pH Alkalinization by Chloroquine Suppresses Pathogenic Burkholderia Type 6 Secretion System 1 and Multinucleated Giant Cells

In the absence of CLQ, MNGC formation was induced after infection with B. mallei, B. thailandensis, or B. pseudomallei (Fig. 1A to F; a complete list of strains that were tested and observed to induce MNGCs is shown in Table 1), as shown by either immunofluorescence (Fig. 1A and E) or Wright-Giemsa staining (Fig. 1C). Without CLQ, multiple nuclei were found to be stacked and clustered at cell centers in large Burkholderia-induced MNGCs. These appeared to be morphologically similar to foreign body-type giant cells (FBGCs), whereby nuclei cluster centrally within the giant cells stacked on top of one another, but distinct from Langhans giant cells, in which nuclei are arranged in a horseshoe pattern (26). The rate at which MNGCs form depended on the multiplicity of infection (MOI) and cell confluence. Larger MNGCs were formed from the fusion of smaller MNGCs (see Movie S1 in the supplemental material). Given optimal conditions, we have observed that some MNGCs contain as many as 1,000 nuclei, likely as a result of fusion between several large MNGCs (data not shown).

CLQ, a lysosomotropic drug used to block phagosome maturation, was found to inhibit the MNGC formation induced by the three species (Fig. 1A to F). B. mallei-, B. pseudomallei-, and B. thailandensis-infected cells remain mostly mononucleated in the presence of CLQ (Fig. 1B, D, and F). CLQ inhibited MNGC formation by strains capable of actin tail and lateral flagellum formation, e.g., B. pseudomallei MSHR305 (27), as well as those strains that form actin tails but lack lateral flagella, e.g., B. pseudomallei AI, derived from B. pseudomallei K96423 (NCBI accession number NC_006350.1 ). All MNGCs that were induced by the strains that we tested (Table 1), including B. pseudomallei MSHR305 and B. pseudomallei AI, were inhibited by CLQ. Furthermore, CLQ blocked the MNGC formation induced by B. thailandensis in other cell types, such as baby hamster kidney cells (data not shown). We also tested the microtubule-depolymerizing agent nocodazole (NOC), a drug that inhibits the formation of FBGCs (28), which are morphologically similar to Burkholderia-induced MNGCs. NOC did not inhibit the MNGC formation induced by B. pseudomallei (Fig. 1E and G) or B. thailandensis (data not shown).

MNGC formation was quantified by counting the number of nuclei and cells of both mononucleated and multinucleated cells in a given field of view and was expressed as the ratio of the number of nuclei/number of cells (the N/C ratio). Figures 1B, D, and F show the results of quantification in B. mallei-, B. thailandensis-, and B. pseudomallei-infected cells. The N/C ratio was >1 for untreated infected control cells at later time points typically starting at 12 h postinfection (p.i.), whereas the N/C ratio of CLQ-treated infected cells was ∼1. Hence, CLQ interfered with the formation of MNGCs.

Burkholderia-induced MNGC formation is a terminal process that is invariably followed by host cell death (18, 29–31). To investigate if CLQ also inhibits the death of infected cells, we used calcein AM to label live cells and ethidium homodimer 1 (EthD-1) to label dead cells. As CLQ had similar effects on all three Burkholderia spp., as shown in Fig. 1, B. thailandensis was used as a model pathogen for B. mallei and B. pseudomallei. Figure 2A shows that CLQ-treated B. thailandensis-infected cells were mostly labeled with calcein AM, which indicates that they were alive, whereas a majority of untreated infected (control) cells were labeled with EthD-1, which indicates that they were dead. Dead cells also retained calcein AM, but labeling was mostly perinuclear and morphologically distinct from the homogeneous staining seen in live cells. In addition, the release of lactate dehydrogenase (LDH) from cells, indicative of cell membrane permeability and death, was measured in B. thailandensis-infected cells left untreated or treated with CLQ. The release of LDH from infected cells (expressed as percent cytotoxicity) was lower with CLQ treatment than no treatment (Fig. 2B) (P = 0.0179 at 18 h p.i.). Of note, CLQ treatment of uninfected cells did cause low levels of cell lysis, although the difference was not statistically significant (Fig. 2C) (P = 0.35 at 18 h p.i.). Lastly, a terminal deoxynucleotidyl transferase-mediated dUTP nick end labeling (TUNEL) assay based on fluorescence imaging was used to look for fragmented DNA, found in either pyroptotic or apoptotic cell death. Figure 2D shows pyknotic nuclei labeled with dUTP in infected cells that received no treatment, in contrast to the lack of labeled cells after CLQ treatment. Thus, cell death induced by Burkholderia can be prevented by CLQ.

On the basis of confocal imaging, fewer bacilli were seen inside CLQ-treated cells, in contrast to untreated infected cells. To quantify intracellular growth, B. thailandensis-infected cells were lysed and plated for determination of the number of CFU. Bacterial counts confirmed that fewer bacilli were present with CLQ treatment than no treatment, and the difference became larger over time (Fig. 3A). The disparity in the numbers of CFU was not a consequence of differences in adherence or uptake; no difference in the number of CFU was seen when cells were treated with CLQ prior to or during infection and when cells were not treated (Fig. 3B). Moreover, no growth defect was observed when B. thailandensis was grown in vitro in Luria-Bertani (LB) broth (Lennox) in the presence of increasing concentrations of CLQ (Fig. 3C). Therefore, CLQ specifically inhibits the growth of Burkholderia inside cells.

A hallmark of pathogenesis is escape from the phagosome into the cytoplasm, a prerequisite to the actin tail polymerization and MNGC formation that follow (15). To quantify phagosomal escape, the bacilli inside phagosomes delineated by LAMP1 were scored. A progressive reduction in B. thailandensis and LAMP1 colocalization was seen in untreated cells, whereas with CLQ treatment, B. thailandensis was retained in the phagosomes at later time points (Fig. 3D and E). Scoring for actin tails or nucleation at the bacillus poles confirmed this result. Figures 3F and G show that the number of bacilli associated with actin increased dramatically at 6 h p.i. in control cells, which coincides with the time when 90% of bacilli no longer colocalized with LAMP1. Of note is the 2-h lag in the formation of the actin tail after escape from a LAMP1-positive (LAMP1⁺) compartment. In contrast, no increase in the number of bacilli with actin tails was seen in CLQ-treated cells. Therefore, one way that the ensuing MNGC formation is prevented by CLQ is by delaying or preventing the pathogen from escaping the phagosome. Taken together, the growth inhibition that was observed with CLQ at 9 to 18 h (Fig. 3A) in the absence of MNGCs (Fig. 1B, D, and F) suggests that CLQ inhibits growth within single cells likely due in part to the inhibition of escape to the cytosol, in addition to reducing growth by inhibiting the subsequent formation of MNGCs, which provides new cells and nutrients for infection. Given the inhibitory effect of CLQ on growth and MNGC formation in macrophages, we next examined the effectiveness of CLQ in the treatment of infected animals.

The innate immune systems of insects and mammals have a high degree of similarity (32). Insect hemocytes and mammalian neutrophils phagocytose and kill microorganisms (33). Antimicrobial peptides are also produced in recognition of pathogen-associated molecular patterns by Toll receptors in insects and Toll-like receptors in mammals (33). Insects are a tractable surrogate host that may be used to reduce the number of mammals used (33, 34), and in particular, Madagascar hissing cockroaches (HCs) have been used in B. mallei, B. pseudomallei, and B. thailandensis virulence studies (35). MNGCs are formed in the hemolymph of HCs upon Burkholderia infection, and the virulence in HCs correlates with that in rodent models of infection (35). Here, in 4 separate experiments, groups of 10 or 12 HCs were infected with B. thailandensis, left untreated or treated with CLQ, and followed for 7 days. Treatment with CLQ increased the percentage of B. thailandensis-infected HCs that survived compared to the number of infected HCs that received no treatment that survived (Fig. 4) (P = 0.0270). Ninety-seven percent of HCs survived when they were inoculated with phosphate-buffered saline (PBS) or CLQ alone. Thus, CLQ treatment partially protects HCs infected with Burkholderia.

We next began studies to address the possible mechanisms by which CLQ prevents MNGC formation. A previous study showed that CLQ may block autophagy by interfering with autophagosome clearance, as evidenced by an increase in the observable number of autophagosomes (36). Similarly, we observed an increase in the number of autophagosomes, stained by monodansylcadaverine (MDC), in uninfected cells treated with CLQ (Fig. 5A), consistent with the blockage of autophagy. Burkholderia infection without CLQ treatment did not result in an increase in the number of autophagosomes; no increase in MDC staining was observed (Fig. 5B). Therefore, we wanted to examine whether CLQ's action involved autophagy by utilizing other inhibitors of autophagy to determine whether they also inhibit MNGC formation. B. thailandensis-infected cells that were treated with the autophagy inhibitors (36) 3-methyladenine (3MA) and leupeptin (LEU) did not inhibit MNGC formation (Fig. 5C and D). Furthermore, rapamycin, an inducer of autophagy, also did not inhibit MNGC formation (Fig. 5C and E). Therefore, autophagy is likely not involved in CLQ's ability to inhibit the replication and MNGC formation of Burkholderia-infected cells.

CLQ raises the pH of acidic compartments, accumulates in endolysosomes, and suppresses the activation of lysosomal proteases (37). Using LysoTracker red to label acidic organelles and DQRed bovine serum albumin (BSA), a substrate that fluoresces upon cleavage by acid-activated lysosomal proteases, we confirmed that acidic organelles were neutralized with CLQ (Fig. 6A and B). Therefore, we asked if alkalinization of acidic compartments creates a suboptimal growing environment for Burkholderia resulting in less intracellular growth and replication. To address this, B. thailandensis was grown in LB broth (Lennox) adjusted to pH 6.8 (neutral pH), pH 5.5, or pH 5.0 (acid pH). There was significantly more growth at pH 5.0 or at pH 5.5 than at pH 6.8 at 6 and 8 h (Fig. 6C). Thus, Burkholderia prefers acidic conditions for optimal growth, which is consistent with the lower intracellular growth of B. thailandensis seen in CLQ-treated cells in Fig. 3A.

To determine if alkalinization also leads to retention in phagosomes, B. thailandensis-infected cells were treated with bafilomycin A (BAF) or ammonium chloride (NH₄Cl), each of which also alkalinizes phagosomes, and scored for LAMP1 colocalization. B. thailandensis was retained in phagosomes in cells treated with BAF (Fig. 6D) or NH₄Cl (Fig. 6E), whereas it was not in untreated infected (control) cells. MNGC formation was also inhibited with BAF or NH₄Cl (Fig. 6F). Notably, the retention of B. thailandensis in phagosomes by BAF directly points to endocytic organelles as the organelles targeted by CLQ because BAF inhibits vacuolar H⁺-ATPase-dependent acidification.

CLQ belongs to a class of antimalarial drugs which are purported to have weak base properties (37). Do these CLQ analogs also inhibit MNGC formation? To address this, B. thailandensis-infected cells were treated with 75 μM amodiaquine dihydrochloride dihydrate (ADQ) or 150 μM primaquine bisphosphate (PMQ) and examined for MNGC formation. Figure 7A shows that both drugs inhibited MNGC formation in infected cells. ADQ also inhibited MNGC formation in B. mallei-infected cells (Fig. 7B). Thus, by creating an alkaline environment inside cells, CLQ prevents optimal growth, phagosomal escape, and MNGC formation.

T3SS-3 and its effectors are important for Burkholderia replication and phagosomal escape (17, 18). Therefore, the expression of T3SS-3 effectors was examined after growth at pH 5 and 7. Western blot analyses showed similar BopC protein levels at both pHs, whereas higher levels of BopE were seen at pH 5 than pH 7 (Fig. 8A), likely because these proteins are in different operons. We conclude that the expression of some T3SS-3 effectors is regulated in response to pH and the acid-induced effectors are likely required for replication and phagosomal escape.

Bacterial genes transcriptionally regulated by the two-component regulatory system VirAG, including the T6SS-1 gene cluster, are important for MNGCs and virulence (19, 20). Therefore, we examined the expression of TssM, a protein regulated by VirAG, at pH 5 and 7. Figure 8B shows that B. mallei ΔtssM::GFP(pBHR2), a mutant in which the tssM gene has been replaced with a promoterless green fluorescent protein (GFP) gene, fluoresced when it was grown at pH 5 but not when it was grown at pH 7. In contrast, B. mallei ΔtssM::GFP(pBHR2-virAG), a mutant carrying a plasmid that expresses VirAG under the control of a constitutive promoter, fluoresced at pH 7 (Fig. 8B), demonstrating that TssM can be expressed at pH 7 when VirAG is upregulated. Further, the increased expression of TssM in B. mallei at pH 5 compared with that at pH 7 was also confirmed by Western blotting (Fig. 8C). Thus, TssM is produced in response to acid pH.

Hcp1 is an essential T6SS-1 protein that forms the channel of the secretion apparatus (19). Western blot analysis showed that Hcp1 is expressed by B. mallei (Fig. 8D) and B. pseudomallei (Fig. 8E) at pH 5 but is absent or is present in reduced amounts at pH 7. As expected, the B. mallei Δhcp1 mutant did not express Hcp1 at pH 5 (Fig. 8D). Interestingly, no Hcp1 expression was seen in the B. mallei ΔvirG mutant at pH 5, demonstrating that the pH-dependent expression of Hcp1 requires a functional VirAG (Fig. 8D). Hcp1 was expressed at pH 7 in B. mallei SR1(pBHR2-VirAG), in which VirAG is constitutively expressed, as anticipated (Fig. 8D). Thus, acidification drives Hcp1 and TssM upregulation, and by preventing acidification, CLQ inhibits this process.

The bacterial strains and plasmids used in this study are described in Table 1. B. pseudomallei and B. thailandensis strains were grown in LB broth (Lennox) (Sigma-Aldrich) for 14 to 16 h at 37°C. For the assays whose results are shown in Fig. 8A, C, and D, M9 minimal broth was used. All media used to grow B. mallei, LB and M9 medium, were supplemented with 4% glycerol. Kanamycin (10 μg/ml) was added to the cultures of strains carrying plasmids. All work with B. pseudomallei and B. mallei was performed in a biosafety level 3 facility.

RAW 264.7 cells were maintained in Dulbecco's modified Eagle medium with 6 mM l-glutamine and 10% fetal bovine serum at 37°C with 5% CO₂. RAW 264.7 cells (4.5 × 10⁵) were seeded on a 12-mm number 1-1/2 coverslip (Electron Microscopy Sciences) for MNGC formation, whereas 2 × 10⁵ cells were seeded for phagosomal escape studies. For Movie S1 in the supplemental material, 1.6 × 10⁶ macrophages were seeded on a 35-mm fluorodish with a 0.17-mm-thick glass bottom.

For MNGC formation, RAW 264.7 cells were infected the day after plating with B. thailandensis at an MOI of 1 or with B. pseudomallei or B. mallei at an MOI of 5 and incubated at 37°C for 1 h. For Movie S1, cells were infected with B. thailandensis at an MOI of 10. The MOIs used were chosen to optimize the rate of MNGC formation within 18 h. For LAMP1 and actin studies, cells were infected with B. thailandensis at an MOI of 20 and incubated at 37°C for 1 h. After uptake, the coverslips were washed vigorously and successively in five 100-ml volumes of PBS to remove extracellular bacilli. The coverslips were then placed in antibiotic-free medium with or without drugs. For the assay whose results are shown in Fig. 2B, cells were infected with B. thailandensis at an MOI of 5 and incubated at 37°C for 45 min. The coverslips were washed and placed in medium with 10 μg/ml gentamicin for 30 min to kill extracellular bacilli. Subsequently, the coverslips were washed again and placed in antibiotic-free medium with or without drugs. The time p.i. begins at the time of infection and ends at the time of cell fixation or assay. For the lactate dehydrogenase (LDH) release assay whose results are shown in Fig. 2B and C, medium was removed from the wells at the times indicated in Fig. 2B and C and assayed according to the manufacturer's protocol (Roche), and the results were read on a SpectraMax M5 plate reader (Molecular Devices). One hundred percent LDH release was obtained from uninfected cells lysed with 0.2% Triton X-100.

Drugs were added after removal of extracellular bacilli and for the remainder of the experiment until fixation or assay. The drug concentrations used were 75 μM (unless otherwise noted) for chloroquine diphosphate (CLQ), 250 nM for bafilomycin A (BAF), 30 mM for ammonium chloride (NH₄Cl), 10 μg/ml for nocodazole (NOC), 75 μM for amodiaquine dihydrochloride dihydrate (ADQ), and 150 μM for primaquine bisphosphate (PMQ). All drugs were from Sigma-Aldrich.

RAW 264.7 cells (8 × 10⁵) were seeded on an 18-mm coverslip, infected the next day with B. thailandensis (at an MOI of 1 for the assay whose results are shown in Fig. 3A) for 45 min, and treated with 10 μg/ml gentamicin for 30 min to kill extracellular bacilli. The coverslips were washed as stated above and then incubated in antibiotic-free medium with or without 75 μM CLQ. For the assay whose results are shown in Fig. 3B, an MOI of 20 was used and CLQ was also added prior to infection for 1 h and during infection. To discount the possibility of extracellular growth of bacilli that escaped from cells, the coverslips were washed in PBS prior to cell lysis with 0.2% Triton X-100. Dilutions were plated on LB (Lennox) agar plates for determination of the number of CFU. For broth cultures for which the results are shown in Fig. 3C and to estimate the MOI, readings of the optical density at 600 nm were taken using a DU530 UV/visible spectrophotometer (Beckman).

Cells were fixed with 4% paraformaldehyde for 1 h and blocked in 0.25% saponin, 0.2% BSA fraction V, and goat or donkey serum (1:100; Sigma-Aldrich) for 2 h. Cells were incubated with rabbit anti-Burkholderia antibody (Ab; 1:1,000; kindly provided by David Waag, USAMRIID) and rat anti-LAMP1 Ab (ID4B; 1:200; Developmental Studies Hybridoma Bank) overnight at 4°C. After they were washed, the cells were incubated with secondary Ab conjugated to a fluorophore (1:600; Jackson ImmunoResearch) for 5 h. Cells were stained with 5 μg/ml DAPI (4′,6-diamidino-2-phenylindole) and phalloidin (1:80; Life Technologies) for 2 h prior to mounting with Fluoromount G mounting medium (Electron Microscopy Sciences). Different sets of coverslips were used to score for LAMP1 and actin. Cell preparation and staining for the following assays were performed according to the manufacturers' protocols: the (i) Click-iT TUNEL Alexa Fluor, (ii) Image-iT live lysosomal and nuclear labeling (LysoTracker red DND-99), and (iii) live/dead viability/cytotoxicity (calcein AM and EthD-1) assays (all from Life Technologies) and the autophagy/cytotoxicity dual staining (MDC and propidium iodide [PI]) assay (Cayman). For the assay whose results are shown in Fig. 6B, cells were pulsed with 10 μg/ml DQRed BSA (Life Technologies) for 1 h and chased for 3 h prior to live cell imaging.

Confocal microscopy and live cell imaging were performed on a Zeiss 700 confocal system using either a 40× (numerical aperture [NA], 1.3) or a 100× (NA, 1.4) oil objective lens and Zen 2012 (black edition) software. The pinhole was set to 1 Airy unit. For long-term live cell imaging (Movie S1), a stage-top incubator was used to maintain humidity, temperature, and CO₂ levels.

Permeabilized cells were subjected to Wright-Giemsa stain solution for 10 min, washed 3 times with PBS, and air dried. Light microscopy was performed on a Nikon E800 microscope equipped with a 60× (NA, 1.4) oil objective (Nikon) and a Microfire charge-coupled-device camera (Optronics), and the images were analyzed using Picture Frame software (Optronics).

For expression of recombinant B. mallei Hcp1 with a C-terminal 6×His tag (rHcp1-6×His), the hcp1 open reading frame (BMAA0742) was PCR amplified from B. mallei ATCC 23344 genomic DNA using primer pair Bmhcp1-HisF1 (5′-CCCAACGGTCTCACATGCTGGCCGGAATATATCTCAAGG-3′) and Bmhcp1-HisR1 (5′-CCCAACGGTCTCAAGCTTCAATGATGATGATGATGATGCGCCGCCGCGCCATTCGTCCAGTTTGCGGC-3′); BsaI sites are underlined. For expression of recombinant B. pseudomallei BopE with a C-terminal 6×His tag (rBopE-6×His), a DNA fragment encoding amino acids 70 to 261 of BopE was PCR amplified from B. pseudomallei strain K96243 using primer pair BpbopE-HisF1 (5′-CCCAACGGTCTCACATGCGCTACGTCGGCAGCTATCG-3′) and BpbopE-HisR1 (5′-CCCAACGGTCTCAAGCTTCAATGATGATGATGATGATGCGCCGCCGCGCCGTCCGCCGCGTTCGTC-3′); BsaI sites are underlined. The resulting PCR products were digested with BsaI and cloned into pBAD/HisA digested with NcoI/HindIII, generating plasmids pBmhcp1-His and pBpbopE-His, respectively. rHcp1-6×His and rBopE-6×His were overexpressed in Escherichia coli TOP10 cells and purified essentially as previously described (48). Purified rHcp1-6×His and rBopE-6×His were used to raise Hcp1- and BopE-specific polyclonal antisera, respectively, in rabbits at Cocalico Biologicals, Inc., using a standard protocol. All procedures involving animals were performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Research Council (49). All protocols were approved by the Cocalico Biologicals, Inc., Animal Care and Use Committee.

For the assay whose results are shown in Fig. 8A, C, and E, B. mallei or B. pseudomallei cultures were grown for 24 h at 37°C in M9 broth with 4% glycerol. Bacterial pellets were suspended in PBS and boiled for 30 min. A micro-bicinchoninic acid protein kit (Thermo Scientific) was used to quantify the proteins. Lysates (10 μg/lane) were run on NuPAGE 10% bis-Tris gel and transferred onto 0.45-μm-pore-size nitrocellulose membranes (Life Technologies). The membranes were blocked with Tris-buffered saline with 0.1% Tween 20 (TBST) and 5% nonfat milk for 2 h. Incubation with rabbit anti-BopC Ab (kindly given by Sunee Korbsrisate, Mahidol University, Bangkok, Thailand), rat anti-BopE Ab (Cocalico), mouse anti-TssM Ab (53), rabbit anti-Hcp1 Ab (Cocalico), mouse anti-DnaK Ab (Stressgen), or rabbit anti-GroEL Ab (Stressgen) occurred at 4°C overnight. All primary Abs were diluted between 1:3,000 and 1:5,000. Following washes with TBST, the membrane was subjected to horseradish peroxidase-anti-rabbit IgG, anti-mouse IgG, or anti-rat IgG Ab (1:10,000; all from Southern Biotech) for 1 h. The blots were visualized using SuperSignal West Pico chemiluminescent substrate (Thermo Scientific) and a Gel Logic Pro 6000 imaging system (Carestream). The blots were stripped using Restore Western blot stripping buffer (Thermo Scientific).

Madagascar hissing cockroaches (HCs; Gromphadorhina laevigata; Carolina Biological Supply) were housed in the dark at room temperature in mouse cages and fed Kibbles 'n Bits dog food (Big Heart Pet Brands), an occasional apple wedge, and water. The 50 lethal dose (LD₅₀), determined in duplicate experiments (n = 2), was established using 8 HCs per group receiving increasing doses of B. thailandensis. The LD₅₀ was calculated through Probit analysis using SAS software (version 9.3; SAS Institute, Cary, NC) and determined to be 11 CFU.

HCs (∼6 g) were acclimated at 37°C for 2 to 3 weeks. B. thailandensis (10 to 20 LD₅₀s) was suspended in PBS alone or with 12 mg/ml CLQ in a 25-μl volume immediately prior to injection. The animals were infected as previously described (35), and deaths were recorded for 7 days. The dose of CLQ was calculated to provide 300 μg CLQ in 25 μl (50 mg/kg), and the HCs received one dose only. The use of HCs for research is not regulated by any department of the U.S. federal government or any animal use committee.

Unless otherwise noted, statistical significance was calculated by an unpaired t test using the online QuickCalcs program at GraphPad. The ratios of the number of nuclei/number of cells in Fig. 1, 5, and 7 and the CFU counts in Fig. 3A and B were calculated using an asymptotic normal test based on a negative binomial regression model. Models of CFU counts included mixed effects to adjust for repeated measurements on biological replicates. The significance of the difference in 7-day survival between the B. thailandensis- and B. thailandensis-infected and CLQ-treated groups shown in Fig. 4 was also calculated using an asymptomatic normal test based on a generalized linear model having a binomial distribution, identity link, and fixed effect adjustment for animal cohort. All analyses were carried out with SAS software (version 9.4; SAS Institute, Cary, NC). In all figures, only those differences with P values of <0.05 are indicated.