SDS) on ice. Total protein concentrations were obtained using the Bio-Rad DC Assay (Bio-Rad Laboratories). Proteins (20 μg) were resolved by SDS-PAGE and transferred to a polyvinylidene difluoride membrane. Western blotting was performed with antibodies (1:1000) for Phlpp1 (ProteinTech, Rosemont, IL, USA; 22789–1-AP), Phlpp2 (Bethyl Laboratories, Waltham, MA, USA; A300–661A), and Actin (Sigma-Aldrich; A4700) and corresponding secondary antibodies conjugated to horseradish peroxidase (HRP) (Cell Signaling Technology). Antibody binding was detected with the Supersignal West Femto Chemiluminescent Substrate (Pierce Technology, Rockford, IL, USA).

2.9. Histology

Bones were decalcified in 15% EDTA for 5 to 7 days, depending on the age of the mouse, then dehydrated and embedded in paraffin for sectioning. Sections (5 μm-thick) were stained with Safranin O and counterstained with Fast Green or stained with Alcian Blue and counterstained with eosin (all dyes from Sigma-Aldrich). Sections were chosen for analysis based on anatomical landmarks in the bone (i.e., depth of the anterior cruciate ligament insertion site, presence and size of the menisci, continuity of the growth plate). Quantification of number of proliferative cells in the growth plate and the secondary ossification center area were completed in ImageJ. Images were acquired on a Zeiss LSM 900 Confocal Microscope (Zeiss, Inc., Thornwood, NY, USA) and Leica Aperio VERSA slide scanner (Leica Biosystems, Deer Park, IL, USA).

2.10. Transcriptomics

RNA was isolated from the chondrocytes of 5-day-old female WT, Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− mice as described in Section 2.4 and submitted to Genewiz (Azenta Life Sciences, Plainfield, NJ) for bulk next generation RNA-sequencing. RNA Sequencing libraries were prepared using the NEBNext Ultra II RNA Library Prep Kit for Illumina per the manufacturer’s instructions (NEB, Ipswich, MA). Samples were sequenced on an Illumina instrument (4000 or equivalent) using a 2 × 150bp Paired End (PE) configuration. Raw sequence data (.bcl files) generated by the sequencer were converted into fastq files and de-multiplexed using Illumina’s bcl2fastq 2.17 software. One mismatch was allowed for index sequence identification. Trimmed reads were then mapped to the reference genome available on ENSEMBL using the STAR aligner v.2.5.2b. Unique gene hit counts were calculated by using feature Counts from the Subread package v.1.5.2. Only unique reads that fell within exon regions were counted.

After extraction of gene hit counts, differential expression analysis was performed with DESeq2. The Wald test was used to generate p-values and log2 fold changes. Genes with adjusted p-values < 0.05 and absolute log2 fold changes > 1 were considered differentially expressed genes for each comparison. A gene ontology analysis was performed on the statistically significant set of genes. Gene expression data were submitted to the Gene Expression Omnibus database (accession #GSE265773).

2.11. Proteomics and phospho-proteomics

2.11.1. Sample preparation

Chondrocytes were incubated in 300 μL of lysis buffer containing 0.1% SDS, 10 mM triethylammonium bicarbonate (TEAB), and Halt^™ protease and phosphatase inhibitor cocktail (ThermoFisher Scientific, Waltham, MA, USA) and lysed during 10–30sec cycles in a BioRuptor (Diagenode, Denville, NJ, USA). Aliquots (96 μg) were dried and denatured with 5% SDS, reduced with TCEP, alkylated with iodoacetamide, and digested with trypsin on S-trap mini columns using the vendor’s recommended protocol (ProtiFi, Fairport, NY, USA). After digestion, an aliquot was used for total peptide assay, and the remaining digests were vacuum concentrated to dryness for TMT adduction.

2.11.2. Sample labeling

Aliquots (50 μg) of tryptic peptides in 100 mM TEAB (pH 8.5) from each of the 12 samples were adducted with Tandem Mass Tag TMTpro reagents (ThermoFisher) dissolved in 20 μL anhydrous acetonitrile (ACN). Samples were quenched with hydroxylamine, mixed, and excess TMT reagent was removed by desalting on a Sep-Pak C18 cartridge (Waters, Milford, MA, USA). Eluted peptides were dried to remove ACN and divided into 96 fractions by C18 bRPLC (Waters 25 cm × 4.6 mm peptide BEH column) at pH 8.5. The 96 fractions were recombined to 12 fractions in a staggered fashion for greater proteome depth of identification and quantification. Each of the 12 fractions were aliquoted with 10% being used for total proteome analysis, and the remainder reserved for enrichment of phosphopeptides by IMAC (Fe-NTA) chromatography using an Agilent AssayMap Bravo automated liquid handler.

2.11.3. Liquid Chromatography Tandem Mass Spectometry (LC-MS/MS)

LC-MS/MS data were collected from an Orbitrap Eclipse interfaced with a Vanquish Neo liquid chromatograph (Thermo Fisher). Aliquots from 12 fractions of the TMT-labeled samples were loaded via autosampler at 10 μL/min to a reversed phase pre-column (0.33μL EXP2, peptide C18, Optimize Technologies, Oregon City, Oregon, USA) before being switched in line to a separation column (PepSep 40cm × 100um × 1.9um C18 column (Bruker Scientific)). TMT-labeled peptides were separated at 350 nL/min and 50°C using an exponential gradient of 4–40% mobile phase B over 120 min where mobile phase A was 2% ACN in 0.2% formic acid and mobile phase B was 80% ACN with 10% IPA, 10% water, and 0.2% formic acid.

Mass spectrometry data were collected on eluting peptides using a data-dependent-acquisition method. Total cycle time between MS1 survey scans was 3s. MS1 data were collected at 120,000 resolving power (FWHM at m/z 200), AGC target = 100%, max ionization fill time of 50 ms. Tandem mass spectra (MS2) were generated from MS1 precursors with charge (z) range of 2–4, and m/z range 350–1400, precursor isolation width = 0.7, NCE=36, minimum precursor intensity = 5e4, AGC target = 300%, and max ionization fill time = 150 ms (total proteome) and 200 ms (phospho-peptides).

2.11.4. Data processing

Raw mass spectrometry data were processed using the MaxQuant workflow (ver. 2.0.3). Total proteome and phosphopeptide fractions were searched against the reviewed mouse database (UniProt.org, 2022, ver. 2) using fixed modifications of carbamidomethylation of cysteine, and TMTpro adduction on peptide N-terminus and lysine. Oxidation of methionine and acetylation of protein N-terminus were selected as variable modifications, and phosphorylation of S, T, Y was added for the phosphopeptide-enriched fractions. Relative quantification data of peptides for each sample was obtained from abundance of their unique reporter ion signals in each MS2 spectrum. Peptide vs mass spectrum matches were filtered at 0.01 FDR using reversed sequence decoys. Only data from spectra where the precursor ion fractions were greater than 0.8 were used. Relative quantification data from total proteome data was summed at the protein level across fractions while phosphopeptide data was summarized across fractions at the phosphosite level.

A set of R-scripts was used to transform data to the log2 scale and perform a median normalization on total summed reporter ion signal from each sample. Protein and phospho-peptide relative abundance values for individual samples and across sample groups were compared using two-tailed Student’s t-tests.

2.12. Multi-omics

2.12.1. Multi-omics Factor Analysis (MOFA) data integration

To determine similarity between WT, Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− chondrocytes across data modalities (Supp Fig 2), data matrices for RNA-sequencing FPKM expression, total protein quantification, and phospho-site protein quantification were calculated as a cohort Z-score based on cohort mean and cohort standard deviation. Conversion to Z-score transformed the distribution of the data to Gaussian. The multi-omics dimensional reduction method MOFA was then used to integrate RNA, protein, and phospho-protein data into a model of covarying features called Factors using default settings. Chondrocytes were grouped within factors in an unsupervised manner and then annotated.

2.12.2. Hierarchical All-Against-All (HAllA) analysis

To determine significant relationships within the multi-omics dataset, hierarchical all-against-all (HAllA) clustering was performed using all data modalities (Supp Fig 3). To define the effect of Phlpp1 or Phlpp2 deletion, samples were annotated based on Phlpp1 status (Yes: WT, Phlpp2^−/−; No: Phlpp1^−/−, Phlpp1/2^−/−) or Phlpp2 status (Yes: WT, Phlpp1^−/−; No: Phlpp2^−/−, Phlpp1/2^−/−). Features with a significant correlation (FDR < 0.05) were selected. The significant features from all data modalities for each comparison were used for unsupervised clustering of samples to show efficacy. Features were further defined based off directionality, creating a set of features from all three data modalities that were increased or decreased. Kinase enrichment of the protein and phospho-protein feature sets was performed using KEA3 [27].

2.13. Statistics

Statistics were performed in GraphPad Prism (Version 9) using Student’s t-test or one-way ANOVA as appropriate with repeated measures and post-hoc tests for multiple comparisons where necessary. Specific statistical tests are detailed in each experiment. Outliers were identified by Grubbs’ tests and removed where necessary. Data are depicted as box and whisker plots or means ± standard deviation (SD) with individual points shown.

3. Results

3.1. Phlpp2 deletion does not delay endochondral bone formation

We previously showed that the long bones of Phlpp1^−/− mice are shorter than wildtype littermates due to increased numbers of proliferating chondrocytes in the growth plate [7] and that they are less dense [28]. In this study, we sought to determine if Phlpp2 deletion affected cartilage and bone development. At four weeks of age, there were no differences in femur or tibia length, body length, or body weight between male or female WT and Phlpp2^−/− mice (Fig 1A–E, Supp Table 1). In addition, there were no differences in proliferative cell number in the growth plate of Phlpp2^−/− compared to WT mice (Fig 1F, G). Four-week-old Phlpp2^−/− mice also had similar bone microarchitecture to WT mice in the distal femur (Fig 1H), as indicated by similar trabecular bone volume fraction and bone mineral density (Fig 1I, J). There were no differences in body metrics or limb length in male and female Phlpp2^−/− mice at 12 weeks of age (Supp Table 2). Thus, deletion of Phlpp2 does not affect endochondral bone formation.

Figure 1. Phlpp2 deletion does not alter bone size or density.

Femur (A) and tibia (B) lengths from 4-week-old WT and Phlpp2^−/− mice were quantified from x-ray images (C, scale bar is 0.25 cm). Body length (D) and body weight (E) were also measured. Alcian blue staining was performed on the proximal tibial growth plate and the number of cells in the proliferative zone were counted (F, G; scale bar is 50 μm; P=proliferative zone; three regions of interest (ROI) were included for n=3 mice/genotype). MicroCT reconstructions of femurs (H). Trabecular bone volume / total volume (I) and bone mineral density (J) were calculated. Statistics were performed using Student’s t-test. Data are shown in box plots from the 25^th to 75^th percentiles, with whiskers extending to the minimum and maximum values and means shown by horizontal lines.

3.2. Phlpp1/2^−/− mice have delayed endochondral ossification similar to Phlpp1^−/− mice

Western blot experiments indicated that chondrocytes from Phlpp1^−/− mice have equivalent levels of Phlpp2 protein as WT cells, while Phlpp2^−/− chondrocytes express more Phlpp1 protein than WT cells (Fig 2A, B). Therefore, we next assessed how deletion of both Phlpp1 and Phlpp2 affected endochondral bone formation. At two weeks of age, Phlpp1/2^−/− hind limbs were shorter than WT and Phlpp2^−/− limbs but similar in size to Phlpp1^−/− limbs (Fig 2C, D). The secondary ossification center was smaller in Phlpp1^−/− and Phlpp1/2^−/− mice compared to WT and Phlpp2^−/− mice at two weeks of age (Fig 2E, F), suggesting delayed replacement of the fetal cartilage template. Overall stunted development was also reflected by smaller body lengths and body weights in Phlpp1^−/− and Phlpp1/2^−/− mice relative to WT and Phlpp2^−/− mice (Supp Fig 4). Together, these data indicate that deletion of Phlpp1, but not Phlpp2, delays endochondral ossification and that Phlpp2 does not contribute to natural growth plate development.

Figure 2. Phlpp1/2^−/− mice have delayed endochondral ossification.

Chondrocytes were isolated from 5-day-old WT, Phlpp1^−/−, and Phlpp2^−/− mice and plated in monolayer for 48h. Protein lysates were collected, resolved by SDS-PAGE, and western blotted with antibodies for Phlpp1, Phlpp2, and Actin (A). Western blots from three biological replicates were quantified (B). WT, Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− male mice were aged to 2-weeks-old. Tibia and femur lengths were measured following x-ray (C, D). Coronal sections of tibiae were stained with Safranin O and Fast Green (E; scale bar is 100 μm). Secondary ossification center (SOC) areas (outlined in E in dashed lines) were calculated (F). Statistics were performed using Student’s t-test or one-way ANOVA followed by Tukey’s post-hoc tests for multiple comparisons. Data are shown as box plots from the 25^th to 75^th percentiles, with whiskers extending to the minimum and maximum value and means shown by horizontal lines.

3.3. Phlpp1^−/− and Phlpp2^−/− chondrocytes have distinct transcriptomic profiles

To decipher molecular events affected by Phlpp1/2 deletion, transcriptomes of chondrocytes from 5-day-old WT, Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− mice were assessed with bulk RNA sequencing (Fig 3A). Principal component analysis (Fig 3B) and distance calculations (Fig 3C) show that transcriptomes of Phlpp1^−/− and Phlpp1/2^−/− chondrocytes are distinct from WT and Phlpp2^−/− chondrocytes. Of the 20 genes upregulated in Phlpp1^−/− chondrocytes, 10 genes were unique to Phlpp1^−/−. Of the 39 genes downregulated in Phlpp1^−/−, 11 were also suppressed in Phlpp2^−/− and Phlpp1/2^−/− chondrocytes (Fig 3D). Consistent with previous data showing a role for Phlpp1 in regulating gene expression [8, 29] and increasing chondrocyte proliferation in Phlpp1^−/− mice [7, 16, 17], genes upregulated in Phlpp1^−/− chondrocytes had significant enrichment of the GO terms “GO:0045944: Positive regulation of transcription from RNA polymerase II promoter” (P=0.03) and “GO:0008284: Positive regulation of cell proliferation” (P=0.05). Among the top GO terms from upregulated DEGs in Phlpp2^−/− chondrocytes were “GO:0007049: Cell cycle” (P<0.0001) and “GO:0051301: Cell division” (P=0.001) (Fig 3E). Phlpp1/2^−/− chondrocytes had 9 upregulated DEGs (6 of which were Gm “pseudogenes” for which a function has not yet been defined) and 23 downregulated DEGs, including Phlpp1 and Phlpp2. Gene ontology analysis did not reveal any common GO terms between Phlpp1^−/−, Phlpp2^−/−, or Phlpp1/2^−/− versus WT chondrocytes.

Figure 3. Chondrocytes lacking Phlpp1 or Phlpp2 have distinct transcriptomic profiles.

Chondrocytes were isolated from 5-day-old female WT, Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− mice (n=3) for transcriptomic analysis. (A) Volcano plots of differentially expressed genes relative to WT. Dotted lines indicate log2 fold change of 1.0. Differentially regulated genes with log2 fold change ≥ 1 are shown in black. Similarity between samples is shown via principal component analysis (PCA; B) and via distance calculation (C). Upregulated or downregulated genes in Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− chondrocytes relative to WT chondrocytes (log2 fold change ≥ 1, P_adj<0.05) are shown in Venn diagrams (D). GO term analysis was performed on Phlpp1^−/− and Phlpp2^−/− chondrocytes compared to WT chondrocytes (E). Statistics were performed using Student’s t-test or one-way ANOVA followed by Tukey’s post-hoc tests for multiple comparisons. Data are shown as box plots from the 25^th to 75^th percentiles, with whiskers extending to the minimum and maximum value and means shown by horizontal lines.

3.4. Phlpp1 deletion alters the chondrocyte proteome and phospho-proteome

We next sought to characterize the proteomes and phospho-proteomes of chondrocytes lacking Phlpp1 and/or Phlpp2. Chondrocytes were isolated from 5-day-old mice and differentially expressed proteins were identified by mass spectrometry (Fig 4A, B). Phlpp1^−/− mice had the largest number of differentially expressed proteins compared to WT mice, with 22 upregulated proteins and 29 downregulated proteins (Fig 4B, C, Supp Fig 5). Unsupervised clustering demonstrated that Phlpp1^−/− and Phlpp1/2^−/− chondrocytes were similar and clustered separately from Phlpp2^−/− and WT chondrocyte proteomes (Fig 4D). STRING analysis of the 29 downregulated proteins in Phlpp1^−/− chondrocytes revealed a network of interacting proteins that was functionally enriched for GO terms including “GO:0030199: Collagen fibril organization” and “GO:0060351: Cartilage development involved in endochondral bone morphogenesis”. Subsequent STRING analysis of the 22 upregulated proteins in Phlpp1^−/− chondrocytes showed only two interacting proteins (Srebf2 and Apob) that were enriched for “GO:0010884: Positive regulation of lipid storage” (Fig 4E).

Figure 4. Phlpp1 deletion alters the chondrocytic proteome.

Chondrocytes were isolated from 5-day-old female WT, Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− mice (n=3). Differentially expressed proteins were identified by mass spectrometry and heat maps (A) were generated. Venn diagrams (B) and volcano plots (C) depict the number of differentially expressed proteins in Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− chondrocytes relative to the WT phospho-proteome. Dotted lines indicate fold change of 1.5 and P-value of 0.05. Proteins with fold expression change ≥ 1.5 are shown in black. PCA was performed to determine similarity between samples (D). STRING analyses were performed on upregulated and downregulated proteins (minimum required interaction score = 0.700; E) in Phlpp1^−/− chondrocytes. The STRING network of upregulated interacting proteins in Phlpp1^−/− compared to WT chondrocytes are shown in the blue dotted box (n=2 upregulated proteins). The STRING network of downregulated interacting proteins in Phlpp1^−/− compared to WT chondrocytes are unboxed (n=11 downregulated proteins).

Chondrocytes isolated from Phlpp2^−/− mice had only 3 upregulated proteins and 12 downregulated proteins compared to WT chondrocytes (Fig 4B, C, Supp Fig 5). Upregulated proteins in Phlpp2^−/− chondrocytes included Rab27b and Fap, while Ttf1, Aldh1b1, and Chad were downregulated, among others. There were no interacting STRING networks within differentially regulated proteins in Phlpp2^−/− chondrocytes. These modest changes align with the phenotypic data from this study on natural development. It is possible that in Phlpp2 is needed in chondrocytes during responses to physiological challenges or in disease states.

Phosphorylated amino acid residues in WT, Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− chondrocytes were then identified by mass spectrometry (Fig 5A). Chondrocytes from Phlpp1^−/− and Phlpp1/2^−/− mice shared 53 commonly phosphorylated residues (including four on Epha2) and 25 commonly unphosphorylated residues (Fig 5B, Supp Fig 6, Supp Table 3, 4). By contrast, there were only 5 commonly phosphorylated (e.g., S1325 and S1327 on Myo9b) and 6 commonly unphosphorylated residues between Phlpp2^−/− and Phlpp1/2^−/− chondrocytes. Between 81–90% of the posttranslational modifications were on serine residues and 6–13% were on threonine residues, dependent on genotype (Fig 5C). Like transcriptomic and proteomic analyses, Phlpp1^−/− chondrocytes clustered separately from WT and Phlpp2^−/− chondrocytes (Fig 5D, E).

Figure 5. Phlpp1 deletion alters the chondrocyte phospho-proteome.

Chondrocytes were isolated from 5-day-old female WT, Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− mice (n=3). Heat maps (A) of the chondrocyte phospho-proteome were generated by genotype. Venn diagrams (B) depict the number of phosphorylated and unphosphorylated residues in Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− chondrocytes relative to the WT phospho-proteome (fold change ≥ 1.5, P<0.05). Summary tables of phosphorylation sites by amino acid (C) and PCA plots show similarity between samples (D). Volcano plots (E) show differentially phosphorylated proteins relative to WT. Dotted lines indicate fold change of 1.5 and P-value of 0.05. Phosphorylated residues with fold change ≥ 1.5 are shown in black.

3.5. Multiparametric data integration reveals altered kinase signaling cascades in chondrocytes lacking Phlpp1 and Phlpp2

To identify intersections between the transcriptomic, proteomic, and phospho-proteomic features of chondrocytes from WT, Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− mice, an unsupervised multiparametric analysis was performed. Unsupervised multi-omics factor analysis (MOFA) integrated the data modalities and weighted five latent factors. The resulting model explained 35% of the variation in RNA features between samples, 70% of the variation in protein features between samples, and 65% of the variation in phospho-protein features between samples (Supp Fig 7A). In the integrated model, Factor 1 explained the greatest variance in the data and Factor 5 explained the least, with Factors 1 and 2 heavily weighted towards protein and phospho-protein data (Supp Fig 7B). Within Factor 1, Phlpp1^−/− and Phlpp1/2^−/− mice clustered together, separate from WT and Phlpp2^−/−. Clustering in the remaining four factors was less definitive, with chondrocytes of different genotypes more randomly dispersed (Supp Fig 7C). Additional UMAP unsupervised clustering with weighted factor values from multiparametric integration also separated samples by genotype (Supp Fig 7D). Furthermore, cells lacking Phlpp1 (Phlpp1^−/− and Phlpp1/2^−/−) were distinct from WT and Phlpp2^−/− chondrocytes within Factor 1 (Supp Fig 8A, C), while no definitive pattern was seen distinguishing any cells lacking Phlpp2 (Supp Fig 8B, D).

We next sought to confer biological relevance across data modalities for chondrocytes lacking Phlpp1 and chondrocytes lacking Phlpp2. While Factor 1 clearly delineated cells lacking Phlpp1 from WT and Phlpp2^−/− cells, none of the weighted factors sufficiently accounted for chondrocytes lacking Phlpp2. We therefore chose to use a directed hierarchical all-against-all (HAllA) approach to generate a list of differentially expressed features across the transcriptome, proteome, and phospho-proteome that were enriched or repressed in chondrocytes lacking Phlpp1 (Fig 6A) or Phlpp2 (Fig 6B). Notably, the significant features in Factor 1 were significantly enriched / overlapping with the features identified by HAllA (using a lenient loading analysis weight of >0.5 for Factor 1, random data distribution would predict 20% overlap with HAllA. We detected 92% overlap, demonstrating highly similar feature datasets).

Figure 6. Phlpp1/2 deletion modulates Pak1/2, Akt, and other kinase signaling in chondrocytes.

Chondrocytes were isolated from 5-day-old WT, Phlpp1^−/−, Phlpp2^−/−, and Phlpp1/2^−/− mice (n=3) for transcriptomic, proteomic, and phospho-proteomic analysis. Chondrocytes were annotated as lacking Phlpp1 (Phlpp1^−/− and Phlpp1/2^−/−) or lacking Phlpp2 (Phlpp2^−/− and Phlpp1/2^−/−). Hierarchical all-against-all (HAllA) clustering generated heatmaps of statistically significant features between chondrocytes with or without Phlpp1 (Phlpp1^−/− and Phlpp1/2^−/−; A) or chondrocytes with or without Phlpp2 (Phlpp2^−/− and Phlpp1/2^−/−; B). Column annotations include both Phlpp1 and Phlpp2 status. Row annotations indicate data modality. KEA3 analysis was then performed to predict the top ten kinase signaling cascades enriched and repressed in chondrocytes lacking Phlpp1 (C & D, respectively) or Phlpp2 (E & F, respectively) based on MeanRank score across libraries (shown in the color-coded key).

Because phosphorylation explained a significant amount of data variation, kinase pathway enrichment was chosen to define differentially regulated phosphorylation cascades. Signaling cascades enriched in chondrocytes lacking Phlpp1, included Pdpk1 and Pak1/2 (Fig 6C; Supp Fig 9A). Within repressed signaling cascades, cells lacking Phlpp1 converged around mTor, Akt1, and Mapk, among others (Fig 6D; Supp Fig 9B). By contrast, chondrocytes lacking Phlpp2 showed enriched phosphorylation signaling cascades centering on Akt1 (Fig 6E; Supp Fig 9C), while repressed signaling was evident in pathways involving Aurka and Aurkb (Fig 6F; Supp Fig 9D). Various kinases (Akt1, Csnk2a1, Aurka) were enriched in both elevated and repressed phosphorylation signaling pathways. Taken together, these data identify key functional pathways altered in chondrocytes in the absence of Phlpp1 and demonstrate that Phlpp1 contributes to the chondrocyte phospho-proteome, impacting endochondral ossification.

4. Discussion

Phlpp1 and Phlpp2 are serine/threonine protein phosphatases that were originally cloned in a screen aimed at identifying phosphatases encoding PH domains that modulate the Akt/PKB pathway [6]. Phlpp1 displays substrate specificity for Akt2 over Akt1 and regulates many developmental and regenerative processes, including those in the metabolic and cardiovascular systems [30–32], immune system [14, 33], and musculoskeletal system [7, 16, 18, 34–36]. In the appendicular skeleton, Phlpp1 facilitates long bone growth by controlling substrate phosphorylation and proliferation in chondrocytes [7]. Phlpp1 also regulates bone mass acquisition [36]. Phlpp2 is highly homologous to Phlpp1, but preferentially dephosphorylates Akt1 instead of Akt2 [5]. Here, we show that Phlpp2 does not affect in vivo chondrocyte proliferation, appendicular bone growth, or bone mass. Phlpp2^−/− mice are phenotypically indistinguishable from their WT littermates and have similar chondrocyte transcriptomes, proteomes, and phospho-proteomes. By contrast, Phlpp1/2^−/− mice are phenotypically similar to Phlpp1^−/− mice, with delayed ossification, shortened limbs, and common chondrocyte transcriptomic and proteomic features. Multi-omic data integration shows that alterations in the phospho-proteome and Phlpp1 status explain most variation across the chondrocyte transcriptome, proteome, and phospho-proteome. Pdpk1 and Pak1/2 signaling cascades were identified in Phlpp1^−/− chondrocytes, while Akt1 and Aurka were evident in Phlpp2^−/− chondrocytes. Taken together, these phenotypic and bioinformatic data demonstrate that Phlpp1 drives greater differential phosphorylation in chondrocytes to regulate endochondral ossification than Phlpp2.

Phlpp1 and Phlpp2 are both expressed in growth plate chondrocytes, but only Phlpp1 deletion regulates the development of the appendicular skeleton. Alterations in cell proliferation in the growth plate and enhanced phosphorylation of Phlpp1 substrates were described at 5 days old in Phlpp1^−/− mice, with stunted growth apparent by 4 weeks old [7]. Here, Phlpp1^−/− and Phlpp1/2^−/− mice have similarly expanded growth plates and stunted whole-body and bone growth at 2 weeks of age. By contrast, Phlpp2^−/− mice are phenotypically indistinguishable from WT littermates at 2, 4, and 12 weeks of age. As such, Phlpp1 is the dominant isoform driving cartilage development in the mouse. As Phlpp1 and Phlpp2 are expressed across a variety of tissues [15] and germline mice were utilized in these studies, future research should define the chondrocyte-specific role of Phlpp1 in endochondral ossification.

While Phlpp2^−/− mice were phenotypically like their WT littermates, they exhibited alterations in the chondrocyte transcriptome, proteome, and phospho-proteome. Differentially expressed genes in Phlpp2^−/− chondrocytes were associated with cell cycle and cell division. Phlpp2 was shown to promote apoptosis and delay entry into the cell cycle, thereby slowing proliferation in non-small-cell lung cancer cells [5]. Deletion of Phlpp2 also slows proliferation in mouse embryonic fibroblasts lacking Pten/Trp53, modeling in vitro prostate cancer progression [37]. By contrast, Phlpp2 promoted proliferation and senescence via Akt-p53-p21 signaling in hepatocellular carcinoma cells [38]. Thus, the effects of Phlpp2 on proliferation are highly dependent on experimental conditions. While no changes were evident in the number of proliferative cells in the growth plate of Phlpp2^−/− mice, these findings suggest that further research is warranted into the effects of Phlpp2 deletion on in vivo chondrocyte dynamics, perhaps in pathological cartilage conditions such as osteoarthritis or fracture healing.

The role of Phlpp2 in osteoarthritis progression is of relevance, as the proteins Rab27b and Fap were upregulated and Akt1 and Aurka were detected in both enriched and repressed signaling cascades via multi-omic data integration in Phlpp2^−/− chondrocytes. Fibroblast activation protein (Fap) is a serine protease that degrades cartilage and induces murine osteoarthritis [39] while Rab27b is elevated in osteoarthritic menisci [40]. Finally, Aurka is elevated in osteoarthritic tissues and degrades superoxide dismutase 2 (Sod2) to induce mitochondrial dysfunction in chondrocytes [41]. Proteomic analysis of human osteoarthritic chondrocytes also identified the phospho-protein myosin IXb (Myo9b), the top phosphorylated protein in Phlpp2^−/− chondrocytes, as differentially regulated in response to dynamic compression [42]. Taken together, Phlpp2 is a compelling future molecular target in osteoarthritis research.

Phlpp1 deletion increases chondrocyte proliferation [7] and protects cartilage from matrix degradation following joint injury [17, 18, 43]. Significantly enriched GO terms in Phlpp1^−/− chondrocytes included “positive regulation of cell proliferation” and “positive regulation of transcription”. Chondrocytes were isolated for transcriptomics, proteomics, and phospho-proteomics after cryopreservation and subsequent plating in monolayer for 24 hours in growth medium. As such, the cells were rapidly proliferating. Enhanced cell proliferation in Phlpp1^−/− chondrocytes perhaps explains the downregulation of a variety of collagens and extracellular matrix components at the level of the proteome in Phlpp1^−/− chondrocytes. These conditions may also account for why previously upregulated pathways identified in Phlpp1^−/− chondrocytes (e.g., Fgf18, Pth1r) were not enriched in these datasets [7, 16].

The phenotypic effect of Phlpp1 deletion driving short limb length was reflected at the molecular level. Unsupervised multiparametric clustering demonstrated that Phlpp1 status was the principal factor explaining variance between all samples. Additionally, most of the variance amongst the data was accounted for by differences in the proteome and phospho-proteome. Due to its reproducibility and relative ease, transcriptomics is the most widely-performed, unbiased screening technique [44]; however, our data demonstrate that alterations in the RNA explain less of the variation in the data than those at the protein or the phospho-protein level for Phlpp1/2. Additional work is needed to validate the proteomic and transcriptomic results.

To provide biological relevance to the multi-omic analysis, pathway enrichment was performed. HAIIA was used to study the effect of Phlpp1 or Phlpp2 deletion. Samples were annotated based on Phlpp1 status (Yes: WT, Phlpp2^−/−; No: Phlpp1^−/−, Phlpp1/2^−/−) or Phlpp2 status (Yes: WT, Phlpp1^−/−; No: Phlpp2^−/−, Phlpp1/2^−/−) (Supp Fig 3). The Phlpp1/2^−/− samples were included in the “No” group in the HAllA analyses for both Phlpp1 status and Phlpp2 status for several reasons. First, during unsupervised multi-omics integration, no dimension of variation was able to separate Phlpp1/2^−/− samples from Phlpp1^−/− or Phlpp2^−/− samples, demonstrating there were no features synergistically influenced by the deletion of both Phlpp1 and Phlpp2. Second, the control group for each analysis included the single knockout of the other Phlpp gene. For example, to analyze the effects of Phlpp1 deletion, WT and Phlpp2^−/− samples were grouped together as controls. If any features were erroneously identified by including the Phlpp1/2^−/− samples with the Phlpp1^−/− samples, they would be cancelled out because Phlpp2 deletion is accounted for in the control group.

Pathway analysis revealed that, while kinase networks largely converged on Akt1 and Aurka in chondrocytes lacking Phlpp2, chondrocytes lacking Phlpp1 had greater variability in activated kinase pathways, including Pdpk1 and Pak1/2. 3-phosphoinositide-dependent kinase 1 (Pdpk1; Pdk1) is an essential serine/threonine ‘master’ kinase, activating at least 23 downstream kinases (including Akt, PKC, and S6k) to regulate cell growth, survival, and proliferation. Pdpk1 participates in one of the two major steps of Akt activation by phosphorylating Thr308 in the activation loop, with the second step being Torc2-mediated phosphorylation of S473 in the C-terminus [45]. Phlpp1/2 then terminate Akt signaling via dephosphorylation of S473. Conditional knockdown of Pdpk1 using a variety of Cre drivers (Agrp, Ctsk, Mlc1f, Rank) negatively regulates femur length and mass, as well as skeletal muscle mass [46–49]. Altered Pdpk1 signaling within Phlpp1^−/− cells reinforces the critical role Phlpp1 plays at the center of various anabolic signaling pathways, including Akt. In addition to Pdpk1, Pak1/2 and Raf1 kinase activity was enriched in chondrocytes lacking Phlpp1.

Beyond its direct role in terminating Akt signaling, Phlpp1 suppresses receptor tyrosine kinase output to dampen Mek/Erk signaling, which is downstream of Pak1/2 and Raf1 [8, 9]. Raf/Erk signaling is essential for endochondral bone development, depending on the Raf isoform [50, 51]. Pak1 has been implicated in inflammation, osteoarthritic cartilage degradation, and synovitis [52, 53]. Inferring directionality (i.e., increased signaling or decreased signaling) using post-translational modifications is difficult, as phosphorylation at specific residues can induce either gain- or loss-of-function for a particular downstream target. Indeed, several kinases (Akt1, Csnk2a1, Aurka) were enriched in both up- and down-regulated phosphorylation signaling pathways in chondrocytes lacking Phlpp1 and/or Phlpp2. While specific relationship of Phlpp1 to Pak1/2 in chondrocytes should be further explored, these data taken together demonstrate that multi-omics is a useful tool to reveal putative kinase targets and show that Phlpp1 deletion has a varied and rich effect on the chondrocyte phospho-proteome.

5. Conclusion

In summary, these phenotypic and multi-omic data demonstrate that Phlpp1 is the major isozyme regulating chondrocyte activity and endochondral ossification in mice. Phlpp2 deletion has minimal effects across the molecular profile of the chondrocyte and Phlpp2^−/− mice are phenotypically indistinguishable from WT littermates. Data integration using multi-omic analysis of the transcriptome, proteome, and phospho-proteome reveal novel pathways regulated by Phlpp1, including Pdpk1 and Pak1/2. In conclusion, Phlpp1, and to a lesser extent Phlpp2, regulates multiple and complex signaling cascades across the chondrocyte transcriptome, proteome, and phospho-proteome.

Highlights.

1. Phlpp1 deletion significantly alters the phospho-proteome of murine chondrocytes.

2. Phlpp2 deletion does not affect bone length or density during murine development.

3. Multi-omics data integration reveals kinase pathways regulated by Phlpp1 and Phlpp2.