Is GPNMB the Achilles' Heel of Mo-MDSC While Marking Their Suppressive Activity?

In this issue of Clinical Cancer Research, Kobayashi and colleagues report on glycoprotein nonmetastatic B (GPNMB), also known as DC-Hil/Osteoactivin, which is expressed on CD14⁺ monocyte-type myeloid-derived suppressor cells (Mo-MDSC), rather than granulocyte-type, and marks the cells that effectively inhibit T lymphocyte through the binding to syndecan-4 (SDC-4) expressed on T cells (1). This is relevant because of the paucity of markers that phenotypically could identify MDSCs endowed with suppressor activity, whereas other markers, such as SiglecF on neutrophils, although protumorigenic, are undefined for functional suppression. Also relevant is the availability of antibodies already tested in the clinic and capable of blocking GPNMB in its checkpoint-inhibitory capacity. Indeed, because of GPNMB overexpression in a variety of tumors, especially melanoma, lung, and triple-negative breast cancer, targeting antibodies have been developed and glembatumumab vedotin, which is a drug conjugate, is currently being tested in clinical trials. Although the drug moiety is active against proliferating cells, thus sparing MDSCs, the blocking of GPNMB–SDC-4 interaction should be conserved. An issue is the specificity of GPNMB expression, which is also on dendritic cells and macrophages, likely under the same transcriptional regulation via MITF in an IL10-rich immunosuppressive milieu or in injured tissue under repair, or attempting to repair (like cancer: a wound that does not heal). GPNMB is also expressed on melanocytes, keratinocytes, and osteoblasts. These commonalities broaden the potential efficacy of GPNMB blockage. Additional evidence underscoring the involvement of GPNMB in maintaining a tolerogenic environment also comes from the analysis of placenta where the CD14 HLA-DR^high population expresses 80 times more GPNMB than the HLA-DR^low counterpart, along with other immune-inhibitory molecules such as p28 (a common subunit of IL27 and IL35), IDO, and PD-L1 (2). Notably, the HLA-DR^high population shows high expression of MITF. Despite the discrepancy in HLA-DR expression, which is invariably low in MDSCs, the functional commonalities for sustaining a tolerogenic environment are consistent between the same cancer microenvironment and placenta.

Also noticeable is the possible coexpression of GPNMB and PD-L1 on myeloid cells with the latter at lower level but increasing in the presence of T cells produced IFNγ, which may well become available in the tumor microenvironment because of the initial blockage of GPNMB–SDC-4 interaction (Fig. 1, enlargement). This event might support the possible combination of Abs to GPNMB and PD-L1/PD-1 for targeting two checkpoint inhibitors, together. However, different from the well-known mechanism of T-cell inhibition exerted by PD-1, through SHP-1 or SHP-2 and final attenuation of the PI3K and Akt pathway, little is known about how SDC-4 could inhibit T-cell function upon binding to GPNMB.

SDC-4 is ubiquitously expressed and woks as low-affinity coreceptor for heparin-binding growth factor. In addition, it has promiscuous binding capacity owing to its participation in several extracellular and intracellular signaling processes (Fig. 1). One converging pattern eventually activating AKT is the SCD-4–dependent activation of PKCα for the assembly of the mammalian target of rapamycin (mTOR) complex 2 and PDK1 (3). Whether this pathway is inhibited by GPNM binding SCD-4 on T lymphocyte remains unknown. Rather, one of the authors has previously reported that GPNMB binding to SDC-4 activates a process that leads SDC-4 assembly with CD148 such as to upregulate the protein tyrosine phosphatases activity of CD148 (4), without specifying which phosphoproteins were hypophosphorylated. Knockdown of CD148 in activated T cells confirmed that the inhibitory effect of GPNMB through SDC-4 in response to CD3 stimulation was largely lost. Other experiments demonstrate that SCD-4 assembly with CD148 occurs away from the TCR/CD3 complex, where instead PD-1 is localized.

In all aspects, GPNMB expressed by myeloid cells seems responsible for inducing inhibition of T lymphocytes expressing SDC-4. This has been further confirmed by Brossard and collaborators showing that addition of IL10 to peripheral blood monocytes, induced toward dendritic cell (DC) differentiation with IL4/GM-CSF, results in cells of low stimulatory capacity characterized by upregulation of both GPNMB and SCD-4, along with GITR/GITRL and PD-L1 and that their blockage by specific antibodies can restore activation of T cells in in vitro coculture (5). Furthermore, they report about IL10-induced upregulation of Dectin-1 that upon biding with the beta-glucan curdlan leads to the reduction of immunosuppressive molecule, GPNMB included, suggesting an additional means to inactivate MDSCs.

In conclusion, GPNMB can be proposed as a relevant marker to identify MDSCs endowed with suppressive activity and proposed as a good target to aim at in order to block its interaction with SDC-4 on T lymphocytes. This increases the armamentarium of molecules inhibiting MDSC-suppressive activity (phosphodiesterase-5 inhibitors, STAT-3 neutralization), MDSC recruitment (chemokine or chemokine receptor blockade), or reducing MDSC frequency (normalization of tumor-induced myeloproliferation, induced cell death or transdifferentiation into macrophage or DC) to unleash T lymphocytes from immunosuppression.

The efficacy of immune-checkpoint inhibitors is remarkable but limited to a fraction of patients, MDSCs can hamper such efficacy in many cases, and their depletion/inactivation is a step that can be immediately undertaken to improve the number of benefiting patients, a possibility that is currently being tested in some clinical trials.

While awaiting a robust and extensive in vivo validation using primary autochthonous mouse models, the work by Kobayashi and colleagues envisages the combination of GPNMB with PD-1/PD-L1 blockers as suitable to avoid Mo-MDSC immunosuppression.

No potential conflicts of interest were disclosed.

The author was supported by Associazione Italiana per la Ricerca sul Cancro (IG 18425).