Cytokine-driven loss of plasmacytoid dendritic cell function in chronic lymphocytic leukemia

Patients

Patients aged over 18 suffering from CLL fulfilling the National Cancer Institute (NCI) Working Group criteria for diagnosis and staging (3), were prospectively recruited from The Alfred Hospital and Peter MacCallum Cancer Centre. Cryosamples of CLL bone marrow aspirates and lymphocytes were obtained from the Australasian Leukaemia and Lymphoma Group (ALLG) Tissue Bank. Age-matched healthy donor (HD) whole blood samples were collected by the Australian Red Cross Blood Service and provided as per non-clinical supply agreement. All samples from CLL patients were divided into two groups: individuals with stable disease not requiring treatment (indolent) and individuals with advanced disease prior to treatment or who would require treatment in the near future (progressive), according to the NCI Working Group criteria (3). All patients herein gave written informed consent prior to inclusion in this study. Patients receiving therapy for CLL were excluded. All studies were approved by the institutional ethics review boards of the Monash University in accordance with the Declaration of Helsinki principles.

Mice

EμTCL1-Tg mice have previously been described (10) and were kept on a B6C3 background. B6C3 mice were used as wild-type (WT) controls and provided by the Alfred Medical Research and Education Precinct (AMREP) Animal Services (AAS). All experimental mice used were age-matched and housed under high barrier protection and handled with approval from institutional Animal Ethics Committees, in compliance with the Australian code of practice for the care and use of animals for scientific purposes.

Lymphocyte isolation

Lymphocytes from HD and CLL patients were purified from peripheral blood diluted 1:1 in PBS (Invitrogen, Carlsbad, CA, USA) and layered onto Ficoll-Paque (GE Healthcare, Piscataway, NJ, USA) for gradient centrifugation. Mouse splenocytes were isolated following disaggregation of the spleen and 70μm filtration (BD, San Diego, CA, USA) to form a single cell suspension. Red blood cells were lysed with RBC lysis buffer (eBioscience, San Diego, CA, USA). Cells were enumerated using a Z™ Series Coulter Counter (Beckman Coulter Inc., NSW, Australia). Absolute numbers of pDCs were calculated by multiplying the percentage of FACS-gated pDCs by total cell count.

Antibodies, flow cytometry and analysis

For flow cytometric analysis, single-cell suspensions were stained as previously described (22). Intracellular staining was conducted using a Fixation/Permeabilization kit from eBioscience (San Diego, CA, USA) as per manufacturers instructions. Fragment crystallizable (Fc) receptors were blocked using purified CD16/32 monoclonal antibody (mAb) (BD, San Diego, CA, USA). Viable lymphocytes were assessed using Live/Dead Fixable Dead Cell Stain Kit (Invitrogen, Carlsbad, CA, USA). Matching isotype mAbs were used to control for background staining.

Anti-mouse CD5-eFluor450 (53-7.3), B220-FITC (RA3-6B2), PDCA1-APC (eBio927), Siglec H-eFluor710 (eBio440c), CD11b-PE (M1/70) and TLR9-FITC (M9.D6) were purchased from eBioscience (San Diego, CA, USA). Gr.1 (Ly-6G)-pacific blue (RB6-8C5) was purchased from Caltag (Buckingham, UK). CD19-PerCP (6D5) was sourced from Biolegend (San Diego, CA, USA). CD11c-PE-Cy7 (HL3) and Flt3-PE (A2F10.1) were purchased from BD (San Diego, CA, USA). Anti-mouse TGF-βR1-PE (Cl141231) was purchased from R&D Systems (Minneapolis, MN, USA).

Human cells were stained with a FITC-conjugated lineage negative (Lin⁻) cocktail containing antibodies to CD3, CD14, CD16, CD19, CD20, CD56 (NCAM16.2, MφP9, L27, SJ25C1, 3G8, SK7), CD11b-FITC (ICRF44), and with CD123-PE-Cy7 (7G3), CD20-APC-Cy7 (L27), CD5-PE-Cy7 (L17F12) and Flt3-PE (4G8), all purchased from BD (San Diego, CA, USA). CD11c-FITC (3.9) and HLA-DR-eFluor450 (LN3) were purchased from eBioscience (San Diego, CA, USA). BDCA2-APC was sourced from Miltenyi Biotec (Bergisch Gladbach, Germany). BDCA4-PerCP (446921) was purchased from R&D Systems (Minneapolis, MN, USA). TLR9-PE (26C593.2) was purchased from Enzo Life Sciences (Farmingdale, NY, USA). A BD LSRII flow cytometer was used to analyze the flow cytometry samples (San Jose, CA, USA). At least 100,000 events were collected for all analyses. The analysis was conducted using FlowJo software (Treestar, Ashland, OR, USA).

Cell cultures

RPMI and DMEM media (Invitrogen, Carlsbad, CA, USA) were used for all mouse and human cultures, respectively. Media were supplemented with 10% FCS, L-glutamine, 2ME, HEPES and Pen/Strep; all from Invitrogen (Carlsbad, CA, USA). For bone marrow cultures, cells were isolated from the femurs of EμTCL1-Tg and age-matched WT controls. Human bone marrow cryosamples were thawed and washed in sterile PBS. Bone marrow cells were seeded at 10⁵ cells/ml in 48-well cell culture plates (BD, San Diego, CA, USA). For lymphocyte cultures, cells were seeded at 10⁶-10⁷ cells/ml in 24- or 48-well cell culture plates (BD, San Diego, CA, USA). All cultures were incubated at 37°C and 5% CO₂. Stimulatory factors were used at the following concentrations: Flt3L, 100-400ng/ml (R&D Systems, Minneapolis, MN, USA); CpG B ODN 7909, 1μM and CpG C ODN 2395, 1μM (Invivogen, San Diego, CA, USA). Optimal CpG stimulations were conducted for 6-12 hrs. In vitro neutralization of TNF and transforming growth factor beta (TGF-β) was achieved using purified anti-mouse TNF (MP6-XT22), anti-mouse TGF-β (TW7-20B9), anti-human TNF (mAb11) and anti-human/mouse TGF-β (19D8) (Biolegend, San Diego, CA, USA), all used at 1μg/ml.

Injections

Mice received one intravenous injection per week for two weeks of either 100μg anti-TNF (MP6-XT22), 100μg anti-TGFβ (TW7-20B9) or 100μg isotype controls (all sourced from Biolegend, San Diego, CA, USA). Sterile PBS (Invitrogen, Carlsbad, CA, USA) was for injection controls.

Serum and secreted cytokine quantification

Serum and secreted IFNα levels were quantified using high-sensitivity ELISA assays from PBL Interferon Source (Piscataway, NJ, USA), as per manufacturer’s protocol. Levels of TNF, TGF-β and Flt3L in serum and supernatant were quantified using ELISA assays purchased from R&D Systems (Minneapolis, MN, USA), as per manufacturer’s protocol. All samples were tested in duplicates.

Statistical analysis

Data are shown as means (± SEM). The student’s t-test was used to determine significant differences between means. All statistical analyses were performed with GraphPad software (Prism Version 6.0d, 2013, San Diego, CA, USA). A P value ≤ 0.05 was considered statistically significant. *P < 0.05; **P < 0.01; ***P < 0.001.

Reduced numbers of splenic pDCs in EμTCL1-Tg mice as disease progresses

Given the higher incidence of chronic viral infection in most CLL patients either at diagnosis or thereafter (11), we first used the EμTCL1-Tg mouse CLL model to examine possible underlying early immune abnormalities, focusing on pDCs, which are central initiators of effective anti-viral immune responses (reviewed in (20)). Interestingly, proportions of splenic pDCs were significantly reduced in EμTCL1-Tg as they age (Figure 1A, top and middle panels). Significant reduction in pDC numbers was also observed in the peripheral blood of EμTCL1-Tg mice (Supplementary Figure 1). No significant difference was observed in the proportion of pDCs in the bone marrow (BM) of EμTCL1-Tg mice and WT mice (Figure 1A, bottom panel) in keeping with previous work showing that pDC response to growth factors is dependent on the microenvironment and distinct in the BM and the periphery (23). Loss of splenic pDCs paralleled the accumulation of the CD5⁺CD19⁺ B cell clone in EμTCL1-Tg mice, which was apparent at 1-mo but striking at 12-mo of age (Figure 1B). Proportions of CD5⁺CD19⁺ B cells in WT control mice remained relatively constant over time (Figure 1B) and corresponded to B1 B cells normally present in the spleen (reviewed in (24)). In addition to the proportions (Figure 1C), absolute numbers of splenic pDCs were reduced in EμTCL1-Tg mice (Figure 1D). Numbers of pDCs and tumor cells inversely correlated (Figure 1E). These data suggest that EμTCL1-Tg mice progressively lose peripheral pDCs as disease progresses.

Impaired pDC numbers in the peripheral blood of CLL patients as disease progresses

We next investigated whether the pDC defect seen in EμTCL1-Tg (Figure 1) could also be detected in blood samples from CLL patients. Indeed, the aberrant pDC phenotype seen in EμTCL1-Tg mice was also seen in CLL patients (Figure 2A). Proportions of pDCs were significantly decreased (5-fold) in the blood of patients with CLL (data not shown). In contrast to the striking absence of pDCs in the peripheral blood, pDCs were detected in the matched BM of patients with progressive CLL (Figure 2A), similar to normal BM sample (Supplementary Figure 2), again confirming different responses of pDCs in distinct microenvironments as described previously(23). pDCs numbers were significantly diminished in patients with progressive CLL but, surprisingly, significantly increased pDC numbers in the blood of patients with indolent CLL relative to healthy donors (HD) (Figure 2B). Overall, pDC numbers in the peripheral blood were found to inversely correlate with the number of CD5⁺CD19⁺ B cells (Figure 2C).

Age-related decline in pDC numbers has been reported previously as a feature of immune senescence in healthy humans over 60 years of age but not to the extent of what we observed in CLL patients compared to age-matched HD (25). Collectively these observations suggest that similar to the EμTCL1-Tg mouse CLL model, progression of CLL is associated with gradual loss of pDCs, which may explain in part the poor anti-viral responses of CLL patients.

Defective pDC function in both EμTCL1-Tg mice and CLL patients

A well-known function of pDCs is to mediate antiviral activity via IFN-I secretion and pDC-derived IFNα accounts for physiological levels of this factor in the serum of healthy individuals (16). Loss of pDCs suggested a possible deficiency in IFNα production in both EμTCL1-Tg mice and CLL patients. Indeed, as they aged, EμTCL1-Tg mice displayed significantly reduced serum IFNα levels relative to WT controls (Figure 3A). Similarly, serum IFNα levels were significantly reduced in patients with indolent CLL relative to HD and further diminished in patients with progressive disease (Figure 3B). Flow cytometry staining for intracellular indoleamine 2,3 dioxygenase (IDO), a well-known immunosuppressive factor abnormally produced by some pDCs in solid tumors (26), was negative in pDCs from EμTCL1-Tg mice and patients with indolent or progressive disease following stimulation (data not shown). This result suggested that pDCs in CLL have not been further subverted beyond an impairment of IFNα production. Moreover, a deficit in pDC-derived IFNα production may also partly explain the vulnerability of CLL patients to infections.

The surprising increase in pDC numbers in patients with indolent CLL despite reduced IFNα levels in the serum of these patients suggested defective IFNα production by their pDCs. pDCs have a restricted TLR profile, primarily expressing TLR7 and TLR9 (16). Ligation of TLR9 by CpG oligonucleotide sequences induces rapid activation and subsequent IFNα production (16). Interestingly, TLR9 expression was significantly reduced in splenic pDCs from 12-mo old EμTCL1-Tg mice with advanced disease (Figure 3C). Likewise, TLR9 expression was significantly reduced in pDCs of patients with indolent CLL and almost completely lost in patients with progressive disease (Figure 3D). As expected, in vitro CpG-stimulated pDCs from WT control mice produced increased levels of IFNα (Figure 3E). In contrast, pDCs from EμTCL1-Tg mice with advanced disease had impaired IFNα production in response to CpG stimulation (Figure 3E). Similarly, blood pDCs from patients with indolent disease had impaired IFNα production in response to CpG stimulation in vitro and negligible levels of IFNα were detected in pDC cultures from patients with progressive disease (Figure 3F). Therefore, the pDC defect in CLL is multifaceted and characterized by both reduced numbers as disease progresses and an inability to produce IFNα upon TLR9 activation. Moreover, increased pDC numbers in the blood of patients with indolent CLL (Figure 2B) may partially offset the reduced production of IFNα by these cells (Figure 3F), hence contributing to a relatively asymptomatic immune state in these patients who do not require therapy.

Elevated serum Flt3L concentration but reduced surface Flt3 expression on pDCs of EμTCL1-Tg mice and CLL patients

Flt3L is a well-known component of pDC differentiation and is essential for their growth and maintenance.(27) In vivo experiments in both mice and humans have demonstrated a significant increase in pDC numbers following administration of Flt3L (27, 28). Surprisingly, serum Flt3L levels were significantly elevated in 12-mo old EμTCL1-Tg mice with advanced disease relative to younger EμTCL1-Tg mice (Figure 4A). Similarly, patients with either indolent or progressive CLL displayed significantly increased serum Flt3L levels compared to HD (Figure 4B). Furthermore, there was no difference in the ability of the BM from either EμTCL1-Tg mice or patients with progressive CLL to generate pDC upon Flt3L treatment ex vivo (Figure 4C and D, respectively). These data suggested that mechanisms underpinning pDC deficiency were confined to the periphery and that increased serum concentrations of Flt3L seen in CLL may be an important compensatory mechanism maintaining high pDC numbers in patient with indolent CLL to offset their reduced IFNα production.

The disparity between decreased pDC numbers and elevation of peripheral Flt3L levels suggested a dampened response to growth signals mediated by the receptor Flt3. EμTCL1-Tg mice with advanced disease (9-mo old) displayed diminished Flt3 expression on splenic pDCs compared to WT mice (Figure 4E). Flt3 expression was significantly higher on BM pDCs from EμTCL1-Tg compared to splenic pDCs from the same mice (Figure 4E). This higher Flt3 expression may explain the maintenance of BM pDC numbers in EμTCL1-Tg mice in contrast to Flt3^lo splenic counterparts.

Likewise, pDCs from CLL patients with indolent disease expressed significantly increased levels of Flt3 compared to HD (Figure 4F). Conversely, significant downregulation of Flt3 expression was observed on pDCs from patients with progressive CLL (Figure 4F). Higher Flt3 expression on pDCs from patients with indolent CLL may account for the increased numbers of circulating pDCs in these individuals (Figure 2B). These observations suggest that loss of pDCs may be linked to gradual loss of Flt3 expression as disease progresses and subsequent impaired maintenance of these cells.

Increased pDC numbers and IFNα secretion following combined Flt3L and CpG stimulation of pDCs from EμTCL1-Tg mice or CLL patients

Exogenous Flt3L increased pDC numbers in 5-d cultures of splenocytes from WT mice (Figure 5A). In contrast, pDCs from EμTCL1-Tg mice with advanced disease failed to expand in response to Flt3L in culture (Figure 5A). A similar pattern was observed in human PBMC cultures. Modest increases in pDC numbers were observed in 4-d Flt3L cultures of PBMCs from patients with indolent CLL, in contrast to Flt3-supplemented cultures of PBMCs from HD in which significantly higher numbers of pDCs were generated (Figure 5B). Strikingly, Flt3-supplemented cultures with PBMCs from patients with progressive CLL failed to increase pDC numbers (Figure 5B), mirroring the effect seen in EμTCL1-Tg splenocyte cultures. This may reflect the lower sensitivity of these pDCs to Flt3L due to reduced Flt3 expression (Figure 4E and F). Furthermore, ex-vivo amplification of pDC numbers from WT mice or HD with exogenous Flt3L alone was enough to modestly elevate IFNα levels (Figure 5C and D). Interestingly, pDCs from patients with indolent CLL also responded in vitro to Flt3L stimulation by producing IFNα in similar amounts to that of control cultures. Importantly, cultures supplemented with Flt3L and CpG resulted in visibly increased IFNα secretion by pDCs from EμTCL1-Tg mice (Figure 5C) and patients with indolent and progressive CLL (Figure 5D). Of note, detected IFNα levels were lower in unstimulated cell culture conditions than in serum (Figure 3B). This is expected in prolonged in vitro culture without exogenous stimuli, and IFNα levels mirrored those of serum when stimulation was provided. We observed no effect of CpG on Flt3 expression on pDCs (data not shown). These observations suggest that combining these immunostimulatory agents could restore some pDC activity (Figure 5C and D).

TNF and TGF-β-dependent downregulation of Flt3 expression in pDCs from CLL patients and EμTCL1-Tg mice

TNF and TGF-β are well-established negative regulators of pDCs via modulation of Flt3 expression during haematopoiesis (29, 30). Interestingly, TLR9 downregulation in pDCs has also been linked to increased TNF production.(31) TNF is highly produced by activated macrophages, and to a lesser extent, activated T cells, B cells and NK cells (reviewed in (32)). TNF has also been described as an autocrine growth factor produced by malignant CLL B cells and is a significant contributor to tumor persistence (33). We detected elevated serum TNF levels in EμTCL1-Tg mice with advanced disease relative to younger EμTCL1-Tg mice, and age-matched WT controls which did not show detectable levels of TNF (Figure 6A). Likewise, we observed a significant increase in TNF levels in the serum of patients with progressive CLL compared to that of both HD and patients with indolent disease (Figure 6B). Furthermore, in vitro TNF stimulation of pDCs from EμTCL1-Tg mice and CLL patients resulted in significantly reduced Flt3 and TLR9 mRNA levels, respectively (Figure 6C and D). This suggested that TNF modulated Flt3 and TLR9 expression at the transcriptional level. Strikingly, multiple IFN-responsive genes and Flt3-inducing genes were also significantly downregulated in TNF treated pDCs from EμTCL1-Tg mice and CLL patients (Supplementary Figure 3). pDCs purified from HD and co-cultured in transwell assays with CLL B cells also displayed reduced TLR9 and Flt3 protein level expression confirming a role for tumor-derived soluble factors in driving TLR9 and Flt3 downregulation (Supplementary Figure 4). These observations strongly suggest that TNF signalling in CLL mediates broad transcriptional deregulation in pDCs to subvert their numbers and function.

In contrast to TNF, no difference was detected in the serum TGF-β levels between WT and EμTCL1-Tg mice. In fact, TGF-β levels were slightly decreased in both WT and EμTCL1-Tg with age (data not shown). In contrast, expression of TGF-βR1 was significantly higher on splenic pDCs from EμTCL1-Tg mice compared to WT (Figure 6E). BM pDCs from EμTCL1-Tg mice showed no difference in TGF-βR1 expression compared to WT mice (Figure 6E). These data suggest that splenic pDCs in EμTCL1-Tg mice might have increased sensitivity to circulating TGF-β.

We next examined the restorative potential of TNF inhibition. In vitro blockade of TNF significantly increased pDC numbers and this effect correlated with restored Flt3 expression on pDCs from EμTCL1-Tg mice with advanced disease (Figure 6F and G). Similar results were obtained using these culture conditions with pDCs from patients with indolent and in particular, the progressive CLL cohort (Figure 6H and I). We next investigated whether this approach would have beneficial effects in vivo. EμTCL1-Tg mice with advanced disease were injected twice, over 14 days, with antibodies blocking either TNF or TGF-β. Mice responded to both TGF-β and TNF blockade with a significant reduction in splenic tumor burden that correlated with a significant increase in splenic pDC numbers relative to control mice (Figure 6J and K). Interestingly, serum IFNα levels were significantly increased in mice treated with either anti-TNF or anti-TGF-β, and this correlated with restoration of pDC numbers (Figure 6L). We also observed increased numbers of both CD8⁺ T cells in anti-TNF-treated EμTCL1-Tg mice (Supplementary Figure 5). Collectively, these observations suggest that TNF production and TGF-βR1 overexpression on peripheral pDCs compromises Flt3-dependent pDC maintenance during CLL progression. Normal TNF levels in the serum of patients with indolent CLL compared to HD may also explain unaffected Flt3 expression on the pDCs from these patients (Figure 6C). It remains to be seen whether the reduction in splenic tumor burden is solely due to neutralization of autocrine TNF signals that usually support tumor persistence or whether increased pDC numbers also restores some immuno-competence. Encouragingly, similar results obtained with TGF-β inhibition, which does not directly affect CLL cells(34), suggest that increased pDC numbers may indeed play an important role.