Peripheral Blood Progenitor Cell Product Contains Th1-Biased Non-Invariant CD1d-Reactive NKT Cells

Abstract

OBJECTIVE

Bone marrow (BM) Th1 populations can contribute to graft-versus-leukemia (GvL) responses. G/GM-CSF-mobilized peripheral blood progenitor cells (PBPC) have become widely accepted alternatives to BM transplantation (BMT). T cells co-expressing NK proteins (NKT) include a CD1d-reactive subset which influence immunity by rapidly producing large amounts of Th1 and/or Th2 cytokines dependent upon microenvironment and disease. There are two types of CD1d-reactive NKT. “iNKT” express a semi-invariant TCR-α. Other “non-invariant” CD1d-reactive NKT from BM and liver produce large amounts of IL-4 or IFN-γ respectively, and within the intestine can be biased in either direction. Recent data suggests that NKT might contribute to clinical benefits of PBPC.

METHODS

To address these issues, we phenotypically and functionally studied PBPC NKT.

RESULTS

Similarly to BM, NKT-like cells were common in allogeneic and autologous PBPC, there were relatively few classical iNKT, but high CD1d-reactivity concentrated in NKT fractions. Significantly, PBPC CD1d-reactive cells were relatively Th1-biased and their presence was associated with better prognosis. G-CSF treatment of BM to yield PBPC in vivo as well as in vitro Th2-polarizes conventional T cells and iNKT. However, G-CSF treatment of BM in vitro produced Th1-biased NKT, providing a mechanism for opposite polarization of NKT from BM versus PBPC.

CONCLUSIONS

These results suggest distinct Th1 CD1d-reactive NKT cells could stimulate anti-tumor responses from those previously described, which can suppress GvHD.

Introduction

Many approaches are being explored to maximize the potential of immune-mediated control of cancer (1,2). The use of G/GM-CSF-mobilized peripheral blood progenitor cells (PBPC) has become a standard alternative to direct allogeneic or autologous bone marrow (BM) transplantation for hematological as well as some solid malignancies with both clinical and practical advantages. However, acute or chronic graft-versus-host diseases (GvHD) limit the effectiveness of different forms of allogeneic BMT (1,2). BM and PBPC contain heterogeneous populations of CD4⁺ and CD4/CD8-negative T cells that include cells expressing NK cell markers (“NKT”) such as CD161 (NKR-P1) (3), which have anti-tumor activity and can suppress acute GvHD in models (3-7) as well as human mixed lymphocyte responses (MLR) in vitro (7-11). Host or donor BM-derived NKT populations can prevent acute GvHD without loss of GvL effects in mice and man (3-7,12).

A distinct subset of NKT-like cells from BM and other sites recognize the non-polymorphic MHC-related CD1d molecule (10,11,13-18). Unlike in the Lab. mouse, where only ∼ 1% of T cells express NK markers such as NK1.1 (murine CD161), the great majority of the ∼25% of adult human blood T cells which are CD161⁺ (3) are not CD1d-restricted, CD161 being a “memory / activation” marker (13-18). Only rare (∼0.03% PBMC) “classical” human CD1d-reactive invariant NKT (iNKT) are analogues of rodent iNKT, also respond to the glycosphingolipid antigen α-galactosylceramide (α-Gal), possess a highly homologous invariant TCRα rearrangement (Vα24Jα18), and preferentially (but not exclusively) utilize Vβ11 (13,14). CD161 is a major co-stimulatory molecule for human iNKT (19). However, CD161 expression is not essential for recognition by some CD1d-reactive T cells (14,16,18).

Distinct major populations of non-invariant CD1d-reactive T cells, which do not react with α-Gal and utilize diverse TCR, are found in rodent and human BM and human liver (10,11,14-16). BM CD1d-reactive NKT have a marked anti-inflammatory “Th2” bias, producing large amounts of IL-4 (10,11), a B cell differentiating cytokine that can have a protective effect in GvHD and suppress mixed lymphocyte reaction (MLR) (7). Quantitative and qualitative defects in iNKT are predictive of progression in certain autoimmune diseases, and CD1d-reactive NKT can mediate specific types of tolerance, implicating these T cells in immune regulation (14). Human CD1d-reactive NKT have also been shown to have CD1d-specific cytotoxic and anti-viral activities (14,19,20). iNKT IFN-γ is essential for model IL-12-dependent anti-tumor responses and cancer patient iNKT have reversible defects (14,21).

Data from mouse models and human in vitro studies (4-11;14;21-25) and most recently a clinical trial (12) show that the anti-tumor activity of CD1d-reactive NKT might contribute to improved GvL without increased GvHD. However, results with human specifically invariant-type NKT suggest they can enhance MLR (26). A recent study has described the phenotype of rare invariant-like NKT from 8 of the 9 allogeneic (healthy donor) PBPC samples that were tested (27).

Here we report that PBPC from 5 allogeneic donors and 14 autologous cancer patients contain low numbers of iNKT, comparable to blood. As with BM (10,11), we found high levels of PBPC NKT, within which fraction was a high level of functionally CD1d-reactive T cells, but with a marked Th1 bias within both allogeneic and autologous samples. G-CSF treatment in vivo (PBPC) and in vitro Th2-biases classical T cells and iNKT (27-29). Interestingly, we found that treatment with G-CSF in vitro caused BM non-invariant CD1d-reactive NKT to produce IFN-γ. The presence of a major fraction of Th1-biased CD1d-reactive NKT in PBPC may influence GvL and was associated with positive prognosis of autologous recipients.

Study Design

Autologous transplant PBPC samples from 14 patients with hematological malignancies aged 21 - 58 (4 Hodgkin’s Disease, HD; 4 Multiple Myeloma, MM; 6 non-Hodgkin’s lymphoma, NHL) and 5 allogeneic healthy donors (ages 25 - 55) were harvested daily by leukopheresis following G-CSF mobilization. All patients received cyclophosphamide conditioning prior to G-CSF, except patient #12 (HD) who was treated with ifosfamide, carboplatin, and etoposide. CD34⁺ counts were monitored for transplantation. BM was from cadaver vertebral bodies or from filter sets from BM harvests. Mononuclear cells were obtained using Ficoll density centrifugation of surplus clinical material under informed consent. All studies were approved by the BIDMC Committee on Clinical Investigations.

For FACS analysis, the following mAbs were used: Vα24 and Vβ11 (Beckman-Coulter, San Francisco, CA), anti-invariant NKT cell TCRα mAb 6B11 (21); CD3, CD56, CD94 (Becton-Dickinson, San Jose, CA), CD161 (Beckman-Coulter), and isotype controls (Becton-Dickinson). Analysis was by FACScan (Becton-Dickinson) as described (10,13,17), with >2×10⁵ cells analyzed for rare invariant NKT.

Functional analysis of CD1d-reactivity by PBPC and BM-derived T cells was performed using healthy PBMC-derived iNKT as positive controls and mock (plasmid alone) or human CD1d-transfected human HLA-A, -B negative C1R B cell lines as described (11,13,17). Briefly, quadruplicate 1×10⁵ / well MNC were co-cultured with an equal number of CD1d transfectamts in 96 well plates, RPMI-1640, 10% FBS, 20U/ml IL-2, and 1ng/ml PMA or 100ng/ml α-Gal (Kirin Pharma, Gunma, Japan) as indicated. Subtracting results with CD1d from mock transfectants provided net CD1d-specific responses. Response to CD1d was specifically blocked by 10μg/ml anti-CD1d antibody 59.1 versus isotype control (11,13,17). Supernatants were collected at 1 day and 3 days for peak IL-4 and IFN-γ respectively. Released cytokine levels were determined by ELISA with matched antibody pairs in relation to cytokine standards (Becton-Dickinson; Endogen, Inc. Cambridge, MA). Results shown with standard errors. Survival curves were generated by Kaplan-Meier method and were compared using two-sided log-rank test, with P < 0.05 considered significant.

The phenotype of Th1-biased CD1d-reactive cells within PBPC and BM was investigated by sequential purification of CD56⁺ and CD161⁺ populations by Miltenyi anti-mIgG microbeads according to the manufacturer’s recommendations using mAbs as above. Most NK cells were removed in the initial CD56⁺ fraction along with a small fraction of CD56⁺ NKT-like cells, leaving most NKT cells (CD56^-CD161⁺; 11) in the CD161⁺ fraction. The “flowthrough” fractions contained most classical CD4 and CD8 T cells and B cells as well as myeloid cell subsets.

Results

NKT-like and invariant NKT cells in autologous and allogeneic PBPC

Previous studies have shown that human as well as rodent healthy resting BM contains substantial levels of both phenotyically NKT and functionally CD1d-reactive NKT (4-11). Published data on human PBPC NKT-like subsets is limited to CD4/CD8-double negative T cells (9), whose relationship to CD1d-reactive NKT remains unclear, and to healthy donor Vα24⁺ Vβ11⁺ double positive iNKT-like cells, which make up <0.1% of healthy donor PBPC T cells (27). To determine the status of ‘NKT’ populations in PBPC, we performed FACS analysis. Figure 1 shows representative and summary results of phenotypic comparison of samples from healthy human BM, allogeneic and autologous PBPC samples, and healthy donor non-mobilized peripheral blood mononuclear cells (PBMC). Regarding generically NKT-type cells, adult PBMC T cells contained 25±5% CD3⁺ CD161⁺ (Figure 1A,B) with fewer CD3⁺ CD56⁺ and CD3⁺ CD94⁺ double positive T cells (both ∼10%). BM T cells contained ∼30% CDC161⁺ T cells (Figure 1A,B) with < 5-< 15% CD3⁺ CD56⁺ and CD3⁺ CD94⁺ cells. Proportions of PBPC NKT cells from autologous donors with a variety of hematological malignancies were similar to BM, with relatively low values for allogeneic healthy donor PBPC (Figure 1A,B). CD161 was more common than CD56 or CD94 on PBPC T cells. The frequency of PBPC CD3⁺CD161⁺ ‘NKT’ was 20 - 40% of T cells (Figure 1A,B) and 5 - 15% for CD56 or CD94, as previously found for PBPC (9) and PBMC (3).

Figure 1. Peripheral blood progenitor cell (PBPC) product NKT & invariant NKT cells.

Representative (A) and summary (B,C) FACS analysis of PBPC compared to BM, and PBMC, gated on live lymphocytes with mAb shown. In Fig. 1A, autologous = HD. In Fig. 1B,C, NHL (n = 6), HD (n = 4), MM (n = 4), and 5 healthy allogeneic donor PBPC were analyzed. Mean % NKT and iNKT shown for each group.

B. Summary data of generic ‘NKT’ cells proportions using mAb against CD3 and CD161.

C. iNKT were identified both by double staining with mAb combinations Vα24 and Vβ11 or with Vα24 and anti-invariant TCR (6B11, 21) as shown. * indicates p < 0.05.

The numbers of rarer Vα24⁺ Vβ11⁺ double positive iNKT-like cells and classic truly invariant Vα24⁺ 6B11⁺ double-positive iNKT expressing the invariant rearrangement (13,14) and therefore recognized by mAb 6B11 (21) were measured to determine if PBPC contained comparable levels to BM or PBMC. Conversely to the high numbers of NKT in both BM and PBPC, only a small number of iNKT were detectable in BM (< 0.02%; Figure 1A,C), as previously shown (10). The levels of iNKT (0.01 - ∼0.1%) in allogeneic and most autologous PBPC were comparable to PBMC, lowest levels being for NHL (Figure 1A,C). Thus PBPC contained relatively high levels of generic NKT cells similar to BM and PBMC, with invariant NKT mostly comparable to PBMC, features intermediate between BM and PBMC.

Functionally CD1d-reactive NKT cells in autologous and allogeneic PBPC

Next it was determined whether PBPC ‘NKT’ populations responded to CD1d as measured directly ex vivo. Intact PBPC were monitored for CD1d-reactivity as previously described for blood and BM-derived NKT cells (10,13,15,20,21). The results showed that PBPC contained readily detectable CD1d-reactive T cells producing high levels of IFN-γ (Figure 2A,B,C). However, unlike Th2-biased BM (10), PBPC CD1d reactivity was Th1-like with minimal IL-4 production (Figure 2A,B,C). PBPC CD1d reactivity was specific, since there was little if any response to CD1a, -b, or -c, and primarily non-invariant since α-Gal induced minimal responses (not shown). Furthermore, CD1d antibody specifically and substantially inhibited PBPC CD1d responses (Figure 2A,B).

Figure 2. PBPC CD1d-reactive NKTex vivo.

A,B. CD1d-specific and T cell mitogen reactivity of PBPC from 2 donors was measured ex vivo as described (10,13,15,21). Briefly, mononuclear cells were stimulated with CD1d⁺ or mock transfectants (“CD1d” = CD1d reactivity - response to mock transfectants) or PHA mitogen. CD1d blocking mAb (“@CD1d”) or isotype control were included as shown.

A. Healthy donor. B. autologous donor. * indicates p < 0.05, ** indicates p < 0.01.

C. CD1d-specific versus T cell mitogen reactivity was measured as described (10,13,21) from PBPC ex vivo and compared to that of control PBMC-derived pure iNKT. PBPC were stimulated with CD1d⁺ or mock transfectants (“tot.CD1d” = total CD1d-specific reactivity - response to mock transfectants) and expressed as a fraction of total T cell responses to PHA mitogen (100%). Summary results shown for 9 patients with functional data and long-term clinical follow-up. Disease and follow-up (in months) are listed. Dec’d. = Deceased donor due to relapse at time shown. 5 donors had measurable CD1d responses, 4 / 5 being alive at follow-up shown (weakest responder of the 5 being deceased), whereas 3 / 4 donors without detectable CD1d responses were deceased due to relapse at 4, 19, and 20 months post-autotransplant.

D. Kaplan-Meier Survival plot from donors with data in Fig. 2C. Dotted line indicates limit of follow-up, median > 53 mth. for CD1d-reactive donors (range > 48 - > 61) and 52 mth. for the surviving CD1d non-reactive donor. Data was subjected to 2-sided log-rank test, p=0.02.

Figure 2B shows on a logarithmic scale that multiple whole PBPC preparations responded specifically to CD1d with some range in levels relative to mitogen control. CD1d-specific IL-4 was detectable in only 1 / 10 PBPC, which also had one of the highest levels of CD1d-specific IFN-γ (Figure 2B). CD1d-specific IL-4 (but not IFN-γ) is readily detectable from bone marrow CD1d-reactive NKT (10). In contrast, the level of PBPC CD1d-specific IFN-γ response was > 5% of total PBPC IFN-γ from 5 / 6 positive donors (Figure 2B). 4 / 5 detectably CD1d-reactive autologous donors as well as the healthy allogeneic donor and pure iNKT cell line from another healthy donor produced CD1d-specific IFN-γ comparable to that induced by mitogen (Figure 2B: only eventually deceased #2C < 1% of PHA).

Neither survival nor CD1d response stratified with autologous donor disease (Figure 2B). Deceased donors were all males, aged 23, 51, and 53 at collection. Surviving patients were males aged 21-52 (median = 41) at collection. However, the presence of detectable CD1d-reactive NKT in the patient PBPC was associated with improved prognosis (Figure 2C). With median follow-up of > 53 mth. (mean = 55; range > 48 - > 61) 5 / 6 survivors were positive for CD1d-reactivity. 3 / 4 autologous donors lacking detectable CD1d-reactivity experienced death due to relapse at 4, 19, and 20 mth. post-collection versus 1 / 5 at 59 months for CD1d-reactive donors (Figure 2C, p=0.02). Of 5 surviving patients with CD1d reactivity, 1 was apparently cured by autologous transplant (remission at 48 mth.), 3 had slow progressive disease (1 allotransplant at 11 mth. post-auto-, 2 allo-transplanted at > 48 mth.), and 1 was lost to follow-up at 60 months. Notably, an HD patient with slow progression and treatment-sensitive disease (#12B) re-tested 27 mth. post-autotransplant maintained peripheral blood CD1d-reactive NKT producing robust IFN-γ response (80.4% of mitogen), comparable to original response (Figure 2B), being allotransplanted ∼ 1.5 years later.

PBPC CD1d-reactivity resides primarily within the CD161⁺ “NKT” fraction

We next determined the phenotype of Th1-biased CD1d-reactive cells within PBPC versus those in PBMC. Both were subjected to sequential purification of CD56⁺ and CD161⁺ populations. This ensured isolation of essentially all NK cells in the CD56⁺ fraction with modest numbers of CD56⁺ T cells with the majority of true NKT cells in the CD161⁺ fraction. Depletion was very efficient, for example in the CD56^hi PBPC sample shown, from > 30% NK (CD3^- CD56⁺) and ∼ 8% CD161⁺ cells to ∼ 1% of each (Figure 3A). PBMC contained modest, but detectable IFN-γ and IL-4 CD1d-reactivity (Figure 3B,C). There was little if any reactivity in the mostly NK cell CD56⁺ fraction or the residual “flow-through” material. However, the reactivity was recovered in the CD161⁺ “NKT” cell fraction (Figure 3B,C).

Figure 3. Defining subsets of PBPC CD1d-reactive NKT ex vivo.

PBPC and PBMC were subjected to sequential fractionation into CD56⁺ (primarily NK cells) and CD161⁺ populations (therefore primarily “NKT cells”). These, the starting material, and ‘flow-through’ (CD56-neg. CD161-neg. primarily classical T cells, myeloid cells, and some B cells) were assessed for % purity by FACS analysis (A) and assayed for total CD1d-reactivity as well as total PHA mitogen responses as in Fig. 2. B,C. PBMC; D,E, autologous PBPC. Production of IFN-γ (B,D) and IL-4 (C,E) shown. Nearly all NK and NKT cells were removed in the purifications, flow-through numbers indicating recovery of other populations in this fraction. Note, highest level PBPC and PBMC CD1d responses (primarily IFN-γ, little or no significant IL-4) were detected in the NKT cell fraction, whereas total responses (PHA mitogen) were comparable on a per cell basis in different fractions.

Next, similar results were obtained for PBPC. The largest and remaining “flow-through” PBPC fraction contained mainly classical CD4 and CD8 T cells and B cells as well as myeloid cell subsets. As was expected, there was minimal IFN-γ based CD1d-reactivity in the NK cell CD56⁺ fraction (Figure 3D). Notably, CD161⁺ populations contained the highest levels of IFN-γ CD1d-reactivity of these fractions. Again, however, there was little CD1d-reactive IL-4 (Figure 3E). The flowthrough also contained little if any significant CD1d-reactivity (Figure 3D). Therefore, PBPC CD1d-reactivity was concentrated in CD161⁺ truly “NKT” cells, as previously found for BM (11).

Treatment of bone marrow CD1d-reactive NKT cells with G-CSF in vitro

Murine and human BM CD1d-reactive NKT have a marked Th2 bias (10,11), as do trace healthy donor PBPC iNKT (27), whereas the major PBPC non-classical CD1d-reactive T cell population has a Th1 phenotype (Figure 2A). Furthermore, G-CSF treatment in vivo and in vitro has been reported to Th2 polarize classical blood T cells as well as allogeneic donor PBMC iNKT (27-29). Therefore, BM was treated with G-CSF in vitro to model stem cell mobilization. Remarkably, BM CD1d-reactive NKT treated with G-CSF produced comparable levels of IFN-γ (Figure 4A,B) as the equivalent liver (15,16) and blood (13-16,21) populations, and negligible IL-4 (not shown). Significantly, there were minimal BM responses to iNKT specific ligand α-Gal (Figure 4B, not shown), confirming modest presence of iNKT in BM as well as PBPC, relative to substantial non-invariant CD1d-reactive NKT cells. In conclusion, these results suggest that PBPC CD1d-reactive NKT are derived from BM CD1d-reactive NKT re-polarized to Th1 phenotype by G-CSF treatment.

Figure 4. Responses of BM CD1d reactive T cells to G-CSF in vitro.

CD1d-specific versus T cell mitogen reactivity of BM and control PBMC-derived iNKT ± treatment with G-CSF. Cells were stimulated with CD1d⁺ or mock transfectants (“tot.CD1d” = total net CD1d-specific reactivity), α-Gal (“a-Gal”) iNKT-specific antigen, or by PHA mitogen as in Fig. 2 and as described (10,13,21). Representative results shown.

As with intact BMMC (Fig. 4A,B), BM CD56⁺ and CD161⁺ populations were separated and also assayed for CD1d-reactivity as described in Figure 3. Production of IFN-γ (Fig. 4C,D) and IL-4 (Fig. 4E,F) from starting intact BM ± G-CSF (as in Fig. 4A,B) are shown (Fig. 4C,E) along with CD161⁺ populations± G-CSF (Fig. 4D,F) and 2 different control iNKT lines. There was no significant CD1d-specific production of IFN-γ or IL-4 from CD56⁺ or ‘flow-through’ (CD56-neg. CD161-neg.) populations, but all produced substantial amounts of both cytokines in response to mitogen (not shown), confirming viability.

Finally, to similarly identify the population involved in BM CD1d-reactivity, BM (± G-CSF treatment in vitro as in Fig. 4A,B) was also separated into CD56⁺, CD161⁺, and ‘flow-through’ fractions. As with PBPC, isolation of CD56⁺ and CD161⁺ cells was almost complete (not shown). As shown in Figure. 4C,D, CD1d-reactivity, readily detected in the starting BM (± G-CSF), was concentrated in the CD161⁺ “NKT” cell fraction. Again, G-CSF treatment specifically enhanced CD1d-dependent IFN-γ production (Fig. 4C,D). Conversely, CD1d-specific IL-4 production was selectively decreased (Fig. 4E,F). Low level invariant NKT responses to the ligand α-Gal confirmed the results seen with total CD1d-reactivity in both cases (Fig. 4C - F). BM CD56 NK cells fraction and flow-through fraction (± G-CSF) had no detectible CD1d reactivity, although PHA responses were strong, indicating viability of the cells (not shown). Therefore, BM contains CD1d-reactive NKT cells, which become more Th1-polarized upon G-CSF treatment in vitro, analogous to the Th1-biased CD1d-reactive NKT cells from G-CSF treated patients PBPC ex vivo.

Discussion

Classic immunoregulatory CD1d-reactive CD161⁺ iNKT interact with CD1d⁺ antigen presenting cells and presumably yet to be defined glycolipid antigens. However, distinct CD1d-reactive populations with different functions are now known to dominate other sites in both rodents and humans (3-12;16-19). A previous study including 8 of 9 tested donors reported that iNKT represent <0.1% of allogeneic PBPC and display a Th2 bias (27) like in BM (10). Here we confirm that PBPC product of both allogeneic donors and autologous cancer patient donors contained very low numbers of classical iNKT. However, the major non-invariant human PBPC T cell population responding specifically to CD1d, as obtained following G-CSF mobilization in vivo, were highly Th1-like cells producing large amounts of IFN-γ and little if any IL-4. Like certain other significant aspects of the otherwise highly conserved NKT cell biology, this contrasts with the situation reported in the murine model, where non-invariant NKT were reported to suppress iNKT anti-tumor responses (30). 3 / 4 autograft recipients who were deceased due to relapse lacked detectable CD1d-reactive PBPC at transplant, the 4^th being the weakest responder, whereas 5 / 6 surviving patients had CD1d reactivity (p = 0.02). Although numbers are small, the observed association between CD1d reactivity and likelihood of survival warrants further study and is consistent with the anti-tumor correlation of classic iNKT in patients as well as models (14,21). Long-term follow-up of the current cohort and further donors is underway.

Unlike conventional blood T cells (28,29) as well as at least healthy donor PBPC iNKT (27), where G-CSF apparently induces Th2 bias, treatment with G-CSF in vitro polarized naturally Th2-biased BM CD1d-reactive NKT to PBPC-like Th1 phenotype. These results extend the spectrum of functionally CD1d-reactive NKT with potentially distinct immunoregulatory functions in different physical locations. PBPC CD1d-reactive non-invariant T cells we have described here appear directly related to BM ‘NKT’, many of which are CD4/CD8-negative, but which are Th2 and inhibit MLR in vitro and prevent GvHD in vivo (4-11). Host or donor BM-derived CD1d-reactive NKT cells can prevent acute GvHD without loss of GvL effects in mice and man (4-7;12;22-25). Therefore, BM and PBPC contain large populations of CD1d-reactive NKT phenotypically and functionally distinct from classic iNKT. Recently, iNKT have been shown to potentially contribute to hemopoiesis, particularly in the context of a CMV infection model (31,32). Clinical application of ‘NKT’ in the context of BMT is in clinical trials (12). Th1-like CD1d-reactive non-invariant NKT may contribute to GvL effects, as their presence was associated with positive prognosis in autologous recipients. However, such cells might also contribute to chronic GvHD in the setting of allogeneic transplantation. Our data offer new perspectives on mobilized BM composition that may affect relapse post-transplant.