Senotherapeutic drugs for human intervertebral disc degeneration and low back pain

We aimed to compare the senotherapeutic activity of the natural senolytic o-Vanillin a natural senolytics with antioxidant and antiinflammatory properties with a pure synthetic senolytic drug. We chose the commercially available FDA approved drug RG-7112 as a candidate. A concentration of 5 μM RG-7112, shown as a safe and effective dose for human IMR90 lung fibroblasts, was used to evaluate its effect on senescent and non-senescent human IVD cells (Weber, 2010; Laberge et al., 2018). Pellet cultures of IVD cells from degenerate NP and AF regions were exposed to RG-7112 for 4 days, the drug was removed, and the treated pellets were maintained for 21 days in standard media. Then, evaluated for cytotoxicity, senolytic and therapeutic activity. No cytotoxicity was observed following treatment. In contrast, both NP (9.87%; p<0.05) and AF (11.87%; p<0.05) cells showed a significant increase in metabolic activity (Figure 1—figure supplement 1A). Pellet cultures were treated as before and the presence of p16^Ink4a, Ki-67 and caspase-3 were analyzed after 21 days (Figure 1A). RG-7112 significantly decreased the percentage of p16^Ink4a positive NP (11.48%; p<0.001) and AF (20.03%; p<0.0001) cells compared to untreated control cultures (Figure 1B). Similarly, RG-7112 significantly increased the percentage of Ki-67-positive NP (23.07%; p<0.05) and AF (34.92%; p<0.0001) cells (Figure 1C). Caspase-3-positive AF cells increased significantly (17.88%; p<0.01) while a non-significant increase (2.94%; p=0.51) was observed in NP cells following treatment (Figure 1D). Caspase 3 and caspase 3/7 activity was confirmed by fluorescence microscopy and activity assays, respectively (Figure 1 —figure supplement 1B-C and d-f). Confocal immunofluorescence confirmed co-localization of p16^Ink4a and caspase-3, whereas proliferating (Ki-67-positive) cells did not co-localize with p16^Ink4a (Figure 1 —figure supplement 1E (a–c). RG-7112 selectively increased apoptosis (Figure 1 —figure supplement 1E) while it maintained comparable metabolic activity (Figure 1 —figure supplement 1F) in NP cells from both degenerate and non-degenerate IVDs. Finally, we investigated proteoglycan synthesis using the DMMB assay following treatment. A significant increase in proteoglycan release in conditioned media was observed after 14 days of treatment (Figure 1 —figure supplement 1G).

To identify molecular pathways affected by RG-7112 and o-Vanillin in human disc cells, gene expression analysis of 96 selected cellular senescence genes was performed and 91 of the 96 genes were expressed at a detectable level. Pellet cultures were treated with RG-7112 (5 μM) or o-Vanillin (100 μM) for 4 days. The relationship between the differentially up- and downregulated genes are depicted in Figure 2A. Compared with the control group, the number of differentially expressed genes for o-Vanillin were 40 (30 upregulated and 10 downregulated) and eight for RG-7112 (6 upregulated and two downregulated). In upregulated DEGs, three are common to both drugs, mitogen-activated protein kinase 14 (MAPK14), cell division cycle 25 c (CDC25c) and cyclin dependant kinase 2D (CDKN2D or p19^ARF). No down-regulated DEGs were common to the drugs. Although cyclin B1 (CCNB1) is significantly expressed following the treatment with the two compounds, it is upregulated in RG-7112 and downregulated in o-Vanillin (Figure 2A). Of the 91 genes identified, 44 were differentially expressed with a p<0.05 in one or both treatments (Figure 2B). The 47 genes that did not meet the significance criteria of p<0.05 are shown in (Figure 2—figure supplement 1A). Next, we compared gene expression profiles of the o-Vanillin group with the RG-7112 group. Of the 91 evaluated, only eight genes were significantly affected by RG-7112 treatment, four were common to o-Vanillin (MAPK14, CCNB1, CDC25c and CDKN2D (p19^ARF)), and four were exclusive to RG-7112 (MDM2, CDKN1A (p21^ARF), E2F1 and RBL1). In contrast, 40 genes were significantly affected by o-Vanillin treatment. Cell cycle and senescence genes were significantly downregulated including cyclin dependent kinase 2A (CDK2A or p16^Ink4a), cyclin dependent kinase 2C (p18^ARF), Cyclin A2, CCNB1, CDC25c, Vimentin, Mitogen-activated protein kinase 6 (MAPK6) and Checkpoint kinase 1(CHEK1) (Figure 2C). Examples of apoptotic and proliferative genes significantly upregulated by o-Vanillin include B-cell lymphoma 2 (Bcl-2), Bcl-2-like 1, Bcl-2-like 2, interferon regulatory factor 5 (IRF5), IRF7 and receptor tyrosine-protein kinase ERBB2. Interestingly, MAPK14, CDC25c and CDKN2D (p19^ARF) were the only genes significantly upregulated after the treatment with both compounds, while CCNB1 showed opposite regulation patterns. In conjunction, apoptotic pathways are significantly upregulated to kill senescent cells while proliferation related pathways are activated in non-senescent cells. Detailed fold change difference of differentially expressed genes and respective p value are included in Supplementary file 3 (a-d).

To gain a further insight into a potential mechanism of action, the differentially expressed genes were mapped to networks in the Ingenuity Pathways Analysis (IPA) database. The scores take into account the number of focus genes and the size of the network to approximate the relevance of the network to the original list of focus genes. The IPA core analysis features allowed identification and determination of one network connecting the cell cycle, cell death and survival, connective tissue development and function pathways (RG-7112, network 1) of NP cells treated with RG-7112 (Figure 2D). In the o-Vanillin-treated pellets, the highest-scoring network revealed a significant link with cell death and survival, neurological disease and organismal injury and abnormalities (o-Vanillin, network 1) (Figure 2E). Furthermore, connective tissue development and function, cell cycle (o-Vanillin, network 2), cancer, cellular movement (o-Vanillin, network 3), were shown to be influenced in the other two networks (Figure 2F–G). All networks were identified and ranked by the score of the calculated p-value of the IPA assay within the selected set of 91 genes, the scores and molecules used to order these networks are shown in Supplementary file 4 (a-b). These results confirmed the expected mode of action for RG-7112 and provide new insights to the predicted pathways involved in senescence and the responses to treatment with o-Vanillin.

Transcriptomic results following the treatment with the two senolytics predicted the activation of apoptotic (in senescent cells) and proliferative (in non-senescent cells) pathways which suggest a reduction in the senescence burden likeley affecting SASP factor release. To verify an effect on SASP, media of untreated NP cell pellets, RG-7112 and o-Vanillin-treated pellets were used. The media collected were from the same donors used for gene expression analysis in Figure 2. Samples were analyzed using an antibody array that simultaneously screen 80 factors. Out of the 80 factors present, 50 were detected in at least four donors and were included in further analysis (Figure 3 —figure supplement 1A). The 25 most affected SASP factors are divided into four classes: Cytokines (Figure 3A–a), chemokines CC (Figure 3A–b), chemokines CXC (Figure 3A–c), growth and neurotrophic factors (Figure 3A–d). The most reduced factors following RG-7112 treatment compared with control were for cytokines (TNF-α; −78.05 ± 4.69%) (p=0.71), CC-chemokines (CCL26; −66.69 ± 7.09%) (p=0.93), CXC-chemokines (CXCL13; −29.02 ± 17.26%) (p=0.68), growth and neurotrophic factors (HGF; −65.24 ± 5.9%) (p=0.87). The top four factors decreased by o-Vanillin treatment were: IL-7 (−60,37 ± 11.46%) (p=0.5), CCL26 (−40,14 ± 3.41%) (p=0.99), CXCL10 (−18,11 ± 6,15%) (p=0.62) and HGF (−52,42 ± 5,45%) (p=0.96) (Figure 3A (a-d)). To better visualize the overall effect of the two drugs on SASPs release, a scatter plot of average changes was generated demonstrating a similar overall effect of the two senolytics (Figure 3A–e). The trends for all 80 factors are visualized in (Figure 3—figure supplement 1B-C). The semi-quantitative cytokine array results were used to select 19 factor that were quantified with the Luminex assay. The 19 factors were selected based on the following criteria: only factors detected in all conditions and in at least four donors were chosen. These factors were selected to represent all the SASP classes and to include the most variably expressed cytokines, chemokines and growth factors. The 19 factors selected for further investigation were (INF-γ, TNF-α, IL1-α, IL1-b, IL-6, IL-8, CCL5, CCL7, CCL11, CCL24, CCL26, CXCL1, CXCL5, CXCL9, CXCL10, CXCL11, CX3CL1, Angiogenin and VEGF-A). The concentrations were measured and expression heatmap showed an overall decrease in the level of proinflammatory factors from both RG-7112 and o-Vanillin when compared to non-treated cultures. (Figure 3B). The levels of INF-γ, IL-6, IL-8 and CCL24 were significantly decreased following both treatments whereas the decrease in CCL7, CXCL1, CXCL5, CXCL9, CXCL10 and VEGF significance was reached only following o-Vanillin treatment (Figure 3C). However, significance was not reached for TNF-α, IL1-α, IL1-β, CCL5, CCL11, CCL26, CXCL11, CX3CL1 and Angiogenin in treated cultures compared to untreated cultures (Figure 3—figure supplement 1C). Although, o-Vanillin reduced more SASP factors than RG-7112 at a significant level when compared to untreated cultures, there was no significant difference in concentrations between o-Vanillin and RG-7112-treated cultures, validating the similar overall effect the two drugs have on SASP factor release observed by cytokine array in Figure 3A–f. These results demonstrate that both RG-7112 and o-Vanillin decreased the overall inflammatory environment in pellet cultures of degenerating tissue and suggest a broader anti-inflammatory effect of o-Vanillin treatment.

RG-7112 or o-Vanillin was injected into the central region of intact human IVDs to verify that the drugs can reach and kill the target cells in native tissue. Discs were pre-cultured for 4–6 days. The IVDs were treated with a single injection of vehicle, o-Vanillin, or RG-7112, delivered to the centre of the NP and were cultured for another 28 days as outlined in Figure 4A. An adapted MRI protocol (Rosenzweig et al., 2018) was used to assess T1ρ-weighted MRI signal that directly correlates with proteoglycan content (Rosenzweig et al., 2018; Mulligan, 2015). The method used to assess uniform NP regions of interest is shown in (Figure 2—figure supplement 1B-D). The same NP regions of interest were scanned in the same orientation pre- and post-treatment. Increased T1ρ values were found in RG-7112 and o-Vanillin-treated discs while vehicle-treated discs displayed decreased T1ρ values. (Figure 4B (a-d)) Quantification of pre- versus post- treatment intensity values showed a non-significant decrease of 13.2 ± 4.7% (p=0.058) in vehicle-treated control IVDs. In RG-7112 treated discs, there was a significant 6.8 ± 1.5% (p=0.024) increase in the T1ρ value post-treatment. Discs that were treated with o-Vanillin, displayed a significant increase of 11.1 ± 1.2% (p=0.001) in T1ρ values following treatment (Figure 4C). The red dye safranin-O, binds negatively charged molecules, which are primarily represented by proteoglycan in the NP. Histological evaluation post-treatment using Safranin-O/fast green staining showed strong (intense red) staining in treated discs (Figure 4D). Finally, p16^Ink4a and Ki-67 immunohistochemistry were performed to verify senolytic activity of the two compounds (Figure 4E). Indeed, immunohistochemical assessment of p16^Ink4a showed a significant decrease in the number of senescent cells of 17.38 ± 1.6% (p=0.0017) in RG-7112 and 22.65 ± 3.6% (p=0.008) in o-Vanillin-treated IVDs compared with vehicle-treated control discs (Figure 4F). Quantification of Ki-67 staining showed a non-significant increase in both RG-7112 and o-Vanillin 3.05 ± 3.4% (p=0.4) treated IVDs (Figure 4G). The data suggest that both senolytics can reach and kill naturally occurring senescent human IVDs cells situated in their native environment and at the same time promote tissue repair and regeneration.

Culture media from treated and untreated discs (in Figure 4) was analyzed using cytokine arrays and were compared to their respective pre-treatment media. Interestingly, a single injection with RG-7112 strongly decreased secretion of several proinflammatory cytokines and chemokines including IL-7 (−60.63% (p=0.01)), IL-6 (−59.39% (p=0.14)), CXCL1 (−36.72% (p=0.26)), GRO-abg (−32.84% (p=0.21)) and CCL24 (−30.57% (p=0.15)). o-Vanillin also strongly decreased IL-7 (−83.37% (p=0.09)), CXCL1 (−65.45% (p=0.12)), IL-6 (−58.68% (p=0.10)), CXCL6 (−52.47% (p=0.009)), IGFBP-2 (−44.51% (p=0.15)), GRO-abg (−41.83% (p=0.03)) and CCL2 (−39.91% (p=0.05)). Both compounds significantly decreased IL-8 by 18.72% (p=0.004) for RG-7112 and by 11.75% (p=0.04) for o-Vanillin. We also detected a moderate increase of CXCL9, CCL22, NT3 and EGF for o-Vanillin and only in CCL2 and OPN for RG-7112-treated media (Figure 5A). The level of the SASP factors released from vehicle-treated discs showed an increase in IL-6 (530.85% (p=0.15)), CXCL6 (196.89% (p=0.07)), CXCL1 (99.83% (p=0.18)), OPN (92.36% (p=0.01)), CCL2 (53.37% (p=0.18)), IGFBP-2 (39.26% (p=0.13)), IL-7 (23.52% (p=0.53), GRO-abg, (14.25% (p=0.91)), CCL24 (9.64% (p=0.97)), EGF (3.18% (p=0.85)), TNSF-14 (2.09% (p=0.66)), IL-8 (1.61% (p=0.82)) and a decrease in CXCL9 (−35.23% (p=0.26)), CCL22 (−11.95% (p=0.56)) and NT3 (−13.46% (p=0.25))(Figure 5B). The complete set of factors from all groups are presented in Figure 5—figure supplement 1 (A-C).

Luminex quantification of the same 19 selected SASP factors that were analyzed in pellet cultures showed an overall decrease in inflammatory mediators in culture media from both RG-7112 or o-Vanillin-treated discs compared with that of vehicle (Figure 5C). All factors decreased in RG-7112 and o-Vanillin-treated discs (Figure 5D and Figure 5—figure supplement 1D). Six factors in RG-7112 and five in o-Vanillin-treated discs were significantly decreased in the post-treatment media compared with their respective pre-treatment media. For RG-7112, these proteins were TNF-α (mean 12.4 pg/ml in pre, 6.8 pg/ml in post, p=0.03), CCL11 (mean 13.1 pg/ml in pre, 7.8 pg/ml in post, p=0.009), CCL24 (mean 201.7 pg/ml in pre, 101.1 pg/ml in post, p=0.04), CXCL1 (mean 261.8 pg/ml in pre, 148.2 pg/ml in post, p=0.001), CXCL10 (mean 197.1 pg/ml in pre, 158.7 pg/ml in post, p=0.04) and Angiogenin (mean 51.7 pg/ml in pre, 37.6 pg/ml in post, p=0.02). o-Vanillin significantly reduced the levels of INF-γ (mean 1.6 pg/ml in pre, 1.1 pg/ml in post, p=0.04), CCL24 (mean 232.7 pg/ml in pre, 73.7 pg/ml in post, p=0.02), CXCL1 (mean 313.9 pg/ml in pre, 130.2 pg/ml in post, p=0.01), Angiogenin (mean 47.4 pg/ml in pre, 35.5 pg/ml in post, p=0.02) and VEGF-A (mean 24802.1 pg/ml in pre, 20905.5 pg/ml in post, p=0.04) (Figure 5D). Finally, we found a significant increase in the release of four factors in vehicle treated discs: INF-γ (mean 1 pg/ml in pre, 1.45 pg/ml in post, p=0.02), CCL11 (mean 7 pg/ml in pre, 19.2 pg/ml in post, p=0.04), CXCL1 (mean 104.3 pg/ml in pre, 476.9 pg/ml in post, p=0.03) and CXCL9 (mean 60.4 pg/ml in pre, 162.4 pg/ml in post, p=0.01).

Human lumbar IVDs were harvested from organ donors through a collaboration with Transplant Quebec. All procedures are approved by and performed in accordance with the ethical review board at McGill University (IRB#s A04-M53-08B). Familial consent was obtained for each subject. Table 1 provides an overview of donor demographics and Table 2 provides detailed description of the used discs. Lumbar spinal columns were removed from organ donors, they were imaged radiographically and visually, and signs of degeneration were noted (Page et al., 1993; Mort and Roughley, 2007). Discs were then dissected from the spinal column and used for cell and organ cultures. Nucleus pulposus (NP) and annulus fibrosis (AF) cells were isolated separately as described previously (Cherif et al., 2019; Kirkland and Tchkonia, 2017). All cultures were mycoplasma free as verified by Mycoplasma PCR Detection Kit (ZmTech Scientific).

10,000 cells were seeded per well in 96-well tissue culture plates for 12 hr. Then, cells were incubated in the presence or absence of RG-7112 (5 μM) for 6 h. Metabolic activity was evaluated by the Alamar blue assay as previously described (Livak and Schmittgen, 2001). We measured the metabolic activity of treated and untreated NP cells cultured in monolayer or in pellet from degenerate and non-degenerate discs to verify a non cytotoxic window of RG-7112. Results are presented as a percentage of metabolic activity compared to their control. Experiments were performed (3–6) times in triplicate wells for each compound and concentration.

300,000 NP cells/tube were collected by centrifugation at 500 × g for 5 min. Pellets were incubated in 1 mL DMEM (2.25 g/L glucose, 5% FBS, ascorbic acid (5 μM) (Sigma-Aldrich, Oakville, ON, Canada)) at 37°C and 5% CO₂. Culture media was changed every 3 days and collected for analysis. After 4 days, the pellet culture is stabilized and a single dose of the senolytics or vehicle was added to the culture media. Metabolic activity measures on pellet culture media from degenerate NP and AF cells, treated or not with RG-7112, was performed to evaluate a potential cytotoxic effect of RG-7112 by comparing, in the same pellet culture, the effect of the treatment (at day 21) to the control before treatment (at day 4). RNA was collected in TRIzol (Thermo Fisher Scientific) for gene expression experiments (as described in Krock et al., 2014). For immunohistochemistry, pellets were fixed in 4% paraformaldehyde and cryopreserved, 5 μm sections were prepared and stained overnight at 4°C for p16^Ink4a and 1 hr at room temperature for Ki-67 and caspase-3 primary antibodies as described (Cherif et al., 2019). Coverslips were mounted using Aqua Polymount, and bright-field images were visualized. Ten fields, randomly distributed across the well, were analyzed, and the number of positive (brown stained) and total cells were counted to calculate the percentage of senescent, proliferative and apoptotic cells. Senescence, Apoptosis and proliferation assessments were performed on cells pellet cultured for 21 days by comparing, in the same subject, the senolytics-treated pellet to their respective controls.

Isolated cells were expanded to Passage 1 (P1) in monolayer cultures. P1 cells were then seeded at 20,000 or 10,000 cells per well in eight-well chamber slides (Nunc Lab-Tek II Chamber Slide System) and 96- well flat clear bottom black microplates (Corning, NY) respectively. Cells were serum-starved in DMEM with ITS (1X) (Thermo Fisher, Waltham, MA) for 2 hr prior to treatment with 5 μM RG-7112 (Selleck Chemicals, TX), 100 μM o-Vanillin (Sigma-Aldrich, Oakville, ON, Canada) or vehicle (DMSO (0.01%, (Sigma-Aldrich, Oakville, ON, Canada) for 6 hr. Immunocytochemistry was performed as previously described (Cherif et al., 2019). Apoptosis was detected using a commercial kit (ab176749, Abcam, Cambridge, MA) according to the manufacturer’s instructions. Photomicrographs were acquired with a fluorescent Olympus BX51 microscope equipped with an Olympus DP71 digital camera (Olympus, Tokyo, Japan).

Caspase 3/7 activity of treated and untreated NP cells from degenerate and non-degenerate IVDs was measured using the Amplite Fluorimetric Caspase 3/7 Assay Kit (AAT Bioquest, Sunnyvale, CA) according to the manufacturer’s protocol. Cells were incubated with the caspase 3/7 assay solution, which contained caspase substrate (Z-DEVD-R110), at room temperature for 1 hr in the dark. Fluorescence intensity was then measured at 490 nm excitation and 525 nm emission. The results are expressed as a percentage of the mean of the control group (set at 100%). Each experiment was performed in triplicate and carried out three times from each round of cell isolation.

Sulphated glycosaminoglycans (GAGs) were quantified using the DMMB assay in the media of degenerate NP and AF pellets with or without RG-7112 treatment, performed as previously described (Wickham, 2016). Chondroitin sulfate was used to generate the standard curve. Conditioned media samples from days 7, 10, 14 and 21 were evaluated separately in triplicate into clear 96-well plates (Costar, Corning, NY). sGAG release in media was normalized to the sGAG concentration in media at day 0 and then normalized to the untreated group.

Following treatment, RNA was extracted using the TRIzol chloroform extraction method, as previously described (Krock et al., 2014). Briefly, 500 ng of RNA was reverse transcribed using a qScript cDNA Synthesis Kit (Quanta Biosciences, Beverly, MA) with an Applied Biosystems Verti Thermocycler (Thermo Fisher, Waltham, MA). RT-qPCR was performed using an Applied Biosystems StepOnePlus machine with TaqMan Fast Universal PCR Master Mix (2×) and Custom TaqMan Array 96-Well Fast Plates (Thermo Fisher, Waltham, MA). The evaluated genes are senescent and apoptotic genes included in the Human TaqMan Array, for Human Cellular Senescence (Thermofisher scientific, Array ID: RPU62T7, Catalog number: 4413255). The Array Plate was customized to include 12 additional recently identified senescence and anti-apoptotic pathway genes⁸⁴. The 96 included genes are described in Supplementary file 1. Each well of the TaqMan Array Plate was reconstituted using a mix of Fast Master Mix and a cDNA sample (20 ng) to a set final volume (10 μl). Five plates were used for each group (CTRL, RG-7112 and o-Vanillin), and fold-change in gene expression was calculated using the 2^−ΔΔCt method ⁸⁵ after normalizing to the housekeeping gene and vehicle-treated cells.

We conducted a gene expression study of a pre-specified set of apoptotic and senescence-genes of interest (Custom Taqman 96-Well Fast Plates, Thermofisher scientific) in nucleus pulposus cells. This single-gene approach offers the advantage that highly relevant genes can be identified and tested first. The candidate genes examined with this approach allowed us also to identify the genes within the selected group that together carry out the drug response in disc cells.

In order to mine the feature genes from different datasets, a Venn diagram analysis was conducted using Venny version 2.1.0 software (Oliveros, 2007). Differences in the expression levels of DEGs for treated and untreated groups were obtained, and the number of DEGs upregulated and downregulated were calculated. The odds ratio (OR) was calculated according to OR = (DEGs_o-Vanillin * DEGs_RG-7112/DEGs_o-Vanillin * NonDEGs_RG-7112) / (NonDEGs_o-Vanillin * DEGs_RG-7112/nonDEGs_o-Vanillin * NonDEGs_RG-7112).

Heatmaps of gene expression pattern were constructed using unsupervised hierarchical clustering using Euclidean distance metric and complete linkage clustering method of the 43 differentially expressed genes in response to either RG-7112 or o-Vanillin (p<0.05) treatment of degenerate NP cells pellet. Differences in gene expression between o-Vanillin, RG-7112 and control groups were respectively compared via unpaired t-tests using the R package. Genes for which met the p<0.05 cut-off point were selected as DEGs, following which gene expression profiles of DEGs were visualized (heatmaps) via the ‘ggplots’ in R package ⁸⁷ and were represented in the heatmaps as Z-scores, which is: (expression value - mean expression value across samples)/divided by the standard deviation. Colors, ranging from blue to grey then red, for each treatment represents the average fold change of each subject in that group.

Differentially expressed genes were subjected to Ingenuity Pathways Analysis (IPA) (Ingenuity Systems, Redwood City, CA) and used as a starting point for building biological networks. This analysis uses computational algorithms to identify networks consisting of focus genes (genes that were present in our list of 91 genes) and their interactions with other genes (‘non-focused’) in the knowledge base. Scores were calculated for each network according to the fit of the network to the set of focus genes and used to rank networks on the Ingenuity analysis. IPA uses the genes from the highest-scoring network to extract a connectivity pathway that relates candidate genes to each other based on their interactions. The involved function and disease significantly associated with these candidates’ genes were shown. To generate the networks, significant pathways were filtered by p-value (α)<0.05 and activation Z-score > −2 or >2, set as the cut-off values and representing a significant deactivation or activation, respectively.

Colors are based on log2 fold changes on these genes. To rank networks of the IPA, p-scores were calculated from p-values. For example, for n genes in the network and f of them are Focus Genes. The p-value is the probability of finding f or more Focus Genes in a set of n genes randomly selected from the Global Molecular Network calculated using Fisher’s exact test. Since interesting p-values are typically quite low, it is visually easier to concentrate on the exponent and the p-score is defined as p-score = -log10(p-value). Networks with a score ≥2 have at least 99% confidence that indicate a 1/100 chance that the focus genes are in a network because of random chance.

Intact lumbar spines were x-rayed, and discs were selected for the study based on the grading system described by Wilke et al. (Page et al., 1993). Discs with a Wilke grade of 2 were included for this study. Three IVDs from the same spine (n = 4 spines, 12 discs) were isolated and cultured as previously described (Gawri et al., 2011; Krock et al., 2014; Rosenzweig et al., 2016). Disc characteristics are described in Table 3. Isolated discs were scanned by MRI before and after treatment at day 28 as described by Rosenzweig et al., 2018. Images were obtained on a 7T Bruker BioSpec 70/30 USR (Bruker Biospin, Milton, ON, Canada) with the high-performance mini-imaging kit gradient upgrade AVIII electronics (Bruker) and a Bruker-issued T1ρ-RARE pulse sequence, as previously established (Rosenzweig et al., 2018; Mulligan, 2015). Briefly, 3D images were acquired and T1ρ values were quantified using the MIPAV software (NIH Center for Information Technology, Bethesda, MD, USA). T1ρ values of the same region of interest (ROI) of ʹbeforeʹ and ʹafterʹ treatment scans of each disc were normalized to the surrounding culture medium using editing features in MIPAV software. After 4 days of culture and one media change, single injection of the discs with vehicle, RG-7112 (5 μM/g disc) or 100 μM o-Vanillin (100 μM/g disc) in a total volume of 200 μl PBS was performed as previously described (Krock et al., 2014; Rosenzweig et al., 2016). Discs were then cultured in DMEM supplemented with 1x Glutamax, 50 μg/ml gentamicin and 1% FBS for 28 days. Media was changed and collected every 4 days. On day 28 and after MRI imaging, discs were prepared for immunohistochemistry as described below (See Intact IVD Tissue Immunohistochemistry). Conditioned media was collected at each change and frozen as individual samples at −80°C for protein analysis.

Preliminary measures to delimitate the NP region of interest (ROI) were calculated using 11-disc images from random organ donors. A contour was drawn around the disc using a Wacom Intuos Pro tablet and stylus (Wacom, Japan). This polygon was used to measure total disc voxel intensity. The NP area was created by reshaping the original disc contour to 30% of its frontal size and 40% of its sagittal size centered around the middle of NP area, which was calculated to be 10% shift below the center of the disc (Figure 2—figure supplement 1B-D). The T1ρ images were manually cropped around the perimeter of the IVDs. The average of the T1ρ values was calculated within the ROIs for the vehicle and the injected discs slices per image. Heat maps representing signal intensity were created using the MIPAV software.

Age of an individual donor is specified in years. The level of the disc injected with either DMSO, O-Vanillin, and RG-7112 from the lumbar region is indicated. Discs were graded based on Wilke et al., 2006 (Page et al., 1993). Disc height was determined by averaging the dorsal, ventral and midsection disc height. No differences in grade or height were found between groups prior to treatment.

Media from the Degenerate NP Pellet used in gene expression analysis was collected from day 4 to day 21 and pooled separately for each group (vehicle, o-Vanillin and RG-7112) and subject. Then, they were analyzed and compared to evaluate the effect of the two senolytics on SASP release. Disc media collected from day 4 (pre-treatment) and day 28 (post-treatment) were analyzed separately for each disc and each factor. The Human Cytokine Antibody Array C5 (RayBiotech, Inc) was used for semi quantitative detection of 80 proteins according to manufacturer’s instructions and as previously described (Krock et al., 2014). Intensity units were detected by the chemiluminescence using an ImageQuant LAS4000 Image Analyzer (GE Healthcare, Baie d'Urfe, QC, Canada) and analyzed with ImageQuant TL array analysis software (GE Healthcare). The relative quantity of each factor present in each media sample was normalized to the positive and negative controls included on the array. Mean relative concentration of each factor of treated and control groups were then calculated. Data was normalized to secretion of vehicle injected discs from the same donor spine. A list of included cytokines is provided in Supplementary file 2.

Nineteen proteins were selected for analysis by Luminex multiplex assay according to manufacturer's instructions. A limitation in the number of factors we could measure was the incompatibility of some factors to be measured simultaneously as indicated by the supplier. Concentrations (pg/mL) (INF-γ, TNF-α, IL-1α, IL-1β, IL6, IL8, CCL5, CCL7, CCL11, CCL24, CCL26, CXCL1, CXCL5, CXCL9, CXCL10, CXCL11, CX3CL1, VEGF-A and Angiogenin) were measured in 40 μl media. Median fluorescence intensity (MFI) from microspheres was acquired with a BD FACSCanto II and analyzed in FlowCytomix Pro2.2.1 software (eBioscience). Concentration of each analyte was obtained by interpolating fluorescence intensity to a seven-point dilution standard curve supplied by the manufacturer.

For intact IVDs, post-MRI analysis, a 2 mm wide sagittal tissue segment from the center of the IVD was fixed in periodate lysine paraformaldehyde (PLP) fixative overnight at 4°C. Samples were then washed in PBS and decalcified using Shandon TBD1 Decalcifier solution (ThermoFischer Scientific) over 72 hr at 4°C, changing solution each day. Tissue segments were washed in PBS and placed in 70% ethanol prior to paraffin embedding. Sections of 5 µm were cut and mounted on glass slides. All sections were heated on a hot plate at 55°C for 45 min and deparaffinized and rehydrated. Next, sections were stained with safranin-O/fast green (Sigma-Aldrich, Oakville, ON, Canada) and with antibodies against p16^Ink4a and Ki-67 and counter stained using the DAB detection IHC Kit (ab64264, Abcam, Cambridge, MA) following the manufacturer’s instructions. All images were acquired using a Zeiss Axioskop 40 and an AxioCam MR (Zeiss) and processed using AxioVision LE64 software (Zeiss).

The data was analyzed using Graph Prism 8 (Graph Pad, La Jolla, CA). Analysis was performed by two-tailed Student's t test for comparison between two groups and a multiple pairwise comparison (Analysis of Variance (ANOVA) was used to evaluate the variance between multiple groups with Turkey’s post hoc test. A p value < 0.05 was considered statistically significant.