Targeted apoptosis of macrophages and osteoclasts in arthritic joints is effective against advanced inflammatory arthritis

Insufficient apoptosis of inflammatory macrophages and osteoclasts (OCs) in rheumatoid arthritis (RA) joints contributes toward the persistent progression of joint inflammation and destruction. Here, we deliver celastrol (CEL) to selectively induce apoptosis of OCs and macrophages in arthritic joints, with enzyme-responsive nanoparticles (termed PRNPs) composed of RGD modified nanoparticles (termed RNPs) covered with cleavable PEG chains. CEL-loaded PRNPs (CEL-PRNPs) dually target OCs and inflammatory macrophages derived from patients with RA via an RGD-αvβ3 integrin interaction after PEG cleavage by matrix metalloprotease 9, leading to increased apoptosis of these cells. In an adjuvant-induced arthritis rat model, PRNPs have an arthritic joint-specific distribution and CEL-PRNPs efficiently reduce the number of OCs and inflammatory macrophages within these joints. Additionally, rats with advanced arthritis go into inflammatory remission with bone erosion repair and negligible side effects after CEL-PRNPs treatment. These findings indicate potential for targeting chemotherapy-induced apoptosis in the treatment of advanced inflammatory arthritis.

15) The sentence beginning on line 151 is confusing and should be rewritten. 16) Figure 2C is not mentioned in the text, although it probably would go on line 157. 17) Why were murine macrophages used for in vitro studies, rather than rat?
Reviewer #2 (Remarks to the Author): This manuscript aims for delivering celastrol (CEL) to selectively induce apoptosis of both types of cells (osteclasts and macrophages) in RA joints. While this is a novel work and nothing wrong with the science, it lacks the required impact to be published as it is in Nat Comm Journal. The work is mostly focused on in vivo studies using AIA rat model, but would be most impactful if validation is done using human cells obtained from patients with RA (e.g. ex vivo models). The fate of PLGA NPs should also be clarified by providing bioelimination and cell internalization (dynamic of endocytosis/exocytosis) data. In addition, some refinement is required with respect to the manuscript contents. Grammar revision should be once performed. Include scale bars in all figures. Some references are also incomplete. In brief, this manuscript requires major revision prior its acepptance for publication.

Reviewer #3 (Remarks to the Author):
This is a manuscript that focuses on a novel method of delivering celastrol for suppressing arthritis. The bulk of the manuscript has been published by other groups looking at AIA model and celastrol treatment of macrophages and osteoclasts. The main novelty is the delivery vehicle. However, to demonstrate the efficacy of the prodrug, the authors should use alpha v Beta 5 intern receptor KO mice and MMP9-/-mice.

Reviewer #1 (Remarks to the Author):
This manuscript describes the development of a nanoparticle therapeutic (CEL-PRNP) capable of inducing apoptosis of macrophages and osteoclasts, which was used to treat rat AIA. The PRNP was designed to express RGD peptides to target v3 integrin positive cells and another protective PEG chain with a MMP9 cleavable peptide. When in an inflammatory environment where MMP9 is expressed, the protective molecule would be cleaved allowing the RGD peptide to bind the integrin positive cells, allowing the CEL to be internalized and released inducing apoptosis. The CEL-PRNP was effective at treating both early and late AIA, although the in vivo mechanism is less clear. An approach to target pathogenic macrophages may be helpful in a subset of patients with RA.
There were a number of concerns with the manuscript as submitted, and addressing them may strengthen the final product.
1) The authors should have the manuscript reviewed by someone more facile with English. It was difficult to follow at times. This was especially true in the introduction, which has many misstatements and exaggerations. A couple of examples: synovial hyperplasia is not a symptom; continuingly suffer; extremely low; and various tumor cells.
Response: Per the Reviewer's suggestion, we have rewritten the manuscript with the help of an English native speaker. In addition, we have proofread our manuscript and corrected the errors in our revised manuscript.
2) While the therapeutic approach described may potentially be appropriate for some patients with RA, there are already many effective therapies that prevent joint erosion and destruction. Also recent studies have shown that JAK inhibitors may reverse erosions, as may other biologic therapies. The rationale should be more realistically presented.

Response:
We appreciate the Reviewer's suggestion and agree that JAK inhibitors may reverse the bone erosions in RA. Such information including the efficacy and potential side effects of JAK inhibition (Arthritis Rheum., 2012, 64:970-981. Nat. Rev. Rheumatol., 2017 has been added into the "INTRODUCTION" section of the revised manuscript. 3) It took a lot of effort to figure out what the different compounds were, and what the abbreviations represented. Figure 2A gives a clear description, but the manuscript text did not. RNP was not described in the text, that I could fine.
Expanding in the text and referring to the diagram would help. Also, Figure 2B is really the hypothesis being tested and is not actually data. 6) The term activated macrophages was over used. In vitro differentiated macrophages treated with LPS would be appropriately referred to as activated.

Response
However, it is unlikely that every macrophage in an AIA joint is activated. It is likely that tissue-resident macrophages are also in the joint. The state of activation of synovial tissue macrophages was not studied.
Response: In agreement with the Reviewer's opinion, there are macrophages including tissue-resident macrophages remaining quiescent in arthritic joints.

7)
The data concerning the uptake in off target organs in figure 4A needs to be quantitated and statistically interrogated. This is a major point for using the PRNPs.
Response: Per the Reviewer's suggestion, we have quantitated and statistically analyzed the drug distribution of various DiD formulations in major organs (heart, liver, spleen, lung and kidney) and inflamed joints. Data represent mean ± SD (n = 3), *P < 0.05, **P < 0.01, ***P < 0.001. (f) The statistical graphs of the fluorescence intensity of inflamed joints and un-inflamed joints from AIA rats with a unilateral inflamed joint after i.v. administration of free DiD or DiD-labeled nanoparticles. Data represent mean ± SD (n = 3), *P < 0.05, **P < 0.01.

OCs, endothelial cells, other cells?
Response: We appreciate the Reviewer's suggestion. To investigate whether PRNPs could target both OCs and activated macrophages in inflamed joints, the in vivo distribution behavior of different DiD formulations in both type of cells has been determined using immunofluorescent staining method. Activated macrophages and OCs in inflamed joints were determined by immunofluorescence analysis of CD68 and CD51, respectively. As shown in Fig. 6, the DiD fluorescence distribution of PRNPs in synovial joint was the highest among the three nanoparticle types, which was consistent with the results of in vivo and ex vivo imaging studies. In addition, free DiD and DiD-labeled NPs showed low levels of colocalization of the red (DiD) and green fluorescence (CD68 and CD51), suggesting the nonspecific distributions in inflamed joints. Whereas, the DiD fluorescence of RNPs and PRNPs was mainly overlapped with the green fluorescence in synovial joints. The DiD fluorescence of PRNPs group was significantly brighter than that of RNPs group. These results have demonstrated that PRNPs efficiently delivered drugs into OCs and activated macrophages in inflamed joints.

Also, we have conducted in vitro experiments to investigate the distribution behaviors of PRNPs in human umbilical vein endothelial cells (HUVECs) and
TNF-α-activated HUVECs (mimicking neovascularization endothelial cells). In HUVECs, all prepared nanoparticles displayed low level of drug distribution, demonstrating the low affinity of prepared nanoparticles to normal endothelial cells.  Supplemental Fig. 15 The relative quantity of IL-1β (a), TNF-α (b), OCN (c) and ALP (d) in arthritic joints in different groups. **P < 0.01, ***P < 0.001. Data represent mean ± SD (n = 5). Response: Thanks for the Reviewer's suggestion. We have applied TUNEL immunostaining method to evaluate the apoptosis in inflamed joints of AIA rats with advanced arthritis after various CEL formulation treatments. As shown in Fig. 7a, CEL-PRNPs group showed the highest green fluorescent signal in arthritic joints, demonstrating the highest level of cellular apoptosis among all the treated groups.

10) Does treatment with CEL-PRNP promote apoptosis of cells in the joints
Furthermore, PRNPs could selectively deliver drugs to synovial macrophages and OCs in inflamed joints. (Fig. 6) CD68 and TRAP staining results also revealed that CEL-PRNPs was the most efficient in reducing the abundance of both synovial macrophages and OCs in inflamed joints. (Fig. 7b) Accordingly, CEL-PRNPs effectively induced apoptosis of synovial macrophages and OCs in arthritic joints of AIA rats with advanced arthritis.
In this study, we aimed to selectively inducing apoptosis of synovial macrophages and OCs in arthritic joints, thus reprograming the inflammatory microenvironment and restoring the bone function balance, in an effort to control joint inflammation and reverse bone erosions in advanced RA. As shown in Fig. 7, we demonstrated that CEL-PRNPs effectively reduced the abundance of both types of cells via inducing increased apoptosis. The reduction of synovial macrophages resulted in the decrease of inflammatory cytokines (TNF-α and IL-1β) secretion in both serum and arthritic joints, (Fig 7c, d and Supplemental Fig. 15) thus reducing swelling in ankle joints and paws. (Fig 8b-d  in blood, and TNF-α secretion in blood from rats receiving the indicated treatment. **P < 0.01, ***P < 0.001. Data represent mean ± SD (n = 5). (d) Detection of IL-1β, TNF-α, OCN and ALP expression levels in arthritic joints in different groups. Arthritic joints in different groups were stained with IL-1β, TNF-α and OCN antibodies, respectively. ALP was stained light dark in arthritic joints from different groups (Scale bar = 100 μm) (n = 5).
Supplemental Fig. 11 Quantitative analysis for the immunofluorescence of TUNEL staining in arthritic joints of AIA rats with late-stage arthritis after receiving the indicated treatment. ***P < 0.001. Data represent mean ± SD (n = 5). Fig 5D should

MMP9 expressed? MMP9 was added to the in vitro activated macrophages. How do we know that the MMP9 target peptide is cleaved in the joint?
Response: Previous studies illustrated that MMP9 was highly expressed in arthritic joints of RA. (Arthritis Rheum., 1996, 39:1576-1587J. Clin. Immunol., 2006, 26:299-307) Our MMP9 staining results also revealed the increased MMP9 expression in arthritic joints from AIA rats. (Fig. 1a) Therefore, the MMP9 target peptide could be cleaved by MMP9 in arthritic joints.

13) How much cell death or other changes occurs in the off target organs of rats treated with CEL-PRNP?
Response: Per the Reviewer's comment, we have conducted TUNEL immunostaining assays to investigate the apoptosis level of major organs (heart, liver, spleen, lung, kidney and brain) after CEL-PRNPs treatment. As shown in Supplemental Fig. 17, CEL-PRNPs induced negligible off-target apoptosis in major organs.
In addition, we have conducted serum enzyme assays and histopathological analyses to investigate the in vivo safety of CEL-PRNPs. As shown in Supplemental   Fig. 18, CEL significantly increased serum levels of ALT, AST, CK and LDH in rats, suggesting the toxicity of CEL to liver and heart. CEL-RNPs also remarkably increased serum levels of ALT and AST, demonstrating the liver toxicity of CEL-RNPs to liver. H&E staining results revealed that CEL caused obvious pyknosis of neuron in the brain, atrophy of myocardial cells and myofibrillar loss in the heart, loss of hepatic cords and dilatation of blood sinus in the liver. CEL-RNPs caused severe atrophy, loss of hepatic cords and dilatation of blood sinus in liver tissue.
CEL-PRNPs caused slight damages in liver tissue. However, CEL-PRNPs caused no significant changes in serum enzyme levels and displayed no obvious damages to heat, spleen, lung, kidney and brain. Therefore, CEL-PRNPs had good in vivo safety and could reduce the neurotoxicity, cardiotoxicity and hepatotoxicity of CEL. Supplemental Fig. 17 TUNEL immunofluorescence staining of major organs (heart, liver, spleen, lung, kidney and brain) were processed 2 days after rats receiving the indicated treatment. (Scale bar = 50 μm) (n = 3).

14) A control using PRNP, without CEL, should be included in the study. What is the in vivo effect of the nanoparticles?
Response: Per the Reviewer's suggestion, the related experiment has been carried out.
Blank PRNPs showed negligible effects on reducing swelling in ankle joints and paws, which was comparable to saline group.
Therapeutic efficacy of blank PRNPs in AIA rats with advanced arthritis. Paw thickness (a) and ankle diameter (b) of AIA rats were recorded every other day during the treatment period. Data represent mean ± SD (n = 7).

15) The sentence beginning on line 151 is confusing and should be rewritten.
Response: Per the Reviewer's suggestion, we have checked and revised our manuscript. Figure 2C is not mentioned in the text, although it probably would go on line 157.

Response:
We appreciate the Reviewer's comment and have added the related data in the revised manuscript.

17) Why were murine macrophages used for in vitro studies, rather than rat?
Response: According to previous studies, murine macrophages have been widely Previous studies demonstrated RGD had high affinity to integrin and RGD mediated endocytosis significantly increased the distribution of drug loaded nanoparticles in integrin-positive cells. (Nat. Biotechnol., 1997, 15:542-546. J. Control. Release, 2015 In this study, taking advantage of the RGD-integrin interaction, CEL was efficiently delivered into OCs and activated macrophages, thus inducing increased cellular apoptosis. Confocal imaging and flowcytometry assay revealed different drug distribution among various prepared nanoparticles on target cells and normal cells. The results have proved that both RNPs and PRNPs (with MMP9) selectively delivered drug to OCs and activated macrophages ( Fig. 3 and Fig. 4) rather than normal cells including non-activated macrophages (BMMs) and normal vascular endothelial cells (HUVECs), (Fig. 4a, d and Supplemental Fig. 8) suggesting a good targeting ability of PRNPs to OCs and activated macrophages.
To investigate the fate of prepared PRNPs after in vivo intravenous injection into AIA rats, we have studied the biodistribution behaviors of PRNPs in major organs and inflamed joints at different time points. As shown in Supplemental Fig. 10, the PRNPs distribution in major organs (heart, liver, lung and kidney) had significantly decreased within 48 h after injection, demonstrating the clearance of prepared PLGA nanoparticles in the body. We also have studied the in vivo safety of CEL-PRNPs.
Results revealed that CEL-PRNPs had negligible off-target toxicity and significantly reduced the neurotoxicity, cardiotoxicity and hepatotoxicity of CEL. (Supplemental Fig. 17 and Supplemental Fig. 18) Given the cell-targeting ability and good safety of PRNPs, the dynamic endocytosis/exocytosis of nanoparticles was not investigated in this study.
Per the Reviewer's suggestion, we have checked the scale bars and magnification of presented figures. In addition, we have proofread our manuscript and corrected the errors in the revised manuscript accordingly. LDH (d) in rats receiving the indicated treatment. *P < 0.05, **P < 0.01. Data represent mean ± SD (n = 5). (e) Hematoxylin and eosin (H&E) staining of the major organs (heart, liver, spleen, lung, kidney and brain) were processed 2 days after rats receiving the indicated treatment.

Reviewer #3 (Remarks to the Author):
This is a manuscript that focuses on a novel method of delivering celastrol for suppressing arthritis. The bulk of the manuscript has been published by other groups looking at AIA model and celastrol treatment of macrophages and osteoclasts. The main novelty is the delivery vehicle. However, to demonstrate the efficacy of the prodrug, the authors should use alpha v Beta 3 intern receptor KO mice and MMP9-/-mice.

Response:
We appreciate the Reviewer's positive comment with respect to the novelty of the present study. Also, we agree that it is a rational strategy to use knockout mice as negative controls. In our study, the in vitro results (including cellular uptake and cell apoptosis assays on activated macrophages and osteoclasts (OCs) derived from mice and RA patients) have demonstrated that CEL-PRNPs could target activated macrophages and OCs through RGD-integrin interaction after responding to MMP9. (Fig. 3 and Fig. 4) Furthermore, the in vivo results (including biodistribution, in vivo antiarthritic mechanism and pharmacodynamic assays) also have shown that PRNPs could be selectively delivered to activated macrophages and OCs in inflamed joints and trigger apoptosis of these cells, (Fig. 6 and Fig. 7) consequently controlling inflammation and reversed bone erosion. (Fig. 8 and Fig. 9) The above results have proved the efficacy of our strategy. Thus, currently, we did not carry out relevant experiments in knockout mice. However, in vivo studies in knockout mice model will be advantageous to the clinical translation of this drug delivery platform and we intend to apply this technology in our future translational research of CER-PRNPs. This information has been added into the "DISCUSSION" section as a limitation of the current study in the revised manuscript.