Suppression of lupus nephritis and skin lesions in MRL/lpr mice by administration of the topoisomerase I inhibitor irinotecan

Background Since the precise mechanism for the pathogenesis of systemic lupus erythematosus (SLE) is unknown, no targeted therapies in addition to immunosuppression are available so far. We recently demonstrated that administration of the topoisomerase I (topo I) inhibitor irinotecan at extremely low concentrations reversed established lupus nephritis in NZB/NZW mice. While profound immunosuppression was absent, we proposed changes in DNA relaxation and anti-double-stranded (ds)DNA antibody binding as the underlying mechanism. To exclude that these effects were restricted to NZB/NZW mice, irinotecan was used in a genetically different strain of lupus-prone mice. Methods MRL/lpr mice were treated with high- and low-dose irinotecan beginning at 8 weeks of age. Treatment was repeated every fourth week. In vitro, DNA was relaxed by recombinant topo I, and altered anti-dsDNA antibody binding was measured by enzyme-linked immunosorbent assay. Results Administration of both high- and low-dose irinotecan prevented proteinuria and prolonged survival in MRL/lpr mice. Moreover, both concentrations of irinotecan significantly improved histopathology of the skin at 18 weeks of age. While only high-dose irinotecan diminished the numbers of plasmablasts and double-negative T cells, no changes in IgG-secreting cells or anti-dsDNA IgG were observed. In vitro, relaxation of DNA by topo I increased the binding of anti-dsDNA IgG but not the binding of anti-dsDNA IgM derived from the plasma of MRL/lpr mice. Conclusion The beneficial effects of topo I inhibition in a second, genetically different strain of lupus-prone mice strongly implicate irinotecan as a new therapeutic option for human SLE.


Background
Systemic lupus erythematosus (SLE) is a chronic autoimmune disease mainly affecting women of childbearing age. It is estimated that, in the USA, up to 275,000 adult women suffer from SLE [1]. The disease involves different organs, but immune complex glomerulonephritis most strikingly influences the course of SLE. Ten to thirty percent of patients with lupus nephritis progress to end-stage renal disease (ESRD) resulting in hemodialysis or kidney transplantation [2]. Due to the application of immunosuppressive drugs, the survival of patients with lupus-associated glomerulonephritis increased from a 5-year survival of 44 % in the 1950s to a 10-year survival of 88 % recently [3]. Despite these advances in the treatment of SLE, the life expectancy of patients with lupus and renal damage was recently demonstrated to be 23.7 years shorter compared to the general population [4]. Moreover, the incidence of ESRD associated with lupus nephritis has not decreased over the last years [5], indicating that current medication is insufficient to treat lupus nephritis.
Unselective immunosuppressive drugs remain the central strategy to control lupus nephritis. Medication consists of an induction therapy with cyclophosphamide and prednisolone or mycophenolate mofetil, followed by a maintenance therapy with azathioprine or mycophenolate mofetil [6,7]. Major side effects of this medication are infections, and they bear the risk of malignancies whereupon cyclophosphamide also causes amenorrhea [8][9][10]. Furthermore, none of the newly developed biological agents such as rituximab, ocrelizumab, atacicept, or abatacept demonstrated beneficial effects in patients with active lupus nephritis [11][12][13][14][15]. Two other biologicals reached statistical significance for the improvement of moderate to severe forms of SLE. However, the benefit of belimumab and tabalumab were modest [16,17], the "number needed to treat" was high, and other phase III studies hardly reproduced the first results [18,19].
The precise mechanism for the pathogenesis of SLE is unknown. Therefore, no targeted therapies beside immunosuppression exist so far. However, a potentially targeted therapy for SLE was suggested by a new finding from our group which was generated by serendipity. We demonstrated that the topoisomerase I (topo I) inhibitor irinotecan efficiently suppressed murine lupus nephritis in New Zealand Black/New Zealand White (NZB/NZW) mice [20][21][22]. It was the first time that topo I was linked to the treatment of lupus nephritis.
Topo I is ubiquitously expressed and highly conserved [23,24]. Its major function is the relaxation of supercoiled DNA in order to release torsional stress from DNA occurring during replication and transcription. To mediate DNA relaxation, topo I binds to DNA and cleaves one DNA strand, subsequently allowing the rotation of the cleaved strand around the other in a controlled reaction [25]. Afterwards, the nicked strand is religated by topo I restoring intact double-stranded (ds)DNA in a relaxed state.
Inhibitors of topo I stabilize a normally very transient catalytic intermediate in which topo I is bound to one strand of the DNA, known as the topo I cleavable complex. When the cleavable complex is stalled by topo I inhibitors, re-ligation of the DNA is impossible [26]. The consequence of prevented re-ligation may vary; in proliferating cells, the stalled topo I cleavable complex can collide with replication forks leading to unrepairable DNA double strand breaks and apoptosis [27,28]. For this reason, topo I inhibitors, like camptothecin and its synthetic derivative irinotecan, are widely used as anticancer drugs for several types of tumors [29]. In non-dividing cells, the treatment with topo I inhibitors results in the production of single-stranded (ss)DNA breaks reducing the replication capacity of a cell, although this is not lethal [26]. In addition to the induction of ssDNA and dsDNA breaks, topo I inhibitors were shown at least in vitro to inhibit DNA relaxation [30,31].
In our first experiments, we applied concentrations of irinotecan that were similar to those used for chemotherapy in humans [20]. Although irinotecan reversed established lupus nephritis, there have been some concerns about using yet another chemotherapeutic for the treatment of SLE in predominantly young women. However, subsequent experiments by our group demonstrated that much lower dosages were still efficient to suppress SLE. We ended up with a dose more than 50 times lower than the dose used for chemotherapy in humans, which still enabled the successful treatment of established lupus nephritis in NZB/NZW mice [21]. These extremely low concentrations guided us to the hypothesis that inhibition of topo I might be a targeted therapy for SLE. While a profound immunosuppression was ruled out [20][21][22], we hypothesized in the beginning that induction of ssDNA breaks as a consequence of topo I inhibition is the underlying mechanism [32]. However, later we found that topo I alone increased binding of anti-dsDNA antibodies [21,22] supposing that DNA relaxation is critically involved in the pathogenesis of SLE.
All previous experiments using irinotecan for the treatment of SLE were performed in NZB/NZW mice. To provide more evidence for first clinical trials treating human SLE we had to rule out that the demonstrated effects of irinotecan are restricted to NZB/NZW mice. We therefore introduced the MRL/lpr mouse model which is characterized by a fast and severe disease progression involving fatal glomerulonephritis, vasculitis, skin lesions, and massive lymphadenopathy [33,34]. In these mice, we tested whether irinotecan has similar beneficial effects on lupus-like disease as shown before in NZB/NZW mice.

Mice
Female MRL/lpr and MRL/MpJ mice, aged 6 weeks, were purchased from The Jackson Laboratory and kept in isolated ventilated cages. Immediately after arrival, mice were randomly assigned to the respective groups (five animals per cage).
Animal study: treatment of MRL/lpr with irinotecan At 8 weeks of age, MRL/lpr mice were injected intraperitoneally with saline, or 1 or 25 mg/kg irinotecan (Campto®; Pfizer). MRL/MpJ mice treated with saline were used as controls. The volume of each injection was 10 ml/kg. Mice were treated three times per week. The treatment cycle was repeated after 4 weeks. Beginning at an age of 7 weeks, mice were monitored for proteinuria and body weight once a week. Proteinuria was measured with Albustix (Siemens Healthcare Diagnostics) and analyzed semiquantitively as grade 0 (negative), grade 1+ (≥30 mg/dl), grade 2+ (≥100 mg/dl), grade 3+ (≥300 mg/ dl), and grade 4+ (≥2000 mg/dl) according to the manufacturer's recommendations. The onset of proteinuria was defined as two instances of grade 4+ proteinuria occurring 1 week apart. Moreover, skin in the dorsal neck region, from the snout, and from the ears was scored individually in a semiquantitative manner using a score system from 0, for no lesion, to 2, for severe manifestation. Mice were killed when disease became severe (proteinuria grade 4+ and a body weight loss of ≥25 % from the onset of disease) and/or the total skin score was ≥4. The experiment was terminated when mice reached 37 weeks of age.

Histopathology of kidney and skin sections
Kidney and skin obtained from the dorsal regions were fixed overnight in 4 % paraformaldehyde and embedded in paraffin. Standard protocols were used for hematoxylin and eosin, periodic acid-Schiff, and methenamine-silver staining.

Isolation of splenocytes and lung cells
Spleens were harvested from mice and immediately transferred into ice-cold phosphate-buffered saline (PBS), and smashed on a sterile grid with a pestle. Cells were incubated in red blood cell lysis buffer for 2 min on ice, debris was allowed to settle out by centrifugation at 65 g for 2 min at 4°C, and cells were re-suspended in RPMI 1640 supplemented with 10 % fetal calf serum (FCS). Cell viability was checked by trypan blue exclusion.
The left lobe of the lung was minced to small pieces and digested in RPMI 1640 without supplement containing 100 μg/ml Liberase (Roche) for 90 min at 37°C and 5 % CO 2 . Digestion was stopped by adding fetal bovine serum (Gibco) to a final concentration of 10 %. Cells were passed through a 40-μm cell strainer, incubated in red blood cell lysis buffer for 2 min on ice and resuspended in RPMI 1640 media for cell counting and further analysis.

Enzyme-linked immunosorbent assay (ELISA)
Ninety-six-well Nunc MaxiSorp plates were coated with 5 μg/ml goat anti-mouse IgG (Sigma) or anti-mouse IgM (SouthernBiotech) overnight at 4°C. Plates were blocked with 1 % bovine serum albumin (BSA) in PBS for 1 h at 37°C followed by incubation with diluted plasma samples for 1 h at 37°C. Subsequently, plates were incubated with alkaline phosphatase-conjugated goat anti-mouse IgG or anti-mouse IgM (SouthernBiotech) for 1 h at 37°C and developed with p-nitrophenyl phosphate (Sigma-Aldrich). Optical density was measured at 405 nm with a reference filter at 490 nm. Sample concentrations were calculated using a standard curve of purified mouse IgG (Sigma) and IgM (SouthernBiotech).
To determine anti-dsDNA antibodies, calf thymus DNA (Invitrogen) was passed through a Millex-HA 0.45-mm filter (Millipore) to remove any ssDNA fragments. Maxisorp plates were half-coated with 100 mg/ml calf thymus DNA in PBS overnight at 4°C. Plates were blocked with PBS containing 1 % BSA for 1 h at 37°C. Diluted plasma samples were incubated at 37°C for 1 h. Bound anti-dsDNA autoantibodies were detected as described above for total IgG or IgM ELISA, respectively.

DNA modification
Calf thymus DNA (50 mg/ml) was incubated with 4 μg/ml recombinant human topo I (Creative Biomart) in 50 mM Tris HCl (pH 7.5), 50 mM KCl, 2 mM dithiothreitol (DTT), and 1 mM EDTA for 30 min at 37°C. Then, 30 ml of sample per well was used for coating a 384-well Nunc Maxi-Sorp plate overnight at 4°C. Plates were blocked with PBS containing 1 % casein (Pierce) for 1 h at 37°C. Next, depending on the experiment, diluted plasma samples from MRL/lpr mice were added to the plates for 1 h at 37°C. Bound antibodies were detected as described above for anti-dsDNA IgG and IgM ELISA.

Statistical analysis
Data were analyzed using either one-way or two-way analysis of variance (ANOVA) followed by Bonferroni's posthoc test, respectively. Some data were assessed by paired t test. Survival data were analyzed using the Mantel-Cox log rank test. For all tests, the software GraphPad Prism version 6.0 was used. P values <0.05 were considered as significant.

Results
Irinotecan attenuated symptoms of lupus-like disease and prolonged survival in MRL/lpr mice MRL/lpr mice possess a rapid and severe form of lupuslike disease. For clinical monitoring, the involvement of the kidneys with proteinuria and the involvement of skin with alopecia, pruritus, and necrotic lesions are important. Both organ contributions influence the survival of these mice, in contrast to the NZB/NZW model preferably showing renal manifestations.
Mice were treated with high-and low-dose irinotecan from 8 weeks of age. The dosages for high-and low-dose irinotecan were obtained from previous experiments in NZB/NZW mice. As described before, low-dose irinotecan corresponds to a dose which is 50 times lower than the dose used for chemotherapy in humans [20][21][22]. Saline-treated MRL/lpr mice started to develop severe proteinuria from 12 weeks of age. At 21 weeks of age, grade 4+ proteinuria measured at two consecutive weeks affected 90 % of saline-treated animals. In contrast, both high-and low-dose irinotecan significantly prevented the onset of proteinuria from 16 weeks of age (Fig. 1a). Looking at the clinically determined skin score, only high-dose but not low-dose irinotecan suppressed the onset of detrimental skin lesions (Fig. 1b). Improvement of the skin score mediated by high-dose irinotecan was statistically significant until 27 weeks of age. Notably, the graph in Fig. 1b implies that MRL/lpr mice treated with low-dose irinotecan developed more severe lupus-like skin lesions than saline-treated animals. However, since saline-treated MRL/lpr mice showed a rapidly progressing glomerulonephritis in contrast to irinotecan-treated animals, they had to be sacrificed before they could reach higher levels of skin lesions.
Excluding animals that were sacrificed due to the skin score, both high-and low-dose irinotecan demonstrated a significantly improved survival at 20 weeks of age (Fig. 1c). When including all animals, only MRL/lpr mice treated with high-dose irinotecan demonstrated an increased survival of 80 % compared to 0 % of saline-treated mice and 10 % of animals treated with low-dose irinotecan at 37 weeks of age (Fig. 1d).
For functional measurements, other groups of MRL/lpr mice were sacrificed at 18 weeks of age. MRL/MpJ mice were used as controls. Determination of the kidney score revealed an amelioration of glomerulonephritis for both irinotecan concentrations; however, only high-dose irinotecan showed a significant improvement (Fig. 2a). The skin score at 18 weeks of age assessed on paraffin sections by a blinded pathologist demonstrated a suppression of lupus-like skin lesions for both irinotecan-treated groups (Fig. 2b). Remarkably, when looking at the clinically determined parameters in Figs. 1 and 2, we noticed some differences between the groups of mice at 17 weeks of age: living saline-treated animals in Fig. 1 showed a proteinuria of 2.6 (mean, data not shown) while saline-treated mice in Fig. 2 presented a proteinuria of 1.5 (mean) at the same age.
Immunomodulatory effects of irinotecan treatment in MRL/lpr mice MRL/lpr mice bear a mutation in the Fas receptor (CD95). As a consequence, autoreactive immune cells fail to undergo apoptosis and accumulate in lymphoid structures [37]. We therefore examined how treatment with irinotecan influenced the immune system of MRL/ lpr mice. Analysis of the spleen from mice sacrificed at 18 weeks of age described above revealed clear signs of splenomegaly in saline-treated MRL/lpr mice compared to MRL/MpJ mice. This effect was significantly reduced in MRL/lpr mice treated with high-dose irinotecan, whereas low-dose irinotecan had no influence on splenomegaly (Fig. 3a). In addition to splenomegaly, a massive proliferation of CD4/CD8 double-negative (DN) T cells is characteristic for MRL/lpr mice. These cells seem to be unresponsive to antigen-receptor stimulation and have no T-helper activity [38,39]. Determination of DN T cells identified as CD3 + B220 + CD4 -CD8cells demonstrated a significant reduction of this cell population in mice treated with high-dose irinotecan. No effect was seen for low-dose irinotecan compared to salinetreated MRL/lpr mice (Fig. 3b). B-cell numbers identified as CD3 -B220 + CD138cells remained unchanged in both irinotecan-treated groups (Fig. 3c). By contrast, the expression of CD69 on B cells and the numbers of plasmablasts (CD3 -B220 + CD138 int ) were significantly decreased in MRL/lpr mice treated with high-dose irinotecan. For plasma cells (CD3 -B220 -CD138 + ), no significant effect of either highdose or low-dose irinotecan compared to saline was observed (Fig. 3c).
Next, we analyzed the numbers of antibody-secreting cells in the spleen. Surprisingly, increased numbers of IgG-and IgM-secreting cells in MRL/lpr mice were not affected by high-or low-dose irinotecan (Fig. 4a).
In line with that, no changes in the plasma levels of total IgG and IgM were detected (Fig. 4b). Measurement of anti-dsDNA IgGs showed increased plasma levels in saline-treated MRL/lpr mice which were not significantly reduced in both irinotecan-treated groups (Fig. 4c, left). Unexpectedly, when analyzing anti-dsDNA IgM we found diminished, albeit not significantly, plasma levels in MRL/lpr mice treated with low-dose irinotecan while there was no reduction of anti-dsDNA IgM in MRL/lpr mice treated with highdose irinotecan (Fig. 4c, right).

High-dose irinotecan decreased the number of DN T cells infiltrating the lung
In addition to inflammatory processes in the kidneys and skin, MRL/lpr mice also exhibit pulmonary autoimmune disease which is caused by infiltration of mononuclear cells [40]. Analysis of different immune cell subsets in the lungs of 18-week-old MRL/lpr mice revealed a significant reduction of the number of CD3 + cells and DN T cells in the group treated with high-dose irinotecan (Fig. 5a). Moreover, there was some reduction in IgG-secreting lung cells in both irinotecan-treated groups; however, the difference did not reach statistical significance (Fig. 5b).
Increased binding of anti-dsDNA antibodies to DNA modified by topo I was restricted to the IgG isotype Previous data from our group established a link between the extent of DNA relaxation and anti-dsDNA antibody binding. We demonstrated that anti-dsDNA antibody binding to DNA relaxed by recombinant topo I was clearly enhanced. This effect was shown for different types of DNA and for anti-dsDNA antibodies from different sources [21,22]. To clarify whether this effect can be also applied to anti-dsDNA antibodies derived from MRL/lpr mice, we modified calf thymus DNA with recombinant topo I and assessed anti-dsDNA IgG binding by ELISA. Compared to BSA, DNA relaxed by topo I revealed a significantly increased binding of anti-dsDNA IgG from the plasma of MRL/lpr mice (Fig. 6a), in line with our previous observations. For the first time we also determined the binding of anti-dsDNA IgM to DNA modified by topo I. Applying the same conditions (See figure on previous page.) Fig. 3 Prevention of splenomegaly by high-dose irinotecan (Irino). Spleens from 18 week-old MRL/lpr mice treated with saline (n = 7), 25 mg/kg irinotecan (n = 6), or 1 mg/kg irinotecan (n = 6) were analyzed by flow cytometry. MRL/MpJ mice treated with saline (n = 7) were used as controls (Ctrl). a Representative image of the spleen from each treatment group (left) and the spleen weights (right). b Representative bivariate contour plots from each treatment group showing the subpopulation of T cells double negative for CD4 and CD8 and positive for CD3 and B220 (DN T cells). Cells were gated for live, singlets, and negative for CD4 and CD8 expression. (Right) Total number of splenic DN T cells. c (Top left) Numbers of CD3 -B220 + CD138 -B cells and (top right) the expression of CD69. (Bottom left) Numbers of CD3 -B220 + CD138 + plasmablasts and (bottom right) CD3 -B220 -CD138 + plasma cells. Each symbol represents an individual mouse; bars show mean ± SEM.*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, versus saline-treated MRL/lpr mice by one-way analysis of variance with Bonferroni post hoc test as described for anti-dsDNA IgG binding, we found that topo I-mediated DNA relaxation did not increase the binding of anti-dsDNA IgM derived from the plasma of MRL/lpr mice (Fig. 6b). Therefore, increased anti-dsDNA antibody binding following DNA relaxation seems to be restricted to anti-dsDNA IgG, which was an unexpected finding.

Discussion
Five years ago, our group reported for the first time that the topo I inhibitor irinotecan is able to suppress lupus nephritis in NZB/NZW mice [20], a finding that was based on serendipity obtained from cancer research experiments. In those experiments, immunocompetent mice were treated with a human protein either alone or in combination with irinotecan. During the second cycle of treatment, all mice from the group treated with the protein alone died within 3 days whereas all mice treated with the protein in combination with irinotecan survived. In the review of these unexpected results we speculated about the occurrence of some form of fatal hypersensitivity reaction and a protective effect of irinotecan against this. Based on this hypothesis we introduced a mouse model for anaphylaxis in terms of fast hypersensitivity reaction. However, irinotecan failed to delay the onset of anaphylactic shock (unpublished results). We next checked whether protection from delayed hypersensitivity reactions was involved and established two mouse models of spontaneous autoimmune disease: diabetes in NOD mice and SLE in NZB/NZW. While irinotecan had no effect on the development of autoimmune diabetes (unpublished data), application of the topo I inhibitor almost completely prevented the onset of lupus nephritis. Encouraged by this finding we also examined whether irinotecan ameliorates the course of two other autoimmune diseases, multiple sclerosis and systemic sclerosis. Both attempts failed (unpublished results) providing evidence that inhibition of topoisomerase I is a specific treatment for SLE. This suggestion was supported by further experiments showing that extremely low concentrations of irinotecan still suppressed lupus nephritis and prolonged survival in NZB/NZW mice [21]. Final support for the potential use of irinotecan as a novel treatment strategy for SLE patients comes from our findings showing that irinotecan, even at a low dose, does not result in profound immunosuppression [20][21][22]. Instead, we presented evidence that alterations in DNA relaxation and subsequent changes in binding of anti-dsDNA antibodies may account for irinotecan-mediated effects in SLE [21,22].
Are these data sufficient enough to legitimize first clinical trials treating lupus patients with irinotecan? Yes, we think so. However, to provide additional support in favor of clinical testing of irinotecan in human SLE, we introduced the MRL/lpr mouse model which is genetically different from NZB/NZW [41]. MRL/lpr mice were treated with either high-dose or low-dose irinotecan. As a result, high-dose irinotecan clearly suppressed the development of lupus-associated glomerulonephritis and prevented the onset of skin lesions. When applying low-dose irinotecan to MRL/lpr mice, the results were more difficult to interpret. Undoubtedly, until week 20, lowdose irinotecan almost completely inhibited the onset of proteinuria, a time point when most of the MRL/lpr miceutilizing studies have finished [42][43][44][45]. Skin lesions were prevented by low-dose irinotecan at least until week 18, which was confirmed through histopathology assessed by a blinded pathologist. After week 18, an increasing number of mice from the group treated with low-dose irinotecan had to be sacrificed due to a skin score ≥4. The remaining mice of this group were suppressed for proteinuria, suggesting that low-dose irinotecan was more efficient at inhibiting lupus nephritis than skin lesions. However, it is very likely that intermediate dosages of irinotecan exist that are considerably lower than the doses used for chemotherapy, but which are still sufficient to suppress both lupus nephritis and skin lesions in MRL/lpr mice. This needs to be clarified in upcoming experiments. Further investigation is also warranted for the finding that topo I-mediated DNA relaxation increased binding of anti- dsDNA IgG but not binding of anti-dsDNA IgM. Notably, anti-dsDNA IgM is known to have a protective function in SLE [46][47][48][49]. However, our previous results showing that DNA relaxation and anti-dsDNA IgG binding can be inhibited by topo I inhibitors explain how irinotecan executes its protective function in SLE. Without the use of a topo I inhibitor, circulating DNA might be relaxed allowing increased binding of anti-dsDNA IgG followed by inflammatory disease. In contrast, when applying a topo I inhibitor, circulating DNA might be less relaxed and, subsequently, binding of anti-dsDNA IgG will be reduced. Then, assuming that binding of anti-dsDNA IgM is independent of DNA relaxation, more anti-dsDNA IgM with its lupusprotective function would be able to bind to circulating DNA. However, such a mechanism is very speculative thus far and needs to be preferentially shown in vivo. Summarizing our results, inhibition of topo I is highly proficient at suppressing lupus nephritis and other symptoms of lupus-like disease in two different mouse models of spontaneous SLE. Although the mechanism is not fully understood yet, it seems that topo I inhibition represents a specific and targeted therapy for SLE beyond immunosuppression. Therefore, first clinical trials applying low-dose irinotecan to human SLE patients are urgently required.

Conclusions
The topo I inhibitor irinotecan suppressed both lupus nephritis and lupus-like skin lesions in MRL/lpr mice. Immunosuppression was not significantly involved in irinotecan-mediated suppression of lupus-like disease in MRL/lpr mice. As an alternative mechanism for irinotecanmediated effects on lupus-like disease, our data propose altered binding of pathogenic anti-dsDNA IgG as a result of changes in DNA relaxation.
In view of the highly conserved nature of topo I, and considering the application of extremely low dosages of the topo I inhibitor as well as the lack of immunosuppression, inhibition of topo I might be developed as the first targeted therapy for human SLE. Authors' contributions AK and SF designed the study, analyzed and interpreted the data, and wrote the manuscript. AK, SRH, and MK performed the experiments. SRH, MK, MH, and RAS made substantial contributions to data analysis, data interpretation, and drafting of the manuscript. All authors read and approved the final manuscript.