The role of macrophages in the susceptibility of Fc gamma receptor IIb deficient mice to Cryptococcus neoformans

Dysfunctional polymorphisms of FcγRIIb, an inhibitory receptor, are associated with Systemic Lupus Erythaematosus (SLE). Cryptococcosis is an invasive fungal infection in SLE, perhaps due to the de novo immune defect. We investigated cryptococcosis in the FcγRIIb−/− mouse-lupus-model. Mortality, after intravenous C. neoformans-induced cryptococcosis, in young (8-week-old) and older (24-week-old) FcγRIIb−/− mice, was higher than in age-matched wild-types. Severe cryptococcosis in the FcγRIIb−/− mice was demonstrated by high fungal burdens in the internal organs with histological cryptococcoma-like lesions and high levels of TNF-α and IL-6, but not IL-10. Interestingly, FcγRIIb−/− macrophages demonstrated more prominent phagocytosis but did not differ in killing activity in vitro and the striking TNF-α, IL-6 and IL-10 levels, compared to wild-type cells. Indeed, in vivo macrophage depletion with liposomal clodronate attenuated the fungal burdens in FcγRIIb−/− mice, but not wild-type mice. When administered to wild-type mice, FcγRIIb−/− macrophages with phagocytosed Cryptococcus resulted in higher fungal burdens than FcγRIIb+/+ macrophages with phagocytosed Cryptococcus. These results support, at least in part, a model whereby, in FcγRIIb−/− mice, enhanced C. neoformans transmigration occurs through infected macrophages. In summary, prominent phagocytosis, with limited effective killing activity, and high pro-inflammatory cytokine production by FcγRIIb−/− macrophages were correlated with more severe cryptococcosis in FcγRIIb−/− mice.

Accordingly, the high susceptibility to cryptococcosis in the Fcγ RIIb− /− strain appeared to be due to the gene defect and less likely a result of autoantibody stimulation or ongoing SLE.
Additionally, the levels of pro-inflammatory serum cytokines, TNF-α and IL-6, but not an anti-inflammatory cytokine, IL-10, were higher in Fcγ RIIb− /− mice ( Fig. 4B-D,F-H). Moreover, infected Fcγ RIIb− /− mice showed a tendency toward higher levels of total immunoglobulin (measured as total gamma globulin protein from protein electrophoresis) but this did not reach a statistically significant level ( Supplementary Fig. S4). Therefore, cryptococcosis, either at 2 weeks or when moribund, was more severe in Fcγ RIIb− /− than wild-type mice, as demonstrated by higher fungal burdens in most internal organs, higher liver enzyme levels, and higher pro-inflammatory cytokine levels but not higher anti-inflammatory cytokine levels.
Additionally, the average number of fungi in each macrophage was higher in Fcγ RIIb− /− cells, as determined by the phagocytosis index (number of internalized fungi/total macrophages) (Fig. 6B). In contrast, the macrophage killing activity, as determined by fungal viability after incubation with macrophages for 2, 4 and 24 h, was not different between wild-type and Fcγ RIIb− /− cells. The numbers of viable fungi in the macrophages with total killing ability (both extruded and intracellular yeasts were determined; see methods), in vitro, at 2, 4, and 24 h of fungal incubation with wild-type cells versus Fcγ RIIb− /− macrophages were 8.2 ± 0.8, 4.4 ± 0.5 and 5.5 ± 0.7 vs. 20.6 ± 2.8, 15.6 ± 2.1 and 14.9 ± 2.1 (x10 4 ) CFU/ml, respectively (Fig. 6C). No difference in macrophage killing activity was observed using the intracellular proliferation assay. The intracellular proliferation assay that determined the viability of only intracellular yeasts also performed (see methods). Indeed, the intracellular proliferation amounts at 2, 4, and 24 h of fungi incubated with wild-type cells versus fungi incubated with Fcγ RIIb− /− macrophages were 1.3 ± 0.2, 1.0 ± 0.2 and 12.3 ± 1.8 vs. 1.3 ± 0.1, 0.7 ± 0.2 and 8.8 ± 1.9 units, respectively (Fig. 6D). Although a trend toward greater intracellular-killing by Fcγ RIIb− /− macrophages was observed, it was not statistically significant. The presence of macrophages did not reduce the colony count of fungi in the fungicidal activity assay, implying that cryptococci were viable, intracellularly.

Discussion
The Fcγ RIIb loss-of-function polymorphism is one of the important genetic associations of SLE, particularly in Asian populations [12][13][14][15] . Accordingly, Fcγ RIIb− /− mice, a lupus nephritis model, may be a useful mouse model of SLE in Asians. Our results suggest the possibility that the increased severity of cryptococcosis in FcγRIIb− /− mice is, at least in part, due to the unique properties of FcγRIIb− /− macrophages including enhanced phagocytosis and elevated pro-inflammatory cytokine responses.
Although intravenous administration is not the natural route of cryptococcal infection, which is transmission through pulmonary system, an intravenous model is adequate for a proof of concept of the difference between the two mouse strains. Indeed, Fcγ RIIb− /− mice showed more severe cryptococcosis than wild-type mice. The immune response against fungal infection is predominantly dependent on cell-mediated immune responses. The defect in Fcγ RIIb, the only inhibitory receptor in the Fcγ R family, induces an enhanced immune response and effectively controls several organisms [7][8][9] . Fcγ RIIb− /− mice were more susceptible to cryptococcosis, either in symptomatic or asymptomatic lupus, compared to the age-matched wild-type controls. There was also an age-dependent increase in the mortality rate for cryptococcosis. The mortality rate in wild-type mice at 8-weeks-old vs. 24-weeks-old was different (p = 0.015 by log-rank test, Fig. 1A,B), but there was no difference between the Fcγ RIIb− /− groups. The severity of cryptococcosis in Fcγ RIIb− /− mice was independent of the manifestation of lupus symptoms and age. These results support the susceptibility to cryptococcosis of patients with Fcγ RIIb polymorphisms, with either SLE or non-SLE [10][11][12] .
Subsequently, the severity of cryptococcosis was evaluated with different parameters at 2 weeks, the earlier stage of infection, and in the moribund stage. The 2 age groups, as expected, showed similar results. Disseminated cryptococcosis, including fungal organisms found in several internal organs, which was observed in Fcγ RIIb− /− mice, was similar to the disseminated cryptococcosis observed in patients with compromised immune systems 17 . However, the lesions were limited mostly to the brain in wild-type mice, a major target organ 18 , as usually observed in the immuno-competent human host. Surprisingly, despite the absence of generalized lesions in wild-type mice, the brain lesions were larger than in Fcγ RIIb− /− mice. This difference remains to be investigated. It seems that the neurotropism characteristic of C. neoformans is not apparent in the infection in Fcγ RIIb− /− mice because the organism could survive in any organ. However, in wild-type mice, C. neoformans may prefer to grow in the brain because of more suitable nutritional conditions [19][20][21] . Enhanced pathogenesis due to disseminated disease was also indicated by more severe liver injury in Fcγ RIIb− /− mice.
Enhanced cryptococcal dissemination in the Fcγ RIIb− /− mice underscores the role of this genetic lesion in pathogenesis. The inflammatory cytokines, TNF-α and IL-6, but not IL-10, were higher in Fcγ RIIb− /− than wild-type mice. This is a response to the higher fungal burden, in addition to the anti-inflammatory defect. Perhaps serum IL-10 is too low to balance out the pro-inflammatory immune responses, leading to more severe organ histopathology and symptoms. However, total immunoglobulin levels did not differ between the Fcγ RIIb− /− and wild-type mice after cryptococcal infection, despite the previous report of immunoglobulin hyperproduction in bacterial infection 8 . This inconsistency is possibly due to the difference in adaptive immune-response activated by these two organisms. Nevertheless, the beneficial effect of humoral immune response in controlling C. neoformans was demonstrated [22][23][24] .
Macrophages and T helper cells are the main immune cells responsible for the immune response to cryptococcosis 25,26 . Fcγ RIIb presents in macrophages but not in T cells 27 . Therefore, the higher fungal burdens in Fcγ RIIb− /− mice may be due to the primary defect in macrophages. Accordingly, we evaluated the phagocytosis activity, killing activity and cytokine responses of macrophages after exposure to C. neoformans. Interestingly, the phagocytosis of Fcγ RIIb− /− macrophages in response to cryptococci was elevated compared with wild-type cells, as reported for other organisms 8,9 . In Fcγ RIIb− /− mice, nearly all of the macrophages incubated with heat-killed C. neoformans phagocytosed approximately 6-7 yeasts per cell at 4 h (phagocytosis index). In contrast, approximately only 50-60% of wild-type macrophages phagocytosed yeast, and they did so with a reduced activity of 2-3 yeasts per cell. On the other hand, the total killing capacity and the intracellular proliferation activity of Fcγ RIIb− /− macrophages was not different from wild-type macrophages, unlike the responses to other organisms 8 , perhaps due to the immune evasion properties of C. neoformans 26 .
Cryptococcus is a facultative intracellular pathogen, which can utilize host macrophages to spread within the body, via the Trojan horse mechanism 16 . Cryptococci typically escape extracellular immune responses, survive and replicate intracellularly, transfer laterally between macrophages, and eventually invade tissue and organs 16 . They can use macrophages as trafficking vehicles for dissemination, particularly to pass through the blood-brain barrier into the central nervous system 28 . Interestingly, the depletion of macrophages, at least in certain situations, is associated with less severe pathogenesis 29 . We hypothesized that the elevated phagocytosis of Fcγ RIIb− /− macrophages and the immune evasion properties of C. neoformans enhanced the Trojan horse mechanism, resulting in more severe cryptococcosis in vivo. We tested cryptococcosis severity in a macrophage depletion model with daily liposomal clodronate injection in Fcγ RIIb− /− and wild-type mice. As expected, at 1 week after fungal administration, macrophage depletion led to lower fungal burdens in the liver, lung and spleen of Fcγ RIIb− /− mice, but not of wild-type mice. Nevertheless, liposomal clodronate not only depleted monocytes/macrophages but also reduced the numbers of dendritic cells and regulatory T cells 30,31 . Although loss of dendritic cells after liposomal clodronate injection might be responsible for the less severe cryptococcosis, the enhanced cryptococcosis severity after Cryptococcus-infected macrophage injection supports the greater pathogenic role of macrophages compared to dendritic cells. The inoculation of fungi-containing Fcγ RIIb− /− macrophages increased the fungal burdens in the brains and livers of wild-type mice at 24 h after administration. These results support the high phagocytosis capacity of Fcγ RIIb− /− macrophages and the enhancement of fungal transmission by macrophages, particularly through the blood-brain barrier. Together, these results support the importance of macrophages in cryptococcosis pathogenesis in Fcγ RIIb− /− mice and in patients with Fcγ RIIb loss-of-function polymorphisms.
In addition, the prominent pro-inflammatory cytokine response (TNF-α and IL-6), but not the anti-inflammatory cytokine (IL-10) response, was demonstrated in the Fcγ RIIb− /− groups, both in vivo and in vitro. This might be because glucuronoxylomannan (GXM), an important cryptococcal capsular polysaccharide, induces potent immunosuppression by direct engagement of Fcγ RIIb, an immunoinhibitory receptor, and stimulates greater IL-10 production 32,33 . In Fcγ RIIb− /− mice, perhaps GXM was unable to induce IL-10, resulting in the more severe pro-inflammatory cytokine storm and fungal sepsis. More studies on this topic are needed to explain the underlying mechanism.
Taken together, we conclude that more severe cryptococcosis in Fcγ RIIb− /− mice was due to enhanced dissemination, possibly through the Trojan horse mechanism, and the hyper-responsiveness of pro-inflammatory cytokine production during sepsis. This is the first report of the disadvantage of the prominent macrophage function of Fcγ RIIb− /− mice in cryptococcosis. In clinical translation, we propose Fcγ RIIb loss-of-functionpolymorphisms as a new risk factor for cryptococcosis. Screening for Fcγ RIIb polymorphisms in patients with SLE, particularly in areas of endemic cryptococcosis, might be beneficial for patient management. Our results also implied the importance of the genetic-background differences among patients with SLE to micro-organism responses. The examination of the genetic background of individual patients with SLE might be clinically beneficial.

Methods
Animal models and Cryptococcus neoformans injection method. Fcγ RIIb− /− mice on the C57BL/6 background were kindly provided by Dr. Silvia Bolland (NIAID, NIH, Maryland, USA). The mice were originally constructed in a 129Sv/B6-hybrid background and were backcrossed onto the C57BL/6 background for 12 generations. Female C57BL/6 wild-type mice, age-matched to Fcγ RIIb− /− mice, were purchased from the National Laboratory Animal Center in Nakornpathom, Thailand. The animal protocols were approved by the Faculty of Medicine of Chulalongkorn University and followed NIH criteria. C. neoformans was isolated from a patient sample (Mycology Unit, King Chulalongkorn Memorial Hospital), identified by morphology, together with urease production and melanin synthesis (L-3,4-dihydroxyphenylalanine or DOPA test), and stored in Sabouraud dextrose Broth (SDB) at − 80 °C. The sample accession process was approved by the Ethical Institutional Review Board, faculty of Medicine, Chulalongkorn University according to the declaration of Helsinki, with written informed consent. The same strain of C. neoformans was used in all of the experiments. Before use in experiments, C. neoformans was sub-cultured on Sabouraud dextrose agar (SDA) at 37 °C for 24 h. Asymptomatic (8-week-old without proteinuria) or symptomatic lupus (24-week-old with proteinuria) mice or age-matched wild-type control groups were injected, via tail vein, with 1 × 10 5 yeast cells C. neoformans diluted in 200 μ l of PBS. For survival analysis, the mice were observed for 90 days after fungal administration. The mice were sacrificed at the moribund stage, as determined by an inability to walk after touch stimulation. In other experiments, the mice were sacrificed at 2 weeks after fungal administration. At the time of euthanasia, blood was collected via cardiac puncture under isoflurane anaesthesia, and the internal organs (brain, lung, kidney, liver and spleen) were fixed with 10% formalin for histology or processed for fungal burden experiments (details below). In addition, to further investigate the importance of Fcγ R deficiency relative to autoantibody and incipient SLE status, 4-week-old Fcγ RIIb− /− mice and age-matched wild-type (non-significantly different anti-dsDNA antibody titers between both groups; Supplementary Fig. S6) were intravenously injected with C. neoformans at 1 × 10 5 yeast cells and internal organ fungal burdens were determined at 2 weeks.
In vivo macrophage depletion. Macrophage depletion with liposomal clodronate injection was performed in order to determine the role of macrophages in cryptococcosis, following a previously published protocol 28 . Female 8-week-old Fcγ RIIb− /− and wild-type mice were administered C. neoformans (yeast form) via the tail vein. Then, 200 μ l/mouse liposomal clodronate (Encapsula Nanoscience, Nashville, TN, USA) (5 mg/ml) or control liposomes were injected to induce sustained monocyte depletion (Fig. 8A). The daily injections began on the third day of fungal administration and continued for 4 consecutive days. At 7 days post-inoculation, the mice were sacrificed and the internal organs were processed for fungal burdens and fixed in 10% formalin to confirm macrophage depletion (by immunohistochemical staining with an F4/80 antibody; Biolegend, San Diego, CA, USA). Macrophages were not detectable in organs after liposomal clodronate treatment in either Fcγ RIIb− /− or wild-type mice (Supplementary Fig. S5).

Transfer of Cryptococcus-containing macrophages in vivo.
Bone marrow (BM)-derived macrophages from Fcγ RIIb− /− and wild-type mice were allowed to phagocytose yeast cells and then they were infused into wild-type mice, as previously described 34,35 . Briefly, BM macrophages cultured in a 96-well plate at 2.5 × 10 4 cells/well with 20% mouse serum were incubated with C. neoformans in the ratio of 5:1 (fungal cells to macrophages) for 2 h. The un-phagocytosed fungi were washed out with DMEM (3-5 washes). Subsequently, the macrophages were detached with cold-PBS washing (3-5 times) and centrifuged at 1,000 rpm for 10 min at 4 °C. Cell pellets were then resuspended with DMEM. The macrophages were counted and stained with trypan blue. Either Fcγ RIIb− /− or wild-type macrophages with internalized Cryptococcus, at 2.5 × 10 4 cells, were intravenously administered to wild-type mice through the tail vein. The mice were sacrificed at 24 h and the fungal burdens determined.

Fungal burdens and organ histology.
To measure internal organ fungal burdens, the organs were weighed, homogenized, plated onto SDA and incubated at 37 °C for fungal colony enumeration at 48 h. For histology, the tissue samples were fixed in 10% formalin and embedded in paraffin; 4-μ m sections were stained with haematoxylin and eosin colour (H&E) and Grocott's silver stain (GMS) for C. neoformans identification. Quantitative measurement of the fungal infection area was performed by 2 blinded observers. Fields (10 selected randomly) were examined at 200x magnification, with the following criteria: 0, no fungi; 1, area of fungal infection < 25%; 2, infected area involving 25-50% of the field; and 3, infected area ≥ 50% of the field.
Blood chemistry, urine chemistry and cytokine analysis. Kidney injury was determined by serum creatinine (Scr) (QuantiChrom Creatinine Assay, DICT-500, BioAssay, Hayward, CA, USA), and liver injury was assessed via alanine transaminase levels (ALT) (EnzyChrom ALT assay, EALT-100, BioAssay). For the evaluation of antibody responses, mouse serum was analysed for total immunoglobulin 36 by capillary protein electrophoresis (MINICAP-2 Sebia, Evry Cedex, France). The percentage of protein in the gamma zone of protein electrophoresis was converted into total immunoglobulin level by multiplying the ratio of protein at the gamma zone by the serum total protein. Serum total protein and urine protein were measured by Bradford protein assay. Urine protein creatinine index (UPCI), a representative of 24 h urine protein, was determined by the following equation; UPCI = spot urine protein/spot urine creatinine. Cytokine measurement (TNF-α , IL-6 and IL-10) in serum and supernatant media were measured using ELISA assays (eBioscience, San Diego, CA, USA).
Anti-dsDNA antibody detection. Calf DNA (Invitrogen, Carlsbad, CA, USA) coated on 96-well plates was used for measuring anti-dsDNA antibodies, following a previously published protocol 37 . In brief, the plates were coated with calf DNA at 100 μ g/well and incubated overnight at 4 °C. The plates were dried, filled with 100 μ l/well of blocking solution, incubated at room temperature for 1.5 h and washed. Subsequently, mouse serum samples at 100 μ l/well were added and incubated for 1 h. Then, the plate was washed, incubated with peroxidase-conjugated goat anti-mouse antibodies (BioLegend, USA) at 100 μ l/well at room temperature for 1 h, washed and developed with ABTS peroxidase substrate solution (TMB Substrate Set; BioLegend) for 10 min in the dark. Finally, the stop solution (2 N H 2 SO 4 ) was added, and the plate was read with a microplate photometer at a wavelength of 450 nm.
Bone marrow-derived macrophage preparation. Bone marrow (BM)-derived macrophages were produced following an established protocol 38 . In brief, BM cells from femurs were centrifuged at 1,000 rpm in 4 °C for 10 min. Then, the cells were incubated in BMM or DMEM complete (DMEM supplement with 10% fetal bovine serum, 1% penicillin/streptomycin, HEPES and sodium pyruvate) plus 5% horse serum (HyClone TM donor horse serum, Thermo Scientific, Waltham, MA, USA) and 20% L929-conditioned media in a humidified 5% CO 2 incubator at 37 °C for 7 days. The cells were harvested at the end of the culture period using cold PBS, and the macrophage phenotype was confirmed by flow cytometry with anti-F4/80 and anti-CD11b antibodies (BioLegend, USA).

In vitro Cryptococcus neoformans-induced macrophage cytokine production. Heat-killed
C. neoformans, (immersion in a 60 °C water bath for 1 h) or live C. neoformans, at a dose of 5 × 10 5 yeast cells/ml, were incubated with macrophages (1 × 10 5 cells/well) in 96-well polystyrene tissue culture plates 39 . The culture supernatants were collected at various time points and stored at − 80 °C until use. After the incubation, the cell viability was measured by the MTS cell proliferation assay (One Solution Cell Proliferation Assay, Promega Corporation, WI, USA) according to the manufacturer's instructions 40 . Briefly, 20 μ l of MTS were added to the culture plates, incubated for 2 h at 37 °C in a 5% CO 2 incubator, and then read with a microplate photometer at a wavelength of 490 nm. All of the experiments showed cell viability of greater than 95% (data not shown).
Phagocytosis, macrophage total killing activity and intracellular proliferation assays. The phagocytosis and macrophage killing activity assessments were performed according to a previous protocol, with slight modifications 34,35 . Briefly, BM-derived macrophages were added to 96-well plates at 2.5 × 10 4 cells/ well in DMEM complete and incubated overnight. After the incubation, LPS from Escherichia coli 026:B6 (Sigma-Aldrich, St. Louis, USA) was added at final concentration of 100 ng/ml and the plates were incubated in 5% CO 2 at 37 °C for 24 h. The medium was then removed and 100 μ l of complete DMEM with 20% normal mouse serum, as a source of opsonin, were added with heat-killed C. neoformans at various ratios of fungal cells to macrophages. These were incubated for various periods. After incubation, the wells were washed with 200 μ l of PBS at least 3 times to remove un-ingested yeast and then the macrophages were detached with 200 μ l of cold PBS. Then, the macrophages were transferred to a CytoSpin chamber (Thermo Scientific) and centrifuged at 600 rpm for 5 min to concentrate the cells into a single cell-layer for easier visualization. The macrophages were stained with Diff-Quick stain (Life Science Dynamic Division, Nonthaburi, Thailand). Macrophages containing yeast were counted as showing phagocytosis. At least 100 macrophages per well were counted. In parallel, the ingestion ability of each individual macrophage was determined as the average number of fungal cells in each macrophage (phagocytosis index), calculated as the total number of ingested fungi divided by the total number of macrophages. The phagocytosis activity was determined as the percentage of macrophages with phagocytosed cryptococci. All of the experiments were performed in triplicate.
Total cryptococcal cell killing activity was assessed using a previously published method which both extruded and intracellular yeasts were determined 34 . Briefly, BM-derived macrophages at 1 × 10 5 cells/well were co-cultured with live C. neoformans at a ratio of 1:1 for 2, 4 and 24 h. LPS and mouse serum were used as mentioned above. After incubation, the culture supernatant was separated, and a lysis medium (distilled water containing 0.01% bovine serum albumin and 0.01% Tween-80) was added to the wells, for 20 min at 37 °C, to rupture the macrophage cell membranes. Then, the supernatant and the lysate were well mixed. Serial dilutions of the mixed lysates were plated on SDA for viable yeast colony forming unit (CFU) counts. Control cultures consisted of incubation medium alone plus C. neoformans. Macrophage total killing activity is inversely correlated with the number of yeast colonies. The intracellular and extracellular killing activity of macrophages was evaluated with this method.
Moreover, to determine only the intracellular killing activity, the intracellular proliferation assay was performed with methods that were slightly modified from those previous published 41,42 . Briefly, BM-derived macrophages, at 2.5 × 10 4 cells/well, were incubated for 17 h with IFN-γ (10 ng/ml final concentration; BioLegend, USA). Then, LPS (100 ng/ml final concentration; Sigma-Aldrich) was added and incubation continued for 24 h. Cells were then washed with PBS. After that, live C. neoformans, at a ratio of 5:1, were added and incubated for 2 h in 20% normal mouse serum containing media to promote phagocytosis. After 2 h, the wells were extensively washed (4-5 times) to remove extracellular fungi. This was set as the 0 h time-point. For some of the culture wells at this time-point, macrophage cell lysis were induced by lysis medium and plated on SDA for the visualization of intracellular fungal viability for "phagocytosis activity at the 0 h time-point" for the further calculation. The remaining culture wells from the 0 h time-point were maintained in DMEM media at 37 °C, and the cells were subsequently lysed at 2, 4 and 24 h, for the determination intracellular fungal viability, as mentioned above. Because the difference in intracellular fungi might be due to the difference in phagocytosis activity at the 0 h time-point, the phagocytosis activity at the 0 h time-point was used for the normalization with the following equation: intracellular proliferation at specific time points = CFU of fungi at 2, 4 or 24 h after the 0 h time-point/ CFU of fungi after phagocytosis at the 0 h time-point. Macrophage intracellular killing activity is inversely correlated with the number of yeast colonies (intracellular proliferation).
Statistical analysis. The mean ± SE was used for data presentation, and the differences among groups were examined for statistical significance using the unpaired Student's t-test or one-way analysis of variance (ANOVA) with Tukey's comparison test for the analysis of experiments with 2 and 3 groups, respectively. The repeated measures ANOVA with Bonferroni post hoc analysis was used for the analysis of the data with several time-points. Survival analyses were evaluated with the log-rank test. P values < 0.05 were considered statistically significant. SPSS 11.5 software (SPSS Inc., Chicago, IL, USA) was used for all statistical analysis.