Derangement of immune responses by myeloid suppressor cells

Abstract

In tumor-bearing mice and cancer patients, tumor progression is often associated with altered hematopoiesis leading to the accumulation of myeloid cells. Extensive studies in preclinical models indicate that these cells share the CD11b and the Gr-1 markers, possess a mixed mature-immature myeloid phenotype, and are responsible for the induction of T-cell dysfunctions, both tumor-specific and nonspecific. Moreover, CD11b⁺Gr-1⁺ myeloid cells are described under different unrelated situations associated with temporary impairment of the T-lymphocyte reactivity. This review examines recent findings on the nature, properties, and mechanisms of action of these myeloid suppressor cells (MSCs).

Introduction

Several studies have reported inhibition of T-cell responses by a cell population that has been identified phenotypically with the mAbs M1/70 and RB6-8C5. These cells are isolated from the spleen, bone marrow, and tumor infiltrate where they appear to be mediating immunosuppressive effects in a wide variety of unrelated pathologic conditions (Table 1). M1/70 mAb recognizes the 170-kDa αΜ subunit of Mac-1 (CD11bCD18, αΜβ2 integrin), the complement receptor 3 (CR3), which mediates adhesion to C3bi and ICAM-1 (CD54). The antigen is present on the surface of granulocytes, macrophages, dendritic cells (DCs), natural killer (NK) cells, and B-1 cells in the body cavities ([44, 51], and http://www.ncbi.nlm.nih.gov/PROW/guide/680557536_g.htm). The RB6-8C5 antibody recognizes Ly-6G previously known as Gr-1, a 21–25 kDa GPI-anchored protein. Gr-1 is generally held as a marker of granulocyte differentiation/maturation in the bone marrow, and transiently expressed during monocyte but not erythroid differentiation [25]. In the periphery, RB6-8C5 mAb stains granulocytes (neutrophils and eosinophils), and monocytes [42]. Some studies but not others have also reported the staining of plasmacytoid DCs, perhaps due to cross-reactivity with other Ly6 molecules [10, 60]. Table 1 shows that CD11b and Gr-1 are the most common markers found in suppressor cells. It must be pointed out, however, that Gr-1 expression can be low-dull in suppressor cells isolated from some mouse strains, such as C57BL/6, whereas the anti-CD11b mAb usually provides high-intermediate intensity staining (unpublished data). The CD11b⁺Gr-1⁺ population is heterogeneous, enclosing mature granulocytes, monocytes, and varying numbers of immature cells of the myelomonocytic lineage identified by another set of markers (commonly CD31, but also ER-MP58, Ly-6C, and ER-MP54) [6, 17, 62]. The phenotype of these cells and their ability to form colonies in agar suggest that a subset of them comprises myeloid precursors, able to originate granulocytes and monocytes with a prevalence of CFU-GM and CFU-GEMM ([52], and V.B. unpublished results). Other markers that might be related to the inhibitory activity of these cells (CD80, [54]) their prevalent monocytic orientation (F4/80, [16, 46, 78]) and possible function (CD16; [49]) have been described either in freshly isolated cells or in cells cultured in vitro without cytokines for few days.

Table 1.

Examples of systemic accumulation of MSCs

Location	Phenotype	Immune dysfunctions	References
Tumors
Transplantable tumor (CNS5, MethA, MCA-26)	CD11b/Gr-1	Yes	[26, 41, 46, 69]
Transplantable tumor (TS/A, CT26)	CD11b/Gr-1/CD31	Yes	[16, 17]
Transplantable tumor (T9)	CD11b/His48a	Yes	[65]
Transplantable tumor (LLC)	CD31	Yes
Spontaneous tumor (BALB/NeuT mice)	CD11b/Gr-1/CD31	Yes	[52]
Chemical carcinogenesis	CD11b	Yes	[32]
Infection
Candida albicans	CD11b/Gr-1/CD80	Yes	[54]
Trypanosoma cruzi	CD11b/Gr-1	Yes	[29]
Schistosoma mansoni	CD11b/Gr-1/CD16	Yes	[49]
Immune stress
Superantigens (staphylococcal enterotoxin A)	CD11b/Gr-1/LFA-1	Yes	[21]
Schistosome oligosaccharide (Lacto-N-neotetraose)	CD11b/Gr-1/F4/80	Yes	[78]
Vaccination with recombinant vaccinia virus	CD11b/Gr-1/CD31	Yes	[15, 17]
Others
Cyclophosphamide treatment (recover phase)	CD11b/Gr-1/CD31	Yes	[6, 62]
Radiation marrow chimeras	CD11b/Gr-1/Ly6-C	Yes	[12]
Graft-versus-host reaction	CD11b	Yes	[13]

^aThis is the only study shown in the table that was performed on rats

Due to the heterogeneity and the absence of conclusive data on the ultimate cell type causing T-lymphocyte dysfunctions, we proposed to use the acronym MSC (myeloid suppressor cell), rather than IMC (immature myeloid cell), to denote this population [18]. Many aspects of the biology, relevance in physiologic and pathologic situations, and mechanisms of actions of MSCs have been addressed in recent publications [18, 20, 40]. The present review will mainly focus on the interaction between MSCs, tumors, and cells of the immune system, to offer some insight on the cellular and molecular mechanisms that can suppress anti-tumor responses.

MSCs and tumor

Mice bearing transplantable tumors, but also mice developing cancer as a consequence of chemical carcinogenesis or activation of tissue-restricted, transgenic oncogenes, undergo a progressive accumulation of MSCs in the spleen and blood where they can deleteriously suppress immune response to various antigens, including tumor-associated antigens (TAAs) (Table 1). The increase roughly parallels the extent of tumor burden and can be taken as an indirect measure of tumor dissemination [52]. The first evidence of a correlation between myeloid cell accumulation in lymphoid organs and immune suppression was found in the 1980s, when several groups described a cellular population named natural suppressor cells. These cells, distinct from T lymphocytes and NK cells, were found to suppress immune reactivity in the tumor-bearing host, in the early maturation of the lymphoid tissues in neonates, and after manipulation of the lymphoid tissues in adult mice [47, 73, 74]. The transplantable tumors commonly used in mouse studies, as well as many human tumor cell lines, share the capacity of secreting different cytokines which act on bone marrow precursors such GM-CSF, IL-3, M-CSF, IL-6, VEGF (reviewed in [18, 40]). These cytokines do not exert direct inhibitory activity on T lymphocytes; indeed for some of them, such as GM-CSF, adjuvant activity has been reported for peptide and cellular-based tumor vaccines [53, 81]. However, when these cytokines are released systemically in large quantities, by growing tumors or in situations of immune distress, they are responsible for the expansion of myeloid cell pool in secondary lymphoid organs, an activity that can even overcome the adjuvant effect, as in the case of GM-CSF and G-CSF [53].

Recent findings suggest that MSCs are not just a tumor-associated phenomenon, but might be relevant for the tumor history itself. The lymphoma BW-Sp3 grows progressively during the first 2–3 weeks from inoculation in syngeneic AKR mice. Subsequently, about 70% of the animals reject the tumor and exhibit a CD8⁺ T-cell-mediated anti-tumor immunity. However, tumor elimination fails in 30% of recipients resulting in a progressive, lethal tumor growth. Tumor progressors but not regressors showed a marked increase in CD11b⁺Gr-1⁺ MSCs expressing arginase 1 (ARG1) among the splenocytes and an impaired anti-tumor CTL activity that could be recovered by in vitro CD11b⁺ cell depletion [46]. Consistent with these findings, a shift from nitric oxide (NO) to urea/ornithine production in intratumor myeloid cells was correlated with tumor progression [55]. ARG1 converts l-arginine to l-ornithine and urea, and the resultant l-ornithine can be used by ornithine decarboxilase to form polyamines (i.e., putresceine, spermidine, and spermine), essential nutrients for mammalian cell proliferation and differentiation [83]. Macrophages transfected with the rat ARG1 gene promoted breast tumor cell proliferation in coculture experiments, an effect that was down-regulated by the ARG1 inhibitor l-norvaline [22]. Moreover, ARG1 was recently implicated directly in the suppressive activity exerted by MSCs on activated T lymphocytes [19, 20]. It is possible to speculate that intratumoral arginase induction might be beneficial for the tumor through different pathways: to support tumor growth and development by providing them with polyamines, and to suppress anti-tumor immune response by negatively affecting tumor-infiltrating T lymphocytes.

In another model, resection of a primary mouse fibrosarcoma was sufficient to normalize the number of CD11b⁺Gr-1⁺ MSCs, correct the p56^lck abnormalities in T lymphocytes, and restore tumor immunity. Resected mice, in fact, rejected a second challenge of the same tumor, unambiguous evidence that anti-tumor T lymphocytes were not deleted in tumor-bearing mice but simply rendered temporarily quiescent [69]. Adoptive transfer of CD11b⁺ splenocytes from moribund animals into rats bearing 5-day-old glioma caused rapid tumor progression, another example of the direct link between MSC presence and tumor development [64]. Transgenic female mice expressing the transforming rat oncogene c-erbB-2 under the MMTV promoter (BALB-NeuT mice) spontaneously develop mammary carcinoma with a progression resembling that of human breast cancer. In these mice activating anti-tumor immunotherapy failed to induce T-cell-mediated cytotoxicity, suggesting a suppression of the immune response. Recently a correlation between tumor multiplicity and an increased proportion of MSCs in the peripheral blood and in the spleen [52] has been shown. This myeloid recruitment depended on the tumor-persistent release of VEGF, which is released by most tumors, and whose production has been associated with poor prognosis [79].

Whereas numerous findings correlate immunosuppression with systemic MSC accumulation in hosts bearing large tumors, there is relatively little data about their role at the tumor site where they can be present at elevated percentage (I. M., unpublished data). One intriguing hypothesis is that while systemic MSC accumulation causes generalized immune dysfunctions as a late event in tumor progression, tumor-associated MSCs could promote tolerance in the first phases, thus assuming a relevant role in the tumor escape mechanism. In a C57BL/6 adenocarcinoma model, it was reported that tumor-associated macrophages (TAMs) can induce apoptosis in activated T cells by a mechanism requiring cell contact, TNF-α, and NO production [68]. In this model, priming and effector functions of CTLs are not affected by MSC accumulation in the spleen, and only CTL proliferation after TCR stimulation was impaired and could only be restored by MSC depletion in vitro [66]. TAM- or tumor-associated MSC-induced apoptosis of T cells is a conceptually attractive mechanism for a tumor to evade immune killing, and could explain some reports in which a clinically evident regression of metastatic tumor was not observed following vaccination with TAAs, even in the presence of a high expansion of tumor-specific CTLs [8, 43, 58].

Properties of MSCs

Although a large body of evidence links MSCs to tumor progression, these cells have also been detected in mouse lymphoid organs under cancer-unrelated situations: during graft-versus-host reactions [13]; infection with viruses, bacteria, fungi, and parasites [2, 29, 49, 54]; during immune stress caused by exposure to superantigens, immunomodulatory oligosaccharides present in helminths and human milk, or recombinant vaccines [15, 21, 78]; following myelosuppressive treatments by either cyclophosphamide or γ-irradiation [6, 12, 62]. All these seemingly unrelated conditions have in common two features: the functional paralysis of T lymphocytes that are unresponsive to mitogens, antigens, or monoclonal antibodies activating the TCR/CD3 complex, and the rapid proliferation in the myeloid compartment, often resulting in ectopic spleen hematopoiesis. It was recently proposed that MSC recruitment is not just a corollary of advanced tumor spreading but can serve an important role as a controller of either runaway immune responses or exaggerated expansion of the myeloid cell pool [18]. Cancer and immune stress, however, differ in one important aspect: accumulation of MSCs in lymphoid organs of infected or immunized mice is a temporary event, dependent on the time of the immune disturbance, and after 2–3 weeks, the number of these cells decreases to baseline levels leading to the recovery of the immune response (V.B, unpublished results). In contrast, the number of MSCs in tumor-bearing hosts increases steadily during tumor progression, since the tumor-derived factors that alter myeloid differentiation accumulate in proportion to tumor burden.

MSCs isolated from the spleen of immunocompromised animals and cultured in vitro in standard medium, without exogenous stimuli, differentiate spontaneously in major histocompatibility complex (MHC) class II negative, adherent macrophage-like cells that retain the inhibitory properties [16, 17, 46], whereas their culture in the presence of cytokines or all-trans-retinoic acid eliminates their suppressive properties [17, 26]. MSCs are powerful and unselective inhibitors since they interfere with the activation of CD8⁺ and CD4⁺ T cells, in an antigen- and MHC-independent fashion. Although investigated in few models, B lymphocytes are also unresponsive when stimulated in the presence of MSCs [63]. The concept of antigen independence might be misleading, since, in any case described so far, MSCs inhibit only activated T lymphocytes, either naïve or memory, whereas resting lymphocytes are spared. This is an aspect of MSC biology relevant for the corrections of MSC-dependent anomalies of the immune system. In fact, a simple removal of MSCs, in vitro or in vivo, is often sufficient to lift the suppression of lymphocyte reactivity, indicating that the antigen-specific lymphocytes are not deleted [15, 69]. On the other hand, stimulation during the windows of MSC presence in lymphoid organs can result in a disappearance of tumor-specific T lymphocytes leading to complete tolerance and loss of protection from tumor overgrowth in mice vaccinated with a model TAA [15].

A strict cell-cell contact between MSCs and T lymphocytes is required to exert their inhibitory activity on lymphocytes, as described in mouse studies [15, 16, 26, 33] and human studies [5]. When MSCs were separated from the T lymphocytes by a semipermeable membrane, in fact, suppression did not occur, indicating the intervention of either a membrane-bound molecule, a short-range effector mechanism, or rapidly degraded soluble factor. Characterization of this membrane factor has been elusive. It was shown that the cell-associated form of TNF-α can synergize with NO to induce apoptosis of tumor-infiltrating lymphocytes; however, common apoptosis-inducing ligands (FasL, TNF-α, lymphotoxin-β, or TRAIL) have been ruled out using cloned MSCs derived from the spleen of tumor-bearing mice [9]. Given the central and almost exclusive role of the l-arginine-metabolizing enzymes in the suppressor mechanism of MSCs (see below), the cell-cell contact might not be required to directly induce apoptosis by a membrane signal but instead to modulate l-arginine metabolism in MSCs.

MSC-dependent inhibition leads to a widespread immune deficiency and death by opportunistic infections only in the final stages of tumor progression. Since immune status deteriorates progressively in tumor-bearing mice, the choice of the immunologic assay can explain the discrepancies found in the degree of functional impairment in different mouse tumor models, which may occasionally lead to the conclusion that late-stage tumors are not associated with weakened systemic T-cell responses. MSCs from mice bearing transplantable tumors inhibited an MHC class I–restricted CD8⁺-specific response (i.e., IFN-γ release), but not ConA-induced proliferation [26]. In our experience, while the anamnestic response to recall antigens is lost in mice bearing large nodules of different transplantable mouse tumors, alloreactivty can be conserved or even enhanced in individual tumor bearers (V. Bronte and E. Apolloni, unpublished results). Strong antigenic stimulation, like that obtained with alloantigen or polyclonal activators such as ConA, might therefore succeed in eliciting lymphocyte activation in mice with a large tumor burden, and in doing so mask the existence of an underlying immune system dysfunction. Interestingly, tumor-bearing BALB-NeuT mice show an impairment of the alloresponse and proliferative response to anti-CD3 and IL-2, concomitant with the increase in circulating MSCs and tumor multiplicity [52]. In these mice the proliferative response to ConA is conserved and they have a normal reaction to haptens in contact hypersensitivity assays. These results can explain why successful response to immunotherapy is mainly observed in young BALB-NeuT animals without tumors and raises the possibility that delayed-type hypersensitivity and cutaneous recall responses are not appropriate indicators of patients’ immune responsiveness or eligibility for immunotherapy vaccination.

Human hematopoiesis disorders similar to those described in mouse studies have been reported in patients with head and neck cancer (HNC). HNCs release GM-CSF and are infiltrated by cells expressing the CD34 antigen, a marker of human totipotent hemopoietic precursors. CD34⁺ cells are also increased in the peripheral blood of HNC patients, and were shown to play a role in the generalized suppression of anamnestic responses to recall antigens that accompanies HNC natural history [61]. Elevated production of GM-CSF by HNC and infiltration with CD34⁺ cells were associated with a poor prognosis (increased rate of recurrence and metastases) [87]. This is similar to what was described for mouse MSCs; moreover, treatment of CD34⁺ suppressors with GM-CSF + IL-4 induced the maturation of these immature cells into fully mature DCs and reversed their immunosuppressive properties [27]. Human MSCs were recently characterized in the peripheral blood of patients with squamous cell carcinoma, HNC, breast cancer, and non–small cell lung cancer. The suppressor cells contained an immature population with the phenotype CD34⁺, CD33⁺, CD15⁻, CD13⁺. According to the expression of HLA-DR and CD11c molecules two main subgroups were identified: one third of the cells were immature monocyte/DCs, and the remaining cells encompassed earlier myeloid differentiation stages. Human suppressor cells inhibited antigen- and mitogen-induced T-lymphocyte proliferation, and once again the combination of GM-CSF and IL-4 drove their differentiation to mature DCs [5].

MSCs and T_h1/T_h2 paradigm

According to the prevalent balance in the type I or type II cytokines produced by T_h lymphocytes, different subsets of myeloid cells can develop: classically activated (MC1) or alternatively activated (MC2). It is likely that all the accessory cells of myeloid origin (macrophages, DCs, and PMNs) are subjected to this dichotomy. MC1 and MC2 subsets are regulated by complex, often antagonistic pathways and are thought to intervene in diametrically opposite functional responses: MC1 are the final targets and effectors of proinflammatory processes, whereas MC2 participate in anti-inflammatory reactions (such as tolerance induction and wound healing) and express a distinct set of molecules and receptors of the innate immunity. The biology of M2 macrophages alternatively activated by IL-4/IL-13, as opposed to classic IFN-γ-dependent activation of M1 macrophages, was recently reviewed [30], and it was suggested that many of the TAMs are polarized M2 macrophages [48]. TAMs produce chemokines such as CCL2 that can orient immunity in a T_h2 direction although the exact mechanism for this action has not yet been disclosed [31]. Thus the effect of microenvironmental cytokines, in lymphoid organs and at tumor site, on myeloid cells can play a crucial role in CTL suppression and tumor progression. T_h1 cytokines are thought to be necessary for induction and the maintenance of anti-tumor CTL responses in vivo, whereas T_h2 reponses may subvert T_h1-mediated immunity, providing a microenviroment leading to disease progression [57]. Studies using a B16 melanoma model have documented a gradual shift of initial T_h0 (mixed T_h1/T_h2 reponse) to T_h2-dominated responses by 14–20 days of progressive tumor growth [35]. Moreover injection of neutralizing anti-IL-4, anti-IL-10, or anti-TGF-β1 antibodies can prevent this tumor-induced functional transition in TH₂ prevalence, resulting in enhanced CTL generation and protection against tumor growth [70]. More recently, in patients with renal cell carcinoma or melanoma, the existence of a correlation between a T_h2 reponse and disease progression was shown [76].

Nonetheless, the paradigm of a strict separation of myeloid subpopulations in one activating (MC1) and one suppressing (MC2) the immune response is challenged by several experimental observations. Both APC subsets were shown to affect negatively adaptive immune responses during the malignancy evolution. Indeed, classic inflammatory macrophages (M1) can exert suppressive activities by the secretion of NO (see below), and superoxide (not generated by M2 macrophages) is released by TAMs [37]; on the other hand, many tumors secrete or induce IL-10 that conditions DC maturation to the DC2 phenotype able to induce immune tolerance.

Are the MSCs that are found in secondary lymphoid organs polarized? Even if tumor-released factors affecting myeloid differentiation are the primary force driving MSC accumulation, there are substantial differences in how MSCs restrain the immune response in various mice strains, which likely reflects the propensity that different genotypes have toward type I or type II cytokine polarization. For example, T lymphocytes from C57BL/6 mice have a strong T_h1 inclination, whereas DBA and BALB/c mouse lymphocytes have a mixed T_h1/T_h2 tendency [56]. As we recently showed [19, 20], two enzymes in l-arginine metabolism, nitric oxide syntase 2 (NOS2) and ARG1, antithetically regulated by T_h1- and T_h2-type cytokines, respectively, are involved in the MSC suppressive pathways. T_h2-type cytokines induce ARG1 and repress NOS2, whereas the opposite is true during T_h1-oriented responses. In the case of a mixed T_h1-T_h2 environment, both enzymes are induced at the same time. Several reports indicate that MSCs can control the CTL activation by a mechanism dependent on NO production [7, 36, 50]. T-lymphocyte suppression by NO is not associated with a loss in IL-2 production or inability to up-modulate the early activation markers, CD69 and CD25, but results in direct impairment of the three signaling pathways of the IL-2 receptor, as demonstrated by the lack of Stat5, extracellular signal-regulated kinase (Erk), and Akt phosphorylation in response to IL-2. NO is secreted by MSCs in response to signals from activated T cells, including IFN-γ and a still unknown contact-dependent stimulus [50]. NO-dependent inhibition of T-cell proliferation is reversible during the first 24–48 hr of stimulation [50], suggesting that this inhibitory pathway prevents activated T cells from entering the cell cycle without killing them. This immunosuppressive mechanism appears to be prevalent in T_h1 predisposed strains such C57BL/6. In a mixed T_h1/T_h2 milieu, as in the case of BALB/c mice, MSCs use another mechanism to restrain the immune response in tumor-bearing mice. In this case both ARG1 and NOS2 are activated and required for immunosuppression. ARG1 increases superoxide production in MSCs through a pathway that likely utilizes the reductase domain of inducible NOS2, and superoxide is required for ARG1-dependent suppression of T-cell functions [19]. In the presence of superoxide, NO gives rise to peroxynitrites that drive apoptosis of antigen-primedT lymphocytes by inhibiting protein tyrosine phosphorylation via nitration of tyrosine residues [14]. This proapoptotic effect might explain why the ARG1- and NO-dependent pathway suppresses at an earlier stage of activation and is irreversible, in contrast to NO-dependent suppression [50]. In BALB/c mice deficient in IL-4 production (IL-4 knockouts) ARG1 is not activated and specific NOS2 inhibitors are able to recover the compromised CTL generation in MSC presence, indicating that the two enzymes can also work independently [19].

Under the influence of a prevalent T_h2 response, ARG1 might be the sole enzyme induced and cause l-arginine deprivation in the extracellular environment. This amino acid is critically required for proliferating T lymphocytes, especially in the case of CD8⁺ T cells [59]. l-arginine availability is necessary for the expression of the T-cell antigen receptor CD3 ζ-chain, the main signal-transduction element of the T-cell receptor (TCR) that is also required for the correct assembly of the receptor complex [67]. Therefore, the regulation of l-arginine concentration in the microenvironment could interfere with signaling circuits (similar to what is seen with the block in IL-2R signaling cascade by NO) and modulate T-cell function at an early stage of lymphocyte activation. Loss of ζ-chain in circulating peripheral blood lymphocytes has been reported in patients under conditions that might all lead to ARG1 activation [34, 67]. Also B lymphocytes are severely affected by chronic l-arginine deficiency. Transgenic mice that overexpress ARG1 in their enterocytes suffer from a 30–35% reduction in circulating l-arginine concentration, retardation of hair and muscle growth, and irregular development of the lymphoid tissues [24]. Whereas cellularity of T cells is not affected, B cells undergo a maturation block in the transition from the pro- to pre-B cell stage in the bone marrow. As a result, the number of B lymphocytes is reduced in peripheral lymphoid organs and small intestine, serum IgM levels decrease, and the architecture of lymphoid organs, especially Peyer’s patches, is profoundly compromised. The reason for this selective effect on B-lymphocyte maturation is not known and models in which ARG1 expression is targeted to other immune sites are required to understand the effects of a more localized l-arginine consumption. In the absence of such a model, one main question that still remains is whether l-arginine depletion, so profound to affect T- and B-lymphocyte functions, might occur in vivo. Decreased levels of the amino acid have been detected in wounds, in liver transplanted individuals, and in patients with acute bacterial peritonitis ([3, 75] and discussed in ref. [67]).

MSCs, and NK and NKT cells

Natural killer (NK) cells are important connectors between innate and adaptive immunity. Since NK cells are the major source of the T_h1-type cytokine IFN-γ, their intervention during immune response is thought to promote a T_h1 polarization essential to maximize CTL responses and activate macrophages [39]. However, depending on the prevailing environment, NK cells can produce a wide panel of both T_h1- and T_h2-type cytokines (TNF-α, GM-CSF, MIP1α, IL-1, IL-8, IL-3, IL-5, IL-10, IL-13, and TGF-β) [38]. Given the potential effects of NK cells on APC differentiation, either via soluble factors or by direct cellular interactions, these cells might be further candidates to regulate MSC activity. In a BW-Sp3 T-cell lymphoma model, depletion of NK cells resulted in tumor progression in all the mice and favored the production of T_h2 type cytokines in splenocyte cultures, concomitant with an outgrowth of CD11b⁺ adherent cells expressing ARG1 (consistent with a MC2 phenotype) [28]. These MC2 cells suppressed the restimulation of memory CTLs and were more susceptible to lysis by lymphokine-activated NK cells than MC1. It was therefore suggested that NK cells that activate MC1, kill MC2; thus, in their absence, MC2 dominate over MC1 and subvert anti-tumor response. Again, these conclusions must take into account the fact that the AKR strain is a strongly type 2–oriented strain.

Another special lymphocyte population, present in both humans and rodents, that has recently received great attention from many immunologists is the paradoxical natural killer T (NKT) cell. These unusual T lymphocytes, which coexpress some NK cell markers, have the capacity to recognize glycolipid antigens in the context of the MHC class I–like molecule, CD1d, via their invariant TCR (V14J281 in the mouse and V24JQ in human). In response to TCR ligation, NKT cells promptly produce large amounts of both proinflammatory T_h1 cytokines, and anti-inflammatory T_h2 cytokines [72]. Using a mouse model in which tumors show a growth-regression-recurrence pattern [77], down-regulation of cytotoxic T-lymphocyte-mediated tumor immunosurveillance was dependent on IL-4R signaling and activation of its downstream transcriptional factor Stat6. Of the two cytokines that signal through the IL-4R–Stat6 pathway, IL-4 was neither sufficient nor necessary for tumor recurrence, whereas IL-13 was mandatory. This was consistent with the observation that T cells and CD4⁺ T cells from mice resistant to recurrence produced less IL-13 in vitro after anti-CD3 stimulation than the cells from susceptible mice. Loss of NKT cells in CD1 knockout mice resulted in decreased IL-13 production and resistance to tumor recurrence. Since T cells do not express receptors for IL-13, involvement of an intermediate cell affecting anti-tumor CD8⁺ T lymphocytes was speculated [18]. More recent data from this group indicate that NKT cell down-regulation of CTL tumor immunosurveillance is mediated by IL-13 induction of a CD11b⁺Gr1⁺ myeloid cell producing TGF-β (J. Berzofsky, personal communication). This data also demonstrates the relevance that MSCs can have in tumor recurrence.

One report by Leite-de-Moraes [45] shows further evidence of a strict link between NKT cells and MSCs. In vivo activation of NKT via injection of their ligand, the α-galactosylceramide, resulted in the production of a large amount of GM-CSF that reached, in the serum, a concentration of 700–800 pg/ml [45], similar to the levels detected in mice bearing GM-CSF-secreting tumors [16]. Growth factor production was accompanied by significant change in the myeloid compartment mobilizing the CD11b⁺Gr1⁺ cells from bone marrow to the spleen and in the blood. Moreover, ligand-activated NKT cells promptly produce IL-4 and IFN-γ [45], suggesting that NKT cells might recruit and also condition MSC activity, a hypothesis that is worth testing.

Conclusions

We have reviewed recent literature with the aim to supply enough evidence for the role that MSCs have in tumor-dependent progressive impairment of T-cell functions. For cancer patients, however, the past history of their tumors is not as relevant as the finding that these cells might be present when we attempt to immunize them with cancer vaccines. It is clear that such a powerful suppressive pathway might render any immunotherapeutic approach questionable. Moreover, it is possible that adjuvants used to enhance a patient’s immune response to weakly immunogenic tumor antigens might even result in activation of MSC-dependent suppression. Blockade of negative regulatory pathways might be necessary to potentiate the effect of cancer vaccines, a sort of push-pull approach against cancer [11]. Among the strategies used to interfere with MSC-dependent suppression, however, a few appear to be realistically transferable to the clinic. Depleting antibodies have been used successfully to eliminate the suppressor cells in a mouse tumor model [71], but they are not selective since they also eliminate mature granulocytes and expose mice chronically treated with the mAb, to opportunistic microbial infections. Selective leukapheresis might be more appropriate. Vitamins can have a relieving effect on tumor-induced immune dysfunctions [26, 82, 84, 85] by driving MSC differentiation to mature antigen-presenting cells, as shown for all-trans-retinoic acid, a natural metabolite of vitamin A [5], or by interfering with the GM-CSF production by tumor cells, as shown for 1α,25-dihydroxyvitamin D3 [84]. On the other hand, cytokines that have been useful to induce MSC differentiation in vitro, might have a very limited usefulness in vivo for the pleiotropic activity and the secondary effects that they might exert in tumor-bearing patients. An alternative strategy is represented by the use of antibodies blocking the cytokines of either recruiting or activating MSCs, as in the case of the administration of anti-VEGF mAb or IL-13-blocking molecules [1, 4]. However, given the plethora of cytokines with similar, redundant activity on MSCs, this approach might require the combination of several anticytokine and possibly antichemokine antibodies. Probably the most promising strategy is represented by the interference with the mechanisms used by MSCs to suppress T-lymphocyte functions. Novel compounds are currently being developed as isozyme-specific inhibitors for either NOS or arginases [23, 80]. Drugs affecting ARG1 and iNOS activity, the enzymes mediating suppression in MSCs, might represent a novel class of immune modulators that would act by limiting the effects of MSC activity in vivo.