DPP4-Truncated CXCL12 Alters CXCR4/ACKR3 Signaling, Osteogenic Cell Differentiation, Migration, and Senescence

Abstract

Bone marrow skeletal stem cells (SSCs) secrete many cytokines including stromal derived factor-1 or CXCL12, which influences cell proliferation, migration, and differentiation. All CXCL12 splice variants are rapidly truncated on their N-terminus by dipeptidyl peptidase 4 (DPP4). This includes the common variant CXCL12 alpha (1–68) releasing a much less studied metabolite CXCL12(3–68). Here, we found that CXCL12(3–68) significantly inhibited SSC osteogenic differentiation and RAW-264.7 cell osteoclastogenic differentiation and induced a senescent phenotype in SSCs. Importantly, pre-incubation of SSCs with CXCL12(3–68) significantly diminished their ability to migrate toward CXCL12(1–68) in transwell migration assays. Using a high-throughput G-protein-coupled receptor (GPCR) screen (GPCRome) and bioluminescent resonance energy transfer molecular interaction assays, we revealed that CXCL12(3–68) acts via the atypical cytokine receptor 3-mediated β-arrestin recruitment and as a competitive antagonist to CXCR4-mediated signaling. Finally, a reverse phase protein array assay revealed that DPP4-cleaved CXCL12 possesses a different downstream signaling profile from that of intact CXCL12 or controls. The data presented herein provides insights into regulation of CXCL12 signaling. Importantly, it demonstrates that DPP4 proteolysis of CXCL12 generates a metabolite with significantly different and previously overlooked bioactivity that helps explain discrepancies in the literature. This also contributes to an understanding of the molecular mechanisms of osteoporosis and bone fracture repair and could potentially significantly affect the interpretation of experimental outcomes with clinical consequences in other fields where CXCL12 is vital, including cancer biology, immunology, cardiovascular biology, neurobiology, and associated pathologies.

1.

Stromal derived factor-1 (SDF-1) or C-X-C motif chemokine ligand 12 (CXCL12) is a highly conserved pleiotropic chemokine.¹ It is produced in many different tissues both constitutively and in response to injury by multiple cell types.² CXCL12-producing cell types include bone marrow skeletal stem cells (SSCs), often referred to as bone marrow-derived mesenchymal stem cells (BMSCs), neural stem cells, cardiomyocytes, and endothelial cells.²⁻⁴ In the brain, astrocytes and neurons are the primary CXCL12-producing cells, and they increase CXCL12 secretion during tissue repair following injury and during development to regulate migration of neuronal cells and organize the adult neurogenic zone.⁵⁻⁷ CXCL12 is also produced by endothelial and/or stromal cells in many tissues including lungs and bone and is involved in multiple functions such as tissue development, angiogenesis, immune responses, injury responses—including white blood cell activation and homing—stem cell niche site maintenance, bone repair, neuronal regeneration, cardioprotection, and many pathologies including tumor cell migration and metastasis.⁸⁻¹⁶ CXCL12 mainly signals through two receptors: CXCR4, a guanine nucleotide-binding protein (G-protein)-coupled receptor (GPCR) that can also signal through β-arrestin, and the atypical chemokine receptor 3 (ACKR3) or CXCR7, a β-arrestin-associated receptor with a significantly higher affinity for CXCL12 than the canonical CXCR4 receptor.¹⁷⁻¹⁹

Although physiological and pathological functions of CXCL12 are well documented and show a high degree of reproducibility in the literature, some functions—including those in bone tissue—are more difficult to characterize, showing variations and even contradictions among different research groups. An example of those results is the ability of CXCL12 to induce osteogenic differentiation in osteogenic progenitor SSCs. Our group and others have previously reported that CXCL12 can enhance bone morphogenetic protein 2-induced osteogenic differentiation.²⁰⁻²² Other reports emphasize the pro-osteoclastogenic role of CXCL12.^23,24 However, a recent report showed that CXCL12 restrained bone turnover by suppressing both osteogenesis and osteoclastogenesis.²⁵ One aim of our study is to explain such discrepancies in the CXCL12 literature as an initial approach to reconcile seemingly contradicting data with the simple explanation that some of these discrepancies may stem from CXCL12 metabolites exerting altered effects from the intact chemokine. Consequently, we suggest that the current understanding of the normal and pathological actions of CXCL12 needs to be reexamined in light of the bioactivity of its metabolites.

CXCL12 possesses a very short half-life due to its rapid proteolysis by various peptidases, the most important one being dipeptidyl peptidase 4 (DPP4), which initiates degradation of the CXCL12 N-terminus.^26,27 All CXCL12 variants can be cleaved by DPP4, which removes the dipeptide lysine-proline (KP)dipeptide amino acids from their N-termini.^26,28 Importantly, the ability of DPP4 to cleave the N-terminal dipeptide is regulated by the C-terminus. Hence, CXCL12α [CXCL12(1–68)] is rapidly cleaved by DPP4 after its C-terminal lysine is truncated by carboxypeptidase M or N. In contrast, CXCL12β [CXCL12(1–72)] possesses a longer half-life due to its C-terminus being protected by four additional amino acids rendering it less sensitive to carboxypeptidases.^2,29 Other than difference in the half-life, there seems to be no significant difference in their functions or signaling pathways, yet CXCL12α remains more thoroughly studied in the literature despite the higher levels of CXCL12β found in circulation in plasma and local tissue environments like the bone marrow.^2,27,29−33

The physiological roles of DPP4-cleaved CXCL12α and β are understudied in the literature. Other work has pointed out the important changes in various DPP4 substrates once they are truncated and the resulting altered functionality.³⁴ Although this is the first article that focuses on DPP4-cleaved CXCL12 in bone tissue, earlier studies on the brain revealed that CXCL12(5–67), another CXCL12 metabolite, acts as a neurotoxin in mice and humans. It was also found that CXCL12(5–67) signals through CXCR3 instead of the canonical CXCR4 receptor and that—contrary to intact CXCL12α—it induces apoptosis in neural stem cells via increasing caspase 3/7 activation. Finally, higher levels of CXCL12(5–67) were associated with HIV-1 infection.^31,35 This emphasizes the underlying importance of understanding the activity and in vivo levels of cytokines’ metabolites in addition to the well-studied activity of the parent cytokines.³⁴

This work focuses on understanding how DPP4 proteolysis can modulate CXCL12 actions on SSCs. We then go a step further to try to explain the molecular basis of the functional differences between intact and cleaved CXCL12. We believe that this is critical for understanding the development of osteoporosis and potential intervention targets. Importantly, metabolite modulation of CXCR4 signaling has a much broader implication both in the interpretation of prior CXCL12 research and in research moving forward, in not only better understanding the musculoskeletal field but also in cell migration, differentiation, fate, and cell function in most major areas of basic and clinical cardiovascular, neurological, immunological, and cancer biology research. Finally, this work aims to present the potential therapeutic benefits in manipulating intact and truncated cytokine levels in different pathologies, with focus on bone-related pathologies like osteoporosis and fractures.

2. Results

2.1. DPP4-Cleaved CXCL12 Inhibits Osteogenic Differentiation of SSCs and Osteoclastogenic Differentiation of RAW264.7 Cells

We compared the effects of intact and DPP4-cleaved CXCL12 on the differentiation of different bone cell populations. First, we used the standard Alizarin Red staining technique to help quantify osteogenic differentiation and mineralization in human SSCs. We found that in human SSC lines used in this work, both CXCL12(1–68) and (3–68) inhibited osteogenic differentiation at 25 nM as shown by Alizarin Red and alkaline phosphatase assays (Figure 1A–C). In both assays, adding sitagliptin (1 μM), a DPP4 enzyme inhibitor, to CXCL12(1–68) diminished or reversed its inhibitory effect on osteogenesis. We also repeated the Alizarin Red experiment in murine SSCs using the CXCL12β variant, that is, CXCL12(1–72) and (3–72), and found similar results (Figure S1A). To confirm that this effect was in fact due to a change in osteogenesis and not a treatment-induced change in cell proliferation, we used the crystal violet assay to test cell viability and found no significant decrease in cell viability in intact or DPP4-cleaved CXCL12 groups (Figures 1D and S1B).

Figure 1.

CXCL12(3–68) inhibits osteogenic and osteoclastogenic differentiation. (A–D) Optical density of Alizarin Red osteogenic assay staining and corresponding wells (A), higher magnification images showing Alizarin Red-stained bone nodules (B), alkaline phosphatase (ALP) activity (C), and crystal violet cell viability assay (D) of human SSCs treated with the control, 25 nM CXCL12(1–68), 25 nM CXCL12(3–68), CXCL12(1–68) + CXCL12(3–68), 1 μM sitagliptin, and CXCL12(1–68) + sitagliptin. The ANOVA test was used to compare between different groups. *P < 0.05, **P < 0.01, and ***p < 0.001.

Next, we wanted to test how DPP4-cleaved CXCL12 affects the osteoclastogenic differentiation of RAW264.7 cells. Using a tartrate-resistant acid phosphatase (TRAP) enzyme assay, we found that both CXCL12(1–68) and (3–68) produce a pattern similar to that of osteogenic differentiation in murine SSCs, with both CXCL12(1–68) and (3–68) leading to a significant almost 40% inhibition of osteoclast formation, importantly with sitagliptin plus CXCL12(1–68) returning osteoclast differentiation to normal control levels. This suggests that DPP4-cleaved CXCL12(3–68) was the isoform responsible for inhibiting osteoclastogenesis (Figure 2A,B). To confirm that the changes in the TRAP assay were due to corresponding changes in osteoclast differentiation and not the effect of different treatments on cell proliferation, we utilized Alamar Blue cell viability staining and found no significant changes in cell viability across all treatment groups (Figure 2C). We also tested CXCL12β variants (1–72) and (3–72) and found that they have a similar effect on osteoclast differentiation and cell viability (Figure S1C–E).

Figure 2.

CXCL12(3–68) inhibits osteogenic and osteoclastogenic differentiation. (A–C) Images of TRAP-stained wells for the osteoclastogenesis assay (A), with quantification of the total number of stained cells per well under each condition (B), and Alamar Blue cell viability assay (C) of RAW 264.7 cells incubated with RANKL-containing osteoclastogenic medium and corresponding treatments for 5 days. The ANOVA test was used to compare between different groups. *P < 0.05, **P < 0.01, and ***p < 0.001.

The results from both osteogenic and osteoclastogenic experiments reveal that some reported effects of CXCL12(1–68) on cell differentiation may in fact be due to CXCL12(3–68). As CXCL12(1–68) possesses a very short half-life, it could be cleaved by the DPP4 enzyme on the cell membrane within minutes of its addition, producing CXCL12(3–68) and a phenotype similar to that produced in cells directly treated with CXCL12(3–68). This is evident with the addition of the DPP4 enzyme-inhibitor sitagliptin along with CXCL12(1–68), where we consistently saw attenuation or even reversal of the phenotype produced by CXCL12(3–68) or CXCL12(1–68). This suggests that DPP4 cleavage of CXCL12 may act to sharply terminate CXCL12 signaling by actively inhibiting the signaling process and inducing contrary or different effects.

2.2. DPP4-Cleaved CXCL12 Induces a Senescent Phenotype in SSCs

Since we observed a robust effect of DPP4-cleaved CXCL12 on osteogenic and osteoclastogenic differentiation, we wanted to test its effects on senescence in SSCs. Using the SA-β-Gal assay, we found that CXCL12(3–68) at 25 nM appears to induce senescence in human SSCs (Figure 3A,B). We also found a similar effect in murine SSCs isolated from 24 month old mice, with CXCL12(1–72) and (3–72) significantly inducing senescence and the addition of sitagliptin to CXCL12(1–72) inhibiting this effect (Figure S2A,B). We also tested the ability of CXCL12(3–68) to modulate the migration of SSCs toward chemokines by using transwell migration and scratch wound healing assays. Pre-incubation of human SSCs with 25 nM CXCL12(3–68) for 18 h at 37°c prior to transwell migration assays significantly inhibited the ability of cells to migrate toward medium containing 62.5 or 25 nM CXCL12(1–68) (Figures 3C and S2C). We then tested pre-incubation of murine SSCs for 18 h at 37°c with 25 nM DPP4-cleaved CXCL12β, that is, CXCL12(3–72), or 1 μM AMD3100, a CXCR4 antagonist, and found a similar effect (Figure S2D). To confirm previous reports that DPP4 cleavage diminishes the chemotactic effect of CXCL12, we measured the ability of SSCs to migrate toward CXCL12(1–72) and (3–72) at different concentrations and found that indeed, cell migration toward CXCL12(3–72) was significantly lower than that toward CXCL12(1–72) (Figure S2E).

Figure 3.

CXCL12(3–68) induces senescence and inhibits migration in SSCs. (A,B) Images (A) and quantification (B) of human SSCs treated for 48 h with the control, 25 nM CXCL12(1–68), 25 nM CXCL12(3–68), CXCL12(1–68) + CXCL12(3–68), 1 μM sitagliptin, and CXCL12(1–68) + sitagliptin and stained for the senescence-associated β-galactosidase marker. (C) Transwell migration assay of human SSCs migrating toward wells containing 62.5 nM CXCL12(1–68) after being incubated at 37 °C overnight with the corresponding treatments for 18 h. (D) Scratch wound healing assay in human SSCs 24 h after adding the corresponding treatment. The ANOVA test was used to compare between different groups. *P < 0.05 and **P < 0.01.

In a scratch wound healing assay, we aimed to compare the ability of CXCL12(1–68) and CXCL12(3–68) to affect the migration of murine SSCs to heal a scratch wound in the cell monolayer. 24 h after adding treatments, neither CXCL12(1–68) nor CXCL12(3–68) at 20 nM affected cell migration, but CXCL12(1–68) combined with 1 μM sitagliptin led to a significantly higher rate of migration and faster wound healing (Figure 3D).

2.3. CXCL12(1–68) and (3–68) Primarily Signal through CXCR4 and ACKR3

To determine if CXCL12(3–68) signals mainly via the two known cognate receptors for CXCL12(1–68), CXCR4 and ACKR3, or potentially through other unidentified receptors, we submitted DPP4-cleaved and intact CXCL12 to the NIMH Psychoactive Drug Screening Program, where CXCL12(1–68) and (3–68) were screened against 320 different GPCRs using a Parallel Receptor-ome Expression and Screening via Transcriptional Output–transcriptional activation following the arrestin translocation (PRESTO-Tango) assay.^36,37 CXCL12(1–68) and (3–68) were screened against each of the 320 receptors in four replicates. A luminescence value representing β-arrestin recruitment to the receptor and an intra-assay coefficients of variation (CV) value representing variability across the four replicates were recorded as the location along the X axis. We found that for CXCL12(1–68), as expected, ACKR3 and CXCR4 were the highest scoring receptors showing high luminesce and low CV. For CXCL12(3–68), both ACKR3 and CXCR4 were among the highest scoring receptors with relatively lower CVs; additionally, other receptors including ADCYAP1R1, GPR61, and GPR150 scored higher on the luminescence scale but with higher CVs (Figure 4A,B). The outcomes for 3–68 were much lower than those for 1–68.

Figure 4.

Initial hits from the PRESTO-Tango GPCRome screen for CXCL12(1–68) and (3–68) appear to be limited to CXCR4 and ACKR3 (see Methods 4.11). (A,B) Initial receptor hits for CXCL12(1–68) (A) and CXCL12(3–68) (B) in the GPCRome screen with relative luminescence units (RLUs) on the Y-axis representing strength of the signal and the intra-assay CV on the X-axis representing reproducibility of results. The further we move to the left on the X-axis, the less the variability between replicates and the stronger the reliability of the Y axis β-arrestin recruitment. Data represent mean from four technical replicates.

Next, we requested follow-up assays for an additional five receptors including CCR6, GPR61, GPR146, GPR150, and GPR151. Other receptors were excluded for a low luminescence readout, a high CV value, or as commonly detected artifactual hits. For all the receptors selected in the follow-up assay, CXCL12(1–68) and (3–68) were used in 16 different concentrations with 4 technical replicates for each concentration. None of the receptors showed a significant concentration-dependent increase in β-arrestin recruitment as measured by the PRESTO-Tango assay (Figure S3). Since CXCR4 and ACKR3 were already known cognate receptors for CXCL12, we used further functional assays in the next sections to quantify activity of CXCL12(1–68) and (3–68) for these receptors.

2.4. CXCL12(3–68) Acts as a Competitive Antagonist for β-Arrestin Recruitment via CXCR4 and ACKR3

We further confirmed ACKR3 β-arrestin recruitment using assays where HTLA cells (an HEK293-derived cell line containing stable integrations of a tTA-dependent luciferase reporter and a β-arrestin2-TEV fusion gene) were co-transfected with ACKR3 (Figure 5A). We found that CXCL12(3–68), while not as efficacious as CXCL12(1–68), significantly recruited β-arrestin to ACKR3 in a concentration-dependent manner (EC50 = 338 nM, Emax = 78% of CXCL12). We were somewhat surprised by CXCR4 showing on the NIMH Psychoactive Drug Screening Program screen for CXCL12(3–68), but we could not confirm it using PRESTO-Tango. As an alternate approach, we used a bioluminescent resonance energy transfer (BRET) assay system to measure the close proximity and therefore direct physical interaction between two fluorescently labeled components of the β-arrestin recruitment and signaling system. In this case, we used a Venus fluorescent molecule fused to a β-arrestin sequence to detect β-arrestin recruitment to a reporter gene Renilla luciferase (Rluc8) fused to the CXCR4 receptor (CXCR4-Rluc8). We found that CXCL12(3–68) did not induce significant β-arrestin recruitment to CXCR4 in this assay (Figure 5B). Instead, CXCL12(3–68) was able to concentration-dependently reverse CXCL12(1–68)-induced β-arrestin recruitment similar to the CXCR4 antagonist, mavorixafor, indicating that CXCL12(3–68) acts as a competitive antagonist at this receptor (Figure 5B). We also found no significant difference between CXCL12(1–68) and (1–72) in recruiting β-arrestin via the CXCR4 receptor at concentrations of 2.5 × 10^–8 and −5 x 10^–8 M (Figure S4A).

Figure 5.

CXCL12(3–68) acts as a partial antagonist for CXCR4 β-arrestin recruitment and an agonist for ACKR3 β-arrestin recruitment. (A) Concentration–response curves showing ACKR3-mediated β-arrestin recruitment for CXCL12(1–68) vs CXCL12(3–68) from the Tango assay. (B) Concentration–response curves showing CXCR4-mediated β-arrestin recruitment for CXCL12(1–68), CXCL12(3–68), mavorixafor (CXCR4 inhibitor), 10 nM CXCL12(1–68) + CXCL12 (3–68), and 10 nM CXCL12(1–68) + mavorixafor from BRET assay. This shows that when CXCL12(3–68) is in the presence of CXCL12(1–68), it inhibits CXCR4 β-arrestin recruitment similar to the CXCR4 inhibitor mavorixafor. (C) Concentration–response curve showing CXCR4-mediated Gi3 dissociation for CXCL12(1–68), CXCL12(3–68), mavorixafor, 10 nM CXCL12(1–68) + CXCL12 (3–68), and 10 nM CXCL12(1–68) + mavorixafor from BRET assay. This shows that CXCL12(3–68) acts similar to the CXCR4 inhibitor in terms of CXCR4 Gi3 dissociation. For B and C, the cells treated with CXCL12(1–68) + mavorixafor and CXCL12(1–68) + CXCL12(3–68) were pretreated with CXCL12(1–68) first. n = 4 per group.

2.5. CXCL12(3–68) Does Not Activate CXCR4 G-Protein-Mediated Signaling

Next, we aimed to test the effect of CXCL12(3–68) on CXCR4 G-protein-mediated signaling. We set up a BRET system with HEK293 cells co-transfected with different RLuc-tagged Gαi/o subtypes, Gβ, and Gγ2-green fluorescent protein (GFP) (Figure S5A).³⁸ This allowed us to also test CXCL12(3–68) activity through different Gα subunits. We found that while CXCL12(1–68) induced an expected concentration-dependent decrease in the BRET signal with Gαi₁, Gαi₂, Gαi₃, GoA, and GoB, CXCL12(3–68) failed to induce any G-protein activity as reported by this BRET system (Figure S5B–F). A control experiment with no CXCR4 overexpression was carried out to confirm that the results were CXCR4-specific (Figure S5G). Importantly, CXCL12(3–68) mimicked the CXCR4 competitive antagonist mavorixafor in inhibiting 10 nM CXCL12(1–68) Gi3 dissociation downstream to CXCR4 (Figure 5C), which was similar in the CXCR4 β-arrestin recruitment assay.

For further validation of second messenger production, we also tested CXCL12(1–68) and (3–68) in a forskolin-treated cyclic adenosine monophosphate (cAMP) inhibition GloSensor assay. CXCL12(1–68) reduced cAMP levels in a concentration-dependent manner, whereas CXCL12(3–68) did not affect cAMP levels, or it potentially non-significantly increased them, thus showing no activity consistent with the lack of agonist action at this receptor, as measured in the G protein BRET assays (Supp. Figure 4B).

2.6. Reverse Phase Protein Array Reveals Differential Modulation of Canonical Signaling Pathways between CXCL12(1–72) and (3–72)

To understand how DPP4-cleaved CXCL12 differentially modulates downstream signaling pathways from intact CXCL12, we collaborated with the MD Anderson RPPA core facility to run a reverse phase protein array (RPPA) assay where we treated murine SSCs isolated from four different ages of mice (3, 6, 18, and 24 month old) with either the control, CXCL12(1–72), CXCL12(3–72), sitagliptin, or sitagliptin plus CXCL12(1–72) in the previously mentioned concentrations. CXCL12(1–72) and CXCL12(3–72), rather than 1–68 and 3–68, were chosen for this assay since CXCL12 β has a longer half-life, and the intact form might show greater activity in the RPPA assay. Additionally, the CXCL12 β isoform appears to have a more abundant expression endogenously.^2,27,29−33 Upon combining the results from the four cell lines, CXCL12(3–72) showed a significantly different profile of proteins and phosphoprotein expression levels compared to either the control or sitagliptin plus CXCL12(1–72) groups, including changes in levels of mitogen-activated protein kinase (MAPK) (MAPK)11/12, H3K9me2, and Src (Figure 6A,B).

Figure 6.

CXCL12(3–72) induces a significantly different signaling pattern from that of CXCL12(1–72) + sitagliptin and control groups. (A,B) Hierarchical clustering of expression levels of proteins and phosphoproteins from RPPA data showing data for control group vs CXCL12(3–72) group (A) and CXCL12(3–72) group vs CXCL12(1–72) + sitagliptin group, used to prevent CXCL12(3–72) generation (B). (C,D) IPA showing the changes in different canonical signaling pathways for the CXCL12(3–72) group compared to those for the control group (C) and the CXCL12(3–72) group compared to those for the CXCL12(1–72) + sitagliptin group (D). Data represent four different murine cell lines each with four different technical replicates.

To look beyond individual protein and phosphoprotein expression levels, we used Ingenuity Pathway Analysis (IPA) software to analyze changes in canonical signaling pathways using pairwise comparisons between groups. IPA’s canonical signaling pathway analysis utilizes right-tailed Fisher’s Exact Test to calculate the statistical significance of the overlap between proteins and phosphoproteins measured in RPPA and proteins in various canonical signaling pathways. We found that compared to the control group, CXCL12(3–72) significantly inhibited many canonical signaling pathways including p70S6K signaling, PI3K/AKT signaling, and extracellular signal-regulated kinase (ERK) (ERK)/MAPK signaling pathways with Z-scores of −1, −1, and −1.342, respectively. When compared to sitagliptin plus CXCL12(1–72), CXCL12(3–72) showed significant downregulation in regulation of elF4 and p70S6K signaling, p70S6K signaling, ERK/MAPK signaling, and mTOR signaling pathways with Z-scores of −1, −1, −1.134, and −1.265, respectively. There were also many other pathways that showed significant upregulation or downregulation as can be seen in Figure 6C,D and Tables S2–S8. Heatmaps of significantly modulated proteins and phosphoproteins in other pairwise comparisons can be found in Figure S6.

3. Discussion

CXCL12 is one of the earliest evolutionary conserved chemokines/cytokines.¹ It is constitutively produced by numerous cell types and tissues, and its expression is induced by most tissues in response to injury.²⁹ All CXCL12 variants, including CXCL12α and CXCL12β, are cleaved by DPP4 removing their N-terminus KP dipeptide amino acids.^26,28 CXCL12 is not only highly studied and vital for a number of diverse functions in most tissues but is also a critical player in many pathologies, including cancer. As such, understanding the distinct roles of its metabolic isoforms is novel and has broad and important translational implications.

Although the biological functions and canonical signaling of CXCL12α [CXCL12(1–68)] have been thoroughly studied, other splice variants and the proteolytically produced metabolites have garnered much less research interest. In this article, we investigated one of the most rapidly formed and stable proteolytic cleavage products of CXCL12: DPP4-cleaved CXCL12α [CXCL12(3–68)]. Most of the splice variants are less abundant and more restricted in their tissue expression; however, CXCL12β appears to have a similar expression pattern and may even be functionally expressed as a protein at higher levels (2). Because of this, throughout the paper, we also aimed to test if the functional and signaling outcomes of DPP4-cleaved CXCL12β have similar effects.

Few previous articles have discussed functional outcomes in response to DPP4 proteolysis of CXCL12. For example, Shioda et al. previously reported that DPP4 truncation of CXCL12α and β abolished their chemotactic and anti-HIV-1 activities.²⁶ A more recent article reported that compared to CXCL12(1–68), CXCL12(3–68) did not induce endothelial cell migration, an observation that they suggested represents a novel regulatory mechanism of migration.³⁹⁻⁴² In this study, we found that incubation of SSCs with CXCL12(3–68) does not promote migration and actually further diminishes or inhibits migratory capacity in a transwell and scratch wound healing assays. Since CXCL12 is critical for immune and stem cell migration, this observation could have important implications in tissue repair, wound healing, and inflammatory responses in different tissues and have an unrecognized regulatory role in CXCL12-influenced cell migration during development.

Another important albeit controversial role of CXCL12 in the literature is its effect on the differentiation of different bone cell populations. Work by others and ourselves shows that CXCL12 potentiates the osteogenic effects of bone morphogenetic protein 2 and in some instances induces osteogenesis on its own.²⁰⁻²² In terms of osteoclastogenesis, Shima et al. recently reported that CXCL12 enhances lipopolysaccharide-induced osteoclast formation and increases in vivo bone resorption.²³ Other work by Ponte et al. found that Cxcl12 gene deletion in murine SSCs led to a higher bone turnover and attenuated estrogen-mediated cortical bone loss.²⁵ The contradiction of these reports may reflect that each of the functional outcomes was assumed to be mediated by intact CXCL12 isoforms, without consideration of the rapid in vivo and in vitro metabolism of intact CXCL12 by DPP4. In fact, one article reported that blocking DPP4 activity led to an inhibition of osteoclast formation in humans, which could suggest that the pro-osteoclastogenic role is due to DPP4-cleaved CXCL12 instead of intact CXCL12.⁴³ In this work, we managed to approach an understanding of this problem which may be able to reconcile previous seemingly contradicting reports. We found that in human and murine SSCs, both CXCL12(1–68) and (3–68) significantly inhibited osteogenic differentiation unless DPP4 activity was blocked by sitagliptin. We also found a similar response in osteoclastogenic assays of RAW-264.7 cells, where both CXCL12(1–68) and (3–68) led to significant inhibition of osteoclast formation. Blocking DPP4 activity by adding sitagliptin to CXCL12(1–68) attenuated this decrease in osteoclastogenesis, indicating that the inhibitory effect on osteoclast formation was probably mediated by CXCL12(3–68) rather than intact CXCL12, helping to clarify different reports on its role in osteoclast activity in the literature.^23−25,43 It is important to note that further in vivo and in vitro studies that primarily focus on osteogenic and osteoclastogenic activities of DPP4-cleaved CXCL12 are currently underway to confirm this pattern, to understand regulation of DPP4/CD26 expression on osteogenic and osteoclast cells in vitro and in vivo, and to rule out the involvement of any further metabolites.

Based on our osteogenesis and osteoclastogenesis data, we conclude that DPP4-cleaved CXCL12 may, or may not, lead directly to lower bone mass by interfering with osteogenesis; however, since it appears to target both osteogenesis and osteoclastogenesis, it may lead to reduced overall bone turnover. A shift in overall bone turnover or differential effects on osteogenesis and osteoclastogenesis could be significant in affecting responses to normal bone maintenance and bone injury and aging. We have previously shown in humans with age that there is an increased level of circulating CXCL12 that is independently linked to decreased bone density.³ At the time, we did not determine the ratio of DPP4-cleaved to uncleaved CXCL12 isoforms. However, given the rapid cleavage of CXCL12 by DPP4, it is likely that the increased plasma levels of CXCL12 with age also represent an increase in the DPP4-cleaved isoforms. Furthermore, mechanistically, it may drive SSC dysfunction at least in part via induction of senescence. We hypothesize that this may be a general mechanism for DPP4-cleaved CXCL12 to inhibit the differentiation of different types of progenitor cells that rely on CXCL12 signaling. Indeed, we found that in vitro CXCL12(3–68) leads to a significant increase in the number of SA-β-gal-positive SSCs.

We also sought to understand mechanistically the signaling pathways that may mediate the observed biological actions of CXCL12(3–68). We previously suggested that one possible explanation for the DPP4-cleaved CXCL12’s different profile of biological functions was that it signaled through receptors other than the cognate receptors CXCR4 and ACKR3.³⁴ To explore this idea further, we chose to profile CXCL12 (3–68) in the PRESTO-Tango assay performed by the NIMH Psychoactive Drug Screening Program to screen compounds for possible interactions at different GPCRs. As shown in the Results section, the initial screen revealed some potential third receptors for CXCL12(3–68), but these hits could not be validated in follow-up screens. This does not eliminate the possibility of CXCL12(1–68) and (3–68) signaling through other receptors. For example, the atypical chemokine receptor 1 (ACKR1) was not included in the PRESTO-Tango screen, and this receptor was recently found to be expressed at high levels in the nucleated erythroid cells that make up the red pulp of the bone marrow.⁴⁴ Although it was previously reported that CXCL12 is not a ligand for ACKR1,⁴⁵ a new study shows that ACKR1 binds preferentially to the dimeric form of CXCL12.⁴⁶ Understanding the extent to which CXCR4, ACKR3, and ACKR1 mediate the biological effects exerted by CXCL12(3–68) and its dimeric isoform will be the focus of future studies.

Other possible explanations for CXCL12(3–68) actions include that it either acts as an antagonist or an inverse agonist for the CXCR4 receptor, decreasing CXCR4 mRNA or protein expression or inducing CXCR4 internalization by means of β-arrestin recruitment. Our next step was to test how CXCL12(3–68) affects β-arrestin recruitment via CXCR4 and ACKR3 and G-protein signaling via CXCR4. Earlier work by both Janssens et al. and Cheng et al. demonstrated that CXCL12(1–68) can efficiently recruit β-arrestin via both receptors, while CXCL12(3–68) was reported to only recruit β-arrestin via ACKR3, although with a 5- to 10-fold reduction in signaling efficiency.^39,47 Work by Ziarek et al. demonstrated that CXCL12(3–68) shows a dramatic drop in CXCR4 binding affinity and loses G-protein agonist activity.⁴⁸ Here, we showed that similar to previous reports, CXCL12(3–68) retained ACKR3-mediated β-arrestin recruitment. Importantly, ACKR3 is thought to also target CXCR4 expression and availability.^49,50 Furthermore, we showed that CXCL12(3–68) could not recruit β-arrestin through CXCR4 but could interfere with and inhibit CXCL12(1–68) CXCR4-mediated β-arrestin recruitment, thus acting similar to the CXCR4 competitive antagonist mavorixafor.

In terms of G-protein activation, previous work by Janssens et al. demonstrated that contrary to CXCL12(1–68), CXCL12(3–68) does not induce an increase in IP₃ response, ERK phosphorylation, or Akt phosphorylation.³⁹ In terms of Ca²⁺ signaling, Ziarek et al. reported that CXCL12(3–68) did not induce Ca²⁺ mobilization at concentrations up to 1 μM, and it even weakly inhibited CXCL12(1–68)-mediated Ca²⁺ mobilization with an IC₅₀ of 4.5 μM.⁴⁸ Interestingly, earlier work by Proost et al. reported that although CXCL12(3–68) did not induce a Ca²⁺ response at concentrations up to 30 nM, 10 nM was able to induce a 50% reduction in the Ca²⁺ assay response toward a second stimulus with CXCL12(1–68), indicating that CXCL12(3–68) may still be able to bind and desensitize CXCR4 for CXCL12(1–68).²⁷ In our study, we found that CXCL12(3–68) was unable to induce a G-protein-mediated signal via Gαi₁, Gαi₂, Gαi₃, Gαo_A, or Gαo_B. However, in cAMP assays, CXCL12(3–68) induced a slight increase in cAMP compared to the concentration-dependent reduction in cAMP caused by CXCL12(1–68) (Figure 7). This suggests that it will be important to further explore whether the DPP4-cleaved metabolite may have an antagonistic effect on G-protein-mediated signaling.

Figure 7.

Suggested signaling pathway of CXCL12(3–68). CXCL12(3–68) signals mainly through ACKR3-mediated β-arrestin recruitment and interferes with CXCL12(1–68)/CXCR4 signaling via partial antagonism to CXCR4-mediated β-arrestin recruitment and G-protein signaling similar to mavorixafor. Another possible mechanism for CXCL12(3–68) is ACKR3-mediated internalization of the CXCR4 receptor.

Another important factor that complicates receptor affinity and activity for CXCL12(3–68) is the ability of both CXCR4 and ACKR3 to readily undergo homo- and hetero-dimerization.⁵¹⁻⁵³ This ability to dimerize regulates receptor functions and affects their downstream signaling events.^51,54−56 Although receptor and ligand dimerization is outside the scope of this work, we decided to investigate downstream canonical signaling pathways to clarify how CXCL12 and its DPP4-cleaved metabolite affect SSC signaling events. In this case, we used the longer-half-life CXCL12 beta isoforms. RPPA screening of 301 individual proteins and phosphoproteins involved in signaling pathways was followed by IPA, and we found that indeed, many proteins, phosphoproteins, and canonical signaling pathways were being differentially modulated by CXCL12(1–72) and (3–72) in multiple SSC lines.

An important question that remains to be answered is whether the herein reported functional outcomes of CXCL12(3–68) are mediated by signaling via its ACKR3 agonism or CXCR4 β-arrestin competitive antagonism to CXCL12(1–68). In other words, does CXCL12(3–68) induce senescence and inhibit differentiation and migration via its own signaling cascades, or does it interfere with CXCL12(1–68) signaling leading to those outcomes? Although we have not completely answered this question, we believe that it could be argued that the effects we have seen here are a mix of both independent ACKR3 signaling and antagonistic CXCR4 signaling. Further studies modulating the levels of these two receptors could be very beneficial to accurately answer this question in future work.

These effects on receptor signaling support that the DPP4 cleavage does not simply end CXCL12’s ability to signal but actively changes the signaling and functional outcomes—often in opposition to the original actions. As such, CXCL12(3–68) may increase the specificity of CXCL12(1–68) signaling and sharpen the end of the signaling. Functionally, CXCL12(3–68) may limit and may even reverse or change the actions of the initial signaling event since it activates a different set of downstream signaling pathways and gene expression. This may also be critical in mediating the roles of CXCL12 in cell dysfunction and pathological conditions.

These studies have significantly added to the literature supporting the idea that DPP4 proteolytic cleavage generates metabolite ligands that display cryptic bioactivity in a growing number of peptide hormones, and cytokines/chemokines, including CXCL12. These metabolites can possess either a modified receptor selectivity, modified bioactivity, new antagonistic activity, or even a novel activity relative to the intact parent ligand.³⁴ Unfortunately, these metabolite ligands with their own bioactivity are understudied and generally misinterpreted as being simply inactive. Importantly, this highlights a gap in prior experimental rigor which should be addressed because of the confusion, misinterpretation of data, and missed translational clinical opportunities in basic and clinical studies.³⁴

In conclusion, the data presented in this work demonstrates that DPP4-cleaved CXCL12 has critical biological actions affecting cell and tissue function, including inhibition of migration, differentiation, function, and cell fate of SSCs and at least cell differentiation of osteoclast progenitors. Understanding this can be exploited for many clinical applications in the bone field and numerous other fields including immunology, cancer, and cardiovascular and potentially neurodegenerative diseases.

4. Methods

4.1. Animals

C57BL/6J mice were either provided by the National Institute on Aging (Bethesda, MD, USA) aged rodent colony or purchased from the Jackson Laboratories (Bar Harbor, ME, USA). Animals were maintained at Augusta University in the Division of Laboratory Animal Services Facility or the Animal Research Facility of the Veterans Affairs Medical Centers in Charleston, South Carolina. All aspects of the animal research were conducted in accordance with the guidelines set by the Augusta University Institutional Animal Care and Use Committee (AU-IACUC) under AU-IACUC-approved Animal Use Protocols or the Ralph H. Johnson VAMC IACUC. Mice were maintained on a standard 12 h light–12 h dark protocol and permitted water and food ad libitum.

4.2. Isolation and Culture of Murine SSC/BMSCs

Murine SSCs were derived from 3, 6, 18, and 24 month old male C57BL/6J mice at the Augusta University Stem Cell Core Facility. The SSC isolation process and the MSC characterization and multi-lineage potential (osteogenic, adipogenic, and myogenic) have been described previously.^2,57−59 In brief, six to eight mice of each age were euthanized by carbon dioxide (CO₂) overdose followed by thoracotomy. Whole bone marrow aspirates were flushed from femora and tibiae and SSCs isolated by negative immunodepletion using magnetic microbeads conjugated to anti-mouse CD11b (#558013) and CD45R/B220 (#551513) (BD Biosciences Pharmingen, San Diego, CA, USA), CD11c, and the plasmacytoid dendritic cell antigen (PDCA)-1 (#130-092-283 Miltenyi Biotec, Auburn, CA) followed by positive immunoselection using anti-stem cell antigen (Sca)-1 microbeads (#130-092-529 Miltenyi Biotec, Auburn, CA), according to the manufacturer’s recommendations. Enriched single-age pooled SSC populations were maintained in Dulbecco’s modified Eagle’s medium (#10-014-CM DMEM; Cellgro, Mediatech, Manassas, VA, USA) supplemented with 10% heat-inactivated fetal bovine serum (FBS) (#S11150 Atlanta Biologicals, Lawrenceville, GA, USA) and used at 60–70% confluency. Cell isolate phenotypes were confirmed by fluorescence activated cell sorting for the expression of positive (Sca-1, CD29, and CD44) and negative (CD45, CD11b) murine BMSC markers, with labeled isotype control antibodies.^58,59

4.3. Isolation and Culture of Human SSC/BMSCs

A direct isolation procedure was used to rapidly capture human SSCs directly from bone. Bone marrow aspirates from the proximal tibia (knee replacement surgery), proximal femur (hip replacement surgery), or iliac crest (spinal fusion surgery) were collected as orthopaedic surgical waste under institutional review board (IRB) approval in ethylenediamine tetraacetic acid (EDTA) blood collection tubes. The bone marrow aspirates were then run over a Ficoll gradient to collect the buffy coat within 30 min of collection. CD271 positive (+) SSCs were then isolated from the nucleated cell layer using CD271 MicroBead Kits (#130-099-023 Miltenyi Biotec, Auburn, CA) according to the manufacturer’s protocol to obtain a highly enriched SSC population within 2 h of bone marrow aspiration. CD271+ MSCs were isolated directly from bone marrow aspirates, washed with standard culture medium composed of DMEM (#10-014-CM Corning), 1% antibiotic–antimycotic (#15240–062 AA; Invitrogen), and 15% FBS, with low glucose (1g/l), transferred to a 100 mm cell culture dish, and incubated at 37 °C in a humidified atmosphere in 5% carbon dioxide (CO₂). After 24 h, the medium with non-adherent cells was removed, and the adherent cells were carefully washed with Dulbecco’s phosphate-buffered saline (DPBS) (#SH30028 HyClone) and further expanded in fresh culture medium. Culture-expanded CD271 + MSCs of passages 1–2 were used for in vitro studies. The isolated cells were first confirmed by fluorescence activated cell sorting analysis (# FMC020, R&D Systems, Minneapolis, MN) to be positive for CD73, CD90, and CD105 and negative for CD34, CD45, CD11b, CD79A, and HLA-DR with corresponding isotype control antibodies being used (R&D Systems Human Mesenchymal Stem Cell Verification Flow Kit Cat# FMC020 with FlowJo software for analysis; see Eisa et al., 2021 and Figure S7).⁶⁰ Cells were sorted according to human MSC phenotype characterization criteria set by the International Society for Cellular Therapy (ISCT) to define hMSCs.⁶¹⁻⁶⁶

4.4. Treatment Preparation

Recombinant human CXCL12(1–68) (#PFP001) and (1–72) (#PFP002) were purchased from Protein Foundry (Milwaukee, WI). DPP4-cleaved variants CXCL12(3–68) and (3–72) were custom made by Protein Foundry. The required concentrations of CXCL12 were prepared by dissolving it in Milli-Q distilled water. The concentrations prepared for functional and RPPA assays were 200 ng/mL which ≈25 nM. For signaling assays including BRET and cAMP assays, a range of concentrations from 10^–13 M to 10^–5 M was used. To inhibit DPP4 activity, sitagliptin (#MK-0431 Selleckchem, Houston, TX) was used in some functional assays in a concentration of 1 μM. AMD3100 and mavorixafor were used in different experiments as competitive CXCR4 antagonists. Depending on the assay, the control group received either cell growth medium free of any treatment (DMEM or MEM Alpha depending on the assay) or assay buffer medium (BRET buffer or PBS).

4.5. Osteogenic Differentiation Assays

The ability of culture-expanded MSCs to differentiate into the osteogenic lineage was validated according to earlier described methods.^2,67 In brief, cells were plated in 12-well plates at 50000 cells/cm² and cultured in DMEM for 24 h. Culture medium was then aspirated and replaced with osteogenic differentiation medium consisting of DMEM supplemented with 10% FBS, 1%AA, 0.25 mM ascorbic acid (#A4544 Sigma-Aldrich), 0.1 μM dexamethasone (#D4902 Sigma-Aldrich), and 10 mM β-glycerophosphate (#G9891 Sigma-Aldrich). Treatment-containing medium was replaced two times per week. The early osteogenic differentiation marker, alkaline phosphatase, was assessed in cell culture medium after 7 days using an Alkaline Phosphatase Assay Kit (#ab83369 Abcam). After 3 weeks, osteogenic differentiation was assessed by staining with Alizarin Red staining solution (#TMS-008 C Millipore Sigma). The cells were fixed with 10% phosphate-buffered formalin for 30 min at room temperature (RT) and stained with Alizarin Red staining solution for 45 min at RT. Stained monolayers were visualized by phase-contrast microscopy using an inverted microscope (Nikon, Melville, NY). Differentiation was quantified as previously described.⁶⁸ In brief, cells were destained using 10% cetylpyridinium chloride (#855561 Sigma-Aldrich), and collected samples were analyzed using a microplate reader at 570 nm.

4.6. Cell Density Assay

To determine whether treatments affected SSC density, we utilized a Crystal Violet Assay Kit (#ab232855 Abcam) according to the manufacturer’s protocol. In brief, SSCs were plated in 96-well plates at 5000 cells/well and cultured in DMEM for 24 h. Culture medium was then aspirated and replaced with osteogenic differentiation medium as previously described. After 3 days, the cell culture medium was removed, and the cells were washed and stained with the crystal violet staining solution for 20 min at RT. Then, the staining solution was removed and washed, and the remaining stain was solubilized for 20 min with the solubilization solution. Finally, the crystal violet stain was quantified using a microplate reader at 595 nm.

4.7. Bioluminescence Resonance Energy Transfer (BRET) Arrestin Assays

To measure CXCR4- or ACKR3-mediated β-arrestin 2 recruitment as measured by BRET,¹ HEK293T cells (ATCC CRL-11268; 59587035; mycoplasma free) in DMEM supplemented with 10% FBS were co-transfected with codon-optimized human ACKR3 or CXCR4 Tango constructs with the V2 tail/TEV/tTA regions removed and replaced with an IDTG linker followed by a fused Renilla luciferase (RLuc8) and a Venus-tagged N-terminal β-arrestin 2 in a 1:15 ratio using TransiT-2020 (Mirus) in a 3:1 ratio. To measure CXCR4 G protein dissociation via BRET,² HEK293T cells in DMEM supplemented with 10% FBS were co-transfected in a 1:1:1:1 ratio with individual RLuc8-fused Gαi/o subtypes, a GFP² fused to the C-terminus of human Gγ2, human Gβ1, and CXCR4. After at least 18–24 h, transfected cells were plated in poly-lysine-coated 96-well white clear bottom cell culture plates in DMEM containing 1% dialyzed FBS at an approximate density of 40,000 cells in 200 μL per well and incubated overnight. The next day, medium was decanted, and cells were washed with 60 μL of drug buffer (1× HBSS, 20 mM HEPES, 0.1% BSA, 0.01% ascorbic acid, pH 7.4), and then, 60 μL of drug buffer was added per well. Plates were pre-equilibrated at 37 °C for at least 15 min before receiving 30 μL of drug (3X) and incubated for 60 min at 37 °C before reading. Fifteen minutes before reading, 10 μL of the 10X RLuc substrate (5 μM final concentration; BRET¹ coelenterazine h; BRET² coelenterazine 400a) was added to each well. BRET¹ plates were read for 485 nm and 530 nm emission wavelengths and BRET² at 400 nm and 510 nm emission wavelengths for 1 s per well using a Mithras LB940. BRET acceptor over donor ratios were calculated and plotted as a function of the ligand concentration using GraphPad Prism (GraphPad Software Inc., San Diego, CA). In some cases, data were normalized to %CXCL12 stimulation and analyzed using nonlinear regression “log(agonist) versus response.”

4.8. GloSensor Gi/o-Mediated cAMP Inhibition Assay

HEK-293T cells were transfected with the codon-optimized CXCR4 Tango construct with the V2 tail/TEV/tTA regions removed and the cAMP GloSensor-22F (Promega) plasmids in a 1:1 ratio using a 3:1 polyethyleneimine transfection mix. Next day, cells were transferred to poly-l-lysine-coated 384-well white clear-bottom plates seeded at approximately 10,000 cells per well in 1% dialyzed FBS in DMEM and incubated for 24 h. On the day of the assays, plates were decanted, and 20 μL of assay buffer (1x HBSS, 20 mM HEPES, 0.1% BSA, 0.01% ascorbic acid, pH 7.4) containing 4 mM luciferin was added to each well. The assay was started with the addition of the chemokine diluted in assay buffer and dispensed using a FLIPR (Molecular Devices). After 15 min at RT, cAMP accumulation was initiated by forskolin (1 μM). After an additional 15 min, luminescence was measured using a MicroBeta TriLux plate counter (PerkinElmer), and data were analyzed using non-linear regression using GraphPad Prism.

4.9. Reverse Phase Protein Array and Ingenuity Pathway Analysis

RPPA assay was done at the MD Anderson’s Functional Proteomics RPPA Core Facility according to their protocol. Briefly, SSCs from 3, 6, 18, and 24 month old mice were treated with growth medium, CXCL12(1–72), CXCL12(3–72), sitagliptin (#S5079 Selleckchem), or sitagliptin plus CXCL12(1–72). After 30 min, medium was aspirated, and the cells were washed twice with cold PBS before incubating the plate on ice for 20 min. Cells were then scraped and collected before being centrifuged at 14,000 rpm for 10 min at 4 °C. The cellular protein concentration was determined using a Pierce BCA protein assay kit (#23225 Thermo Scientific). Finally, the protein concentration was adjusted to be the same for all samples, and they were mixed with sodium dodecyl sulfate and beta-mercaptoethanol (B-ME) sample buffer before being shipped to MD Anderson.

Tissue or cell lysate samples were serially diluted twofold for five dilutions (undiluted, 1:2, 1:4, 1:8; 1:16) and arrayed on nitrocellulose-coated slides in an 11 × 11 format to produce sample spots. Sample spots were then probed with antibodies by a tyramide-based signal amplification approach and visualized by a 3,3′-diaminobenzidine colorimetric reaction to produce stained slides. Stained slides were scanned on a Huron TissueScope scanner to produce 16-bit tiff images. Sample spots in tiff images were identified and their densities quantified using an Array-Pro Analyzer. Relative protein levels for each sample were determined by interpolating each dilution curve produced from the densities of the five-dilution sample spots using a “standard curve” (SuperCurve) for each slide (antibody). SuperCurve is constructed by a script in R, written by Bioinformatics. Relative protein levels are designated log₂ (log base 2) values. All relative protein level data points were normalized for protein loading and transformed to linear values, and linear values were subsequently transformed to log₂ values.

RPPA slides were stained for 301 unique antibodies, which were analyzed using Array-Pro Analyzer 6.3 and then using SuperCurve_1.5.0 via SuperCurveGUI_2.1.1. QC tests were performed for each antibody staining (slide). A QC score above 0.8 indicates good antibody staining. We included only the data for the 301 individual antibodies with QC scores higher than 0.8 in the heatmaps. The list of used antibodies is included in Table S1.

RPPA fold changes with p-values (<0.15) for individual proteins and phosphoproteins were then uploaded into the IPA pathway analysis system. IPA utilizes a proprietary protein-interaction database to calculate a probability that a specific canonical signaling pathway is activated or inhibited under different conditions. Analysis for canonical signaling pathways was run for all relevant pairwise comparisons. To understand the changes in signaling pathways occurring with CXCL12(3–72) treatment, we focused on pairwise comparisons of CXCL12(3–72) versus control and CXCL12(3–72) versus CXCL12(1–72) + sitagliptin. Although the pairwise comparison of CXCL12(3–72) versus CXCL12(1–72) shows significant changes in protein and phosphoprotein expression (Figure S6A and Table S8), it is less than when DPP4 is inhibited from the beginning of the experiment. The IPA output was exported as Microsoft Excel files to prepare Tables S2–S7. Changes in canonical pathways with the p-value <0.05 or -log(p-value) ≥1.3 were considered statistically significant, and a Z-score of ≥1 was considered significant pathway activation and ≤1 significant inhibition.⁶⁹

4.10. PRESTO-Tango GPCRome Screen

To determine whether intact or cleaved CXCL12 could signal through GPCRs other than the two known cognate receptors CXCR4 and ACKR3, we reached out to the National Institute of Mental Health (NIMH) Psychoactive Drug Screening Program where we could test the 2 peptides on 320 different GPCRs using a PRESTO-Tango assay as part of their guanine nucleotide-binding protein (G-protein)-coupled receptor “omics” screen.

The primary PRESTO-Tango screen was performed as previously described.^36,70 Briefly, HTLA cells, stably expressing a tTA-dependent luciferase reporter and a β-arrestin2-TEV protease fusion gene, were maintained in DMEM supplemented with 10% FBS, pen–strep, and 2 μg/mL puromycin and 100 μg/mL hygromycin B in a 384-well plate format. At least 1 h before transfections, the cells were fed with 10 μl/well DMEM supplemented with 50% FBS, using a Multidrop automated liquid dispenser. HTLA cells were transfected with DNA plasmids using the calcium phosphate precipitation protocol. One GPCRome (up to 320 Tango constructs) screening for each compound consists of a total of 8 384-well assay plates, with each receptor being screened in quadruplicate. After overnight transfection and incubation, cells were removed from medium and received 40 μL/well fresh DMEM supplemented with 1% FBS about 2 h before compound stimulation. Test compounds and the assay control (quinpirole) were made in DMEM with 1% FBS at 5x working concentration. Drugs (10 μl/well) were transferred using the Hamilton Microlab Star with a 384-well pipetting head. Assay plates are then incubated overnight at 37 °C. The following day, medium and drug solutions were removed, and 20 μL per well of the Bright-Glo reagent (#E2610 Promega) was added. Plates were incubated for 20 min at room temperature in the dark before luminescence was measured.

Secondary assays for receptors that showed a high relative luminescence and low CV in the primary screen were done in 16 different concentrations and also in quadruplicate.

4.11. Transwell Migration Assay

SSCs were seeded in 100 mm dishes in low-serum medium (1% FBS). 16 h prior to the experiment, cells were washed three times with 1x PBS and incubated at 37°c with fresh DMEM (phenol red-free) containing 1% FBS and treatment. Medium containing different concentrations of intact CXCL12 was added to the 96-well black receiver plate (#3583 Corning). The transwell inserts (#3374 Corning) were then transferred to the black receiver plate. The cells pre-incubated with different treatments for 18 h were then harvested by trypsinization and resuspended in serum-free medium, before adding 50 μL of the cell suspension (0.4 × 10⁵ cells/ml) to the transwell inserts. The plate was then incubated in a cell culture incubator for 6 h. Prior to the end of incubation, 1 ml of phenol red-free trypsin (#15400–054 Gibco), CyQuant dye (400x), and cell lysis buffer (20x) from the CyQuant Cell Proliferation assay kit (#C7026 Molecular Probes) were diluted in 9 ml of Hanks’ balanced salt solution (HBSS) buffer. The medium from the transwell plate was aspirated, and 100 μL of the CyQuant dye-containing solution was added to each well before incubation in the cell culture incubator for another 1 h. Finally, the insert tray was removed, and fluorescence (485 nm excitation and 520 nm emission) in the black receiver plate was measured with a Spectramax M5 plate reader (Molecular Devices).

4.12. Scratch Wound Healing Assay

Migration of SSCs stimulated by 25 nM CXCL12(1–68) and (3–68) was assessed using a scratch wound healing assay.^71,72 Cells were grown in 24-well plates until confluence. A scratch was created in each well using a 300 μL pipette tip. Images were taken on an Olympus IX73 microscope (Olympus Scientific Solutions Americas Corp., Waltham, MA, USA) 24 h after treatment. The scratch wound healing % was quantified by measuring the average width of the scratch at time 0 and 24 h using open-source image analysis software ImageJ.

4.13. Senescence-Associated β-Galactosidase Assay

To detect senescence in SSCs treated with CXCL12(1–68) or (3–68), the increase in senescence-associated beta-galactosidase (SA-β-gal), which accumulates in senescent cells,^73,74 was detected using a SA-β-gal senescence assay kit (Cell Signaling, cat# 9860, Denver, MA). Briefly, 48 h after treatment, SSCs were washed with PBS and fixed with the kit’s fixative solution and stained with β-gal reaction solution at 37 °C for 24 h according to the manufacturer’s protocol. Images were taken using an Olympus IX73 microscope with a bright field, and the number of SA-β-gal-positive cells was counted for each treatment condition.

4.14. Osteoclastogenesis Assay

For osteoclast differentiation, 5 × 10³ RAW-264.7 cells/well were cultured in a 96-well plate in complete DMEM supplemented with 100 ng/ml receptor activator of nuclear factor kappa-Β ligand (RANKL). To test the effect of intact and cleaved CXCL12 treatment on osteoclast differentiation, cells were treated with 25 nM CXCL12(1–68) or (3–68) or 1 μM sitagliptin with or without CXCL12(1–68) and kept in a humidified 5% CO₂ incubator at 37 °C for 5 days with medium being replaced every other day. Multinucleated cells, indicating the fusion of osteoclast precursors into multinucleated osteoclasts, started to appear at day 3. On day 5, numerous mature osteoclasts with three or more nuclei were formed. At the end of day 5, cells were fixed and stained following instructions of the TRAP staining kit (#387, Sigma-Aldrich, St. Louis, MO, USA). The assay demonstrates the differentiation of precursor cells into an osteoclast phenotype as TRAP activity increases; the acid phosphatase enzyme activity as naphthol AS- BI phosphate is released by enzymatic hydrolysis which then couples with fast garnet GBC salt forming highly insoluble reddish/brown dye deposits.

Four images per well, covering all the well surface, were taken using an ImageXpress Micro High Content Imager (Molecular Devices, San Jose, CA, USA). TRAP-positive cells with three or more nuclei were blindly counted as osteoclasts. Representative images for each well were taken using Keyence BZ-X710 microscope at 10× magnification.

4.15. Statistical Analysis

Experiments were performed at least three independent times with different cell lines (i.e., biological replicates) when possible and with different vials of cells when the experiment was cell-line-specific (e.g., BRET assays in HEK293 cells). The figures shown are representative of three independent experiments, each performed with at least three technical replicates. The exact number of technical replicates is shown for each panel or mentioned in the figure legend. Data were expressed as means ± SD unless stated otherwise. Data were analyzed using GraphPad Prism 7.0 software (GraphPad Software Inc., La Jolla, CA, USA). Student’s t-test was used for comparisons between two groups, and analysis of variance (ANOVA) followed by Tukey’s multiple comparison test was used for comparisons between three or more groups. Null hypotheses were rejected at the 0.05 significance level. Statistical significance is shown in figures and figure legends with * meaning p-value <0.05, ** meaning p-value <0.01, *** meaning p-value <0.001, and **** meaning p-value <0.0001.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsptsci.2c00040.

• CXCL12(3–728) inhibiting osteogenic and osteoclastogenic differentiation; CXCL12(3–72) inducing senescence and inhibiting migration in murine SSCs; follow-up screen for selected receptors CCR6, GPR61, GPR146, GPR150, and GPR151 using the PRESTO-Tango system; CXCR4 β-arrestin recruitment assays and forskolin-induced cAMP levels for CXCL12(1–68) and CXCL12(1–72); BRET assay for measuring CXCR4-mediated G protein activation at individual Gi and Go subunits; heatmaps for different pairwise comparisons from RPPA; and human BMSC phenotyping assay (PDF)

• Antibody list used in RPPA analysis at MD Anderson, significant changes in IPA canonical signaling pathways for CXCL12(3–72) group versus control group, significant changes in IPA canonical signaling pathways for CXCL12(3–72) group versus CXCL12(1–72) + sitagliptin group, significant changes in IPA canonical signaling pathways for CXCL12(1–72) group versus control group, significant changes in IPA canonical signaling pathways for sitagliptin group versus control group, significant changes in IPA canonical signaling pathways for CXCL12(1–72) + sitagliptin group versus control group, significant changes in IPA canonical signaling pathways for CXCL12(1–72) group versus CXCL12(1–72) + sitagliptin group, and significant changes in IPA canonical signaling pathways for CXCL12(3–72) group versus CXCL12(1–72) group (XLS)

Author Contributions

Study design: A.M.E. and W.D.H. Study conduct: A.M.E., N H.E., S.P.-T., G.K., X.-M.S., C.R., M.M.C., J.D.M., and A.A.-P. Data collection: A.M.E., N.H.E., S.P.-T., M.M.C., J.D.M., and G.K. Data analysis: A.M.E., J.B.B., L.M.L., R.C.M., A.C.L., M.M.C., J.D.M., and M.E.M.-L. Statistical analysis: A.M.E., J.C., and M.J.. Data interpretation: J.B.B., L.M.L., M.B.D., M.M.C., J.D.M., B.F.V., and W.D.H. Drafting manuscript: A.M.E., N.H.E., D.K., and W.D.H. Revising manuscript content: A.M.E., J.B.B., L.M.L., R.C.M., A.C.L., M.E.M.-L., M.M.C., M.B.D., J.D.M., B.F.V., and W.D.H.. Approving final version of the manuscript: A.M.E., J.B.B., L.M.L., M.M.-L., J.D.M., W.D.H.. A.M.E., and W.D.H. taking responsibility for the integrity of the data analysis.

The authors declare the following competing financial interest(s): Brian F. Volkman reports a potential COI based on ownership of the Protein Foundery, LLC, which generated the CXCL12 isoforms used in this study.