Novel Proteasome Inhibitors Have a Beneficial Effect in Murine Lupus via the dual inhibition of Type I Interferon and autoantibody secreting cells

Mice and experimental design

NZB/NZWF1, MRL/lpr, and C57BL/6 mice were supplied by Jackson Laboratories. All mice were housed in the animal facility in the University of Rochester Medical Center or Onyx Pharmaceuticals. All experiment protocols were reviewed and approved by the University of Rochester or Onyx Committee on Animal Resources. NZB/NZWF1 mice with established nephritis (24-30 weeks with durable proteinuria ≥2+ proteinuria or 100 mg/dl) (female) and female MRL/lpr mice (10 weeks) were treated with the indicated proteasome inhibitors intravenously via tail vein route, bortezomib (0.50-0.75 mg/kg D1D3), carfilzomib (3-5 mg/kg D1D2), ONX 0914 (10 mg/kg QOD × 3 or 20 mg/kg QW) or vehicle solution. Proteinuria was monitored weekly once or twice with urine dipstick (Uristix by Bayer). Spleen, peripheral blood and bone marrow lymphocytes were collected after sacrifice for flow cytometry and ELISPOT analysis. Kidneys and a portion of spleen were collected for histological analysis.

ELISA assay

Sera from PI treated mice blood samples were collected and frozen prior to quantifying total IgG antibody levels as described (20). Serum was also used for anti-dsDNA analysis following the manufacturer's instructions (Alpha Diagnostic International). Animal identification and treatment information were blinded for all ELISA samples. A proteasome active site ELISA was utilized for proteasome analysis and to monitor inhibition of constitutive and immunoproteasome active sites in PBMC, pDCs and tissue samples from treated animals, and was performed as described previously (21).

Flow cytometry analysis

FITC-CD21, biotin-CD23, PE-CD23, PE-CD1d, APC-B220, FITC-B220, APC-AA4.1,biotin-IgM, FITC-IgM, biotin-CD5 (BD Biosciences) and PE-CD11b (BD Biosciences) were used for B cell subset identification. APC-CD4 (BD Biosciences) and biotin-CD69 (BD Biosciences) were used for T cell analysis. PE-CD11b (BD Biosciences) and FITC-CD11c (BD Biosciences) were used as dendritic cell markers. Plasma cells were identified as k (kappa) light chain+ and CD138+ among the lymphocyte gated cells. Samples were run on a FACSCanto. All antibodies were purchased from eBioscience except where indicated.

Histological analysis

Kidneys of mice from different treatment groups and control mice were fixed in 10% formalin and paraffin embedded or frozen. Kidney sections (4 mm) were stained with hematoxylin and eosin. Pathology was analyzed and scored in a blinded fashion (B.I.G.). Briefly, the severity of glomerular, interstitial, and vascular lesions was determined on a scale of 0 to 4+. Multiple sections at a minimum of two different levels were examined, with each section typically containing >50 glomeruli and >25 blood vessels as described. Spleen sections were frozen and cut into 4 mm sections. Immunohistochemistry slides were stained for FDC-M1 (BD Bioscience), PE-B220, and Biotin-GL7 using a Dako LSAB2 kit. 3-color fluorescent slides were stained for FITC-MOMA (ABD serotec), AMCA-IgM (Vector labs), ALEXA647-IgD or -FITC-B220, BIOTIN-GL7 followed by SA-PE, and APC-CD4 (BD Bioscience) and sections quantitated as described (20).

Detection of antibody-secreting cells by enzyme-linked immunosorbent spot assay

Anti-dsDNA and total IgG secreting cells were detected as previously described (20). Briefly, Millipore were coated with poly–L-lysine (Sigma) and calf thymus DNA (Sigma), blocked with 2% fetal calf serum in PBS, and spleen or BM cell suspensions incubated as serial dilutions starting with 5E5 cells/well overnight at 37 °C. After incubation, plates were washed and incubated with alkaline phosphatase–conjugated goat antibody to mouse IgG (Jackson) for 1 h at RT and detected with Vector Blue Alkaline Phosphatase Substrate Kit III (Burlingame, CA). For assessment of IgG-secreting cells, serial dilutions starting with 1E5 cells/well were incubated in goat anti-mouse IgG coated plates (Southern Biotech). The developed spots were measured by ImmunoSpot 5.0 (Cellular Technology Ltd, Shaker Heights, OH).

Inhibition of IFN-α by proteasome inhibitors and pDC counts

Mouse bone marrow cells were cultured in RPMI complete medium at 0.25 million per 200 ul in 96 flat bottom culture plate over night in the presence of 500 ng/ml CpG (2216, Invitrogen) with or without PIs. Human PMBCs were cultured in serum free media (invitrogen) containing 1ng/mL GM-CSF (Sigma) and 50 U/mL IFN-α (Sigma) at 2.5 × 10⁶ cells/ml with DMSO control or proteasome inhibitor. After 1 h incubation, cells were washed twice with media and cultured in media containing CpG2216. Cells were harvested 24 h after initial treatment. IFN-α production in the culture supernatants was measured by commercial ELISA (Mouse Interferon Alpha ELISA Kit or human ELISA kit PBL interferon source, Piscataway, NJ) following the manufacturer's instructions. In separate experiments mouse cells were harvested and stained for B220, CD11b, CD11c and Ly6C to count pDC numbers by flow cytometry analysis. In some experiments intracellular cytokine staining for IFN-α was performed (Fix & Perm Medium A and B; Invitrogen) with GolgiPlug added during the final 3 h of the culture (BD Bioscience).

RNA analysis

Total RNA was extracted from mouse splenocytes using the Qiagen RNeasy Mini kit (Qiagen, Valenica, CA) and reverse-transcribed into cDNA using the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA). Mx1 expression was defined in duplicate or triplicate real-time quantitative reverse transcriptase-polymerase chain reactions using the Rotor-Gene 3000 thermal cycler (Corbett Life Science, Sydney, New South Wales, Australia) with SYBR Green I (Applied Biosystems, Foster City, CA) for detection of DNA synthesis. Primers were as described (22) with normalization to the β2-microglobulin housekeeping gene and the following cycling conditions: 95°C for 4 minutes followed by 50 cycles of 95°C for 15 seconds, 60°C (annealing temperature) for 60 seconds, and 72°C for 15 seconds.

Statistical analysis

Student's t-test or non-parametric Mann-Whitney U was used for comparison between treatment groups. Chi-squared test was performed on protein survival data. Significance is based on a value of p<0.05.

Novel proteasome inhibitors prevent nephritis progression in Lupus prone mice

To evaluate the ability of carfilzomib and ONX 0914 to prevent lupus nephritis, 10 week-old female MRL/lpr mice were treated for 13 weeks. Both carfilzomib and ONX 0914 inhibited progression of nephritis to a similar level as bortezomib (Fig. 1a left panel and supplemental data). High levels of proteinuria (100 mg/dl) were observed in all the vehicle treated mice by the end of the treatment, whereas less than 20% of treated mice reached this level of proteinuria (Fig. 1a right panel). Similarly, NZB/NZW F1 mice with established nephritis (2+ proteinuria) showed a halt in disease progression (Fig. 1a, right). There was also a significant decrease in the severity of glomerulonephritis (GN) and interstitial inflammation after treatment with ONX 0914 (p=0.03 and 0.003, respectively) or bortezomib (p=0.001 and 0.002, respectively). The impact of carfilzomib was less marked achieving significance only for GN (p=0.05) (Fig 1b). In contrast, the control group displayed severe GN with crescents, necrosis, and mesangial hypercellularity and massive interstitial nephritis (Fig. 1b, left).

Serum anti-dsDNA IgG levels were lowered by carfilzomib and ONX 0914 treatments to a level comparable to that of bortizomib treated mice (Fig 1c). The total IgG levels were also significantly reduced by bortezomib and ONX 0914. Although carfilzomib had effects on total IgG levels early in treatment, this effect became less pronounced over time. This may be because the maximally tolerated dose for carfilzomib in the mouse results in less inhibition of LMP7 (50 – 60%) relative to ONX 0914 and bortezomib (≥80%) (data not shown). Taken together, these data support the hypotheses that proteasome inhibition, including selective inhibition of the immunoproteasome, results in therapeutic improvement in mouse models of SLE.

Elimination of plasma cells and germinal center cells in Lupus prone mice by proteasome inhibition

It has been previously demonstrated that bortezomib decreases plasma cell numbers in the spleen and bone marrow of lupus prone mice (9). Furthermore, we have demonstrated that carfilzomib and ONX 0914 reduce both anti-dsDNA and total IgG levels in the sera of treated animals. Therefore, we measured the impact of proteasome inhibition on plasma cell numbers in spleen and bone marrow of 23-week old MRL/lpr mice after treatment with bortezomib (0.5 mg/kg), carfilzomib (3 mg/kg), ONX 0914 (10 mg/kg), or vehicle solution for 13 weeks. Plasma cells decreased in bortezomib treated animals by 90% and 95% in both spleen and bone marrow, respectively. Carfilzomib and ONX 0914 treatment also decreased plasma cell numbers although not as markedly, by 50% and 65% in spleen and bone marrow, respectively (Fig. 2a). In the NZB/W F1 mice model, splenic plasma cell numbers also decreased approximately 80% after 8 week-treatment of established nephritis with ONX 0914 (20 mg/kg) (Fig. 2d).

Although flow cytometric analysis is sensitive for assessing numbers of plasma cells, it does not allow analysis of sub-populations of antigen specific antibody secreting cells (ASC). Thus, total IgG and anti-dsDNA IgG ASC numbers were measured by ELISPOT assay in 24 week-old MRL/lpr mice that were treated with bortezomib (0.5 mg/kg), carfilzomib (3 mg/kg), ONX 0914 (10 mg/kg) and vehicle solution for 13 weeks (Fig. 2b). In spleen, the level of decrease in total IgG ASC numbers was highest with bortezomib (74%) followed by carfilzomib (17%) and ONX 0914 (15%). The level of decrease in anti-dsDNA IgG ASC numbers in spleen was even more pronounced: bortezomib (87%), ONX 0914 (44%), and carfilzomib (39%) (Fig. 2b). In bone marrow, the level of decrease in total IgG ASC number was highest with bortezomib (63%) followed by ONX 0914 (44%) and carfilzomib (43%), again with more pronounced declines in anti-dsDNA IgG ASC number [bortezomib (91%), ONX 0914 (79%) and carfilzomib (64%)] (Fig. 2b). Thus, it appears that proteasome inhibition preferentially affects autoantibody secreting plasma cells. The impact on the plasma cell compartment was confirmed by histologic analysis of spleen, where decreases in both numbers and size of IgM and IgG bright plasma cells were observed with proteasome inhibitor treatment (Fig. 2c). Therefore, proteasome inhibition not only eliminates plasma cells but also may impact their antibody secreting capacity or preferentially inhibit high antibody secretors.

Given that lupus prone NZB/W F1 mice develop germinal centers (GC) spontaneously as they age, GC cells (B220+, GL-7+) were enumerated in mouse spleen after 8 weeks of treatment with ONX 0914 (20 mg/kg) and compared to vehicle treated control mice. The GC cells in the spleen of treated mice were 78% lower in numbers compared to the control, and no GC formation was observed histologically (Fig. 2d). Therefore, one explanation for the lower numbers of plasma cells found in spleen and bone marrow after long-term treatment with proteasome inhibition is a reduction in the de novo synthesis of plasma cells in GC reactions.

Another pathway for reduction of the plasma cell compartment is direct killing of preexisting plasma cells. As an initial assessment of the in vivo plasma cell killing by proteasome inhibition, we treated mice for a short-term (one week) during which the de novo synthesis of GC experienced plasma cells should be negligible. Plasma cell numbers and ASCs were enumerated by flow cytometry and ELISPOT analysis as previously described and were reduced (supplemental Fig. 2). As with long-term treatment, the effects of proteasome inhibition on autoreactive ASCs were more marked than for total IgG ASCs (supplemental Fig. 2). Although GC cells from spleen were reduced modestly with ONX 0914 treatment there was no apparent reduction with carfilzomib treatment (data not shown).

Proteasome inhibitors preferentially target active autoantibody secreting cells

To more firmly establish that PIs directly and preferentially kill anti-dsDNA ASCs, we examined their effects on antibody secretion in vitro. When PIs were added in titrating amount to ELISPOT wells, the spot numbers of total IgG ASC did not decrease (Fig. 3a and b). However, the size distribution of spots changed from random to small size spots (Fig. 3a and b), suggesting that more active IgG ASCs are more sensitive to proteasome inhibiton (Fig. 3d, left shift from vehicle). In contrast, both the numbers and size of anti-dsDNA ASC spots decreased as proteasome inhibitor concentrations increased (Fig. 3a and c). The most plausible explanation for this result is that the active anti-dsDNA ASCs are more sensitive to proteasome inhibition, with either preferential killing and/or inhibition of antibody secretion activity.

Proteasome inhibitors abrogate IFN-α production in vitro

One of the signature pathogenic cytokines found in lupus patients and mice is IFN-α produced in large quantities by pDCs. Given that PIs have been demonstrated to have effects on the production of other pro-inflammatory cytokines by a variety of immune cells, and pDCs effectively become factories for production of IFN-α, we postulated that one of the important effects of proteasome inhibition in lupus may be abrogation of the IFN-α signature. First, we tested if PIs prevent the production of IFN-α by cells stimulated ex vivo via TLR (CpG) activation. Remarkably, all three PIs significantly suppressed the production of IFN-α by mouse BM cells in a concentration dependent manner (Fig. 4a). To evaluate the impact of selective immunoproteasome inhibition, we compared the IFN-α production in CpG stimulated bone marrow cells exposed to ONX 0914 or PR-893 at concentrations resulting in selective inhibition of LMP7 or β5, respectively (Fig. 4b). LMP7 inhibition blocked production of IFN-α by over 90% whereas selective inhibition of β5 did not alter cytokine release. Bortezomib and carfilzomib, which have dual inhibition of β5 and LMP7, also abrogated IFN-α production.

Next, we examined the impact of proteasome inhibition on IFN-α production by human PBMCs. Similar to the mouse studies, LMP7 inhibition blocked cytokine production (Fig. 5a). Of note, this occurred with multiple TLR ligands, including TLR9 (CpG) and TLR3 agonists, at multiple agonist concentrations (Fig. 5a and supplemental data). A similar result was observed with purified pDCs, the cell responsible for the largest production of IFN-α (data not shown). Moreover, analysis of purified pDCs revealed that over 95% of their proteasome activity is mediated by the immunoproteasome (Fig. 5b).

Proteasome inhibitors block IFN-α production in vivo

The in vivo effects of proteasome inhibition were assessed using bone marrow cells from treated NZB/W F1 mice activated ex vivo by CpG. Both bortezomib and ONX 0914 reduced IFN-α production by approximately 75%, whereas carfilzomib caused a 40% reduction (Fig. 4c). These results demonstrate that proteasome inhibition can suppress the production of IFN-α in lupus. As further demonstration of the in vivo relevance of this inhibition to the disease process, we observed a significant decrease in the expression of the interferon-inducible gene Mx1 in spleen cells from NZB/W F1 mice treated with the proteasome inhibitors (Fig. 4d).

To further define the mechanisms of this effect, we determined whether pDCs were altered in numbers after in vivo or in vitro treatment with PIs. Notably, the fractions of pDCs in bone marrow were similar after in vivo treatment (Fig. 6a), as were absolute numbers (vehicle treated mice 97.3 × 10³±21.5 vs. PI treated mice 98.8 × 10³±29.2). However, upon in vitro stimulation there were moderate decreases in pDC numbers in a PI concentration dependent fashion. Notably, IFN producing pDCs appeared to be particularly susceptible to proteasome inhibitor induced cell death (Fig. 6c). Given the presence of residual pDCs at PI concentrations that completely abrogate IFN production, it is likely that proteasome inhibition suppresses pDC function as well via inhibition of IFN production and/or secretion.