A Widefield Light Microscopy-Based Approach Provides Further Insights into the Colonization of the Flea Proventriculus by Yersinia pestis

ABSTRACT

Yersinia pestis (the agent of flea-borne plague) must obstruct the flea’s proventriculus to maintain transmission to a mammalian host. To this end, Y. pestis must consolidate a mass that entrapped Y. pestis within the proventriculus very early after its ingestion. We developed a semiautomated fluorescent image analysis method and used it to monitor and compare colonization of the flea proventriculus by a fully competent flea-blocking Y. pestis strain, a partially competent strain, and a noncompetent strain. Our data suggested that flea blockage results primarily from the replication of Y. pestis trapped in the anterior half of the proventriculus. However, consolidation of the bacteria-entrapping mass and colonization of the entire proventricular lumen increased the likelihood of flea blockage. The data also showed that consolidation of the bacterial mass is not a prerequisite for colonization of the proventriculus but allowed Y. pestis to maintain itself in a large flea population for an extended period of time. Taken as the whole, the data suggest that a strategy targeting bacterial mass consolidation could significantly reduce the likelihood of Y. pestis being transmitted by fleas (due to gut blockage), but also the possibility of using fleas as a long-term reservoir.

IMPORTANCE Yersinia pestis (the causative agent of plague) is one of the deadliest bacterial pathogens. It circulates primarily among rodent populations and their fleas. Better knowledge of the mechanisms leading to the flea-borne transmission of Y. pestis is likely to generate strategies for controlling or even eradicating this bacillus. It is known that Y. pestis obstructs the flea’s foregut so that the insect starves, frantically bites its mammalian host, and regurgitates Y. pestis at the bite site. Here, we developed a semiautomated fluorescent image analysis method and used it to document and compare foregut colonization and disease progression in fleas infected with a fully competent flea-blocking Y. pestis strain, a partially competent strain, and a noncompetent strain. Overall, our data provided new insights into Y. pestis' obstruction of the proventriculus for transmission but also the ecology of plague.

INTRODUCTION

Plague is caused by the bacillus Yersinia pestis. This disease remains active in several countries worldwide, including Madagascar and the United States (1). In 1898, Simond discovered that plague is transmitted by fleas—a mechanism confirmed by Gauthier and Raybaud in 1903 (2). Eleven years later, Bacot and Martin developed the first model of the transmission of Y. pestis by this hematophagous insect (3). Since then, several researchers have deepened and refined our knowledge of the pathophysiological mechanisms leading to flea-borne transmission of Y. pestis (4–10).

In the current model, the flea bites a septicemic mammalian host and ingests blood infected by Y. pestis. During the blood meal, the infected blood is pulled through the esophagus and the proventriculus (a valve that prevents blood regurgitation but that opens periodically to allow blood ingestion), and then arrives in the midgut (8). Within a few hours of the blood meal, the infection causes the production of a soft bactericidal matrix that traps Y. pestis and entirely fills the lumen of the proventriculus (4). Due to its softness, this “proventricular cast” can be totally or partially dislodged by the following blood meal; a new Y. pestis-containing proventricular cast then forms and fills the lumen of the proventriculus. The new cast can again be partially dislodged and replaced by subsequent feeding (4). During these “cast destruction-reconstruction” cycles, Y. pestis replicates in the cast and produces a polymer of β1-6-N-acetylglucosamine that progressively consolidates the initially soft mass (3, 4, 9, 11). Eventually, the mass becomes so hard that it blocks the proventriculus and prevents the flea from ingesting blood. The so-called “blocked” flea then regurgitates some drawn blood at the biting site and continues to attempt to feed until it starves to death. However, it is during one of the many unsuccessful feeding attempts that the blocked flea will regurgitate Y. pestis-containing blood into the host’s dermis and thus inoculates the host with the plague bacillus (12–14).

In 1996, Hinnebusch et al. used fluorescent Y. pestis to demonstrate that the hemin storage (hmsHFRS) gene locus is essential for blockage of the flea’s proventriculus (9). It is therefore possible to use microscopy to monitor the progression of foregut colonization by fluorescent Y. pestis. Indeed, we have previously acquired fluorescence microscopy images of the proventriculus from infected fleas and have classify them according to the approximate number of bacteria present (i.e., none, very few, few, or many) (4). Next, we developed an algorithm that analyzed the fluorescence microscopy images and measured the surface area occupied by fluorescent bacteria within each individual proventriculus (15, 16). Although this computerized approach is more accurate than the visual method, it still provides only a rough estimation of the state of proventriculus colonization; only the area occupied by the bacteria is taken into account, and the technique does not consider the fluorescence signal’s intensity in the various part of the proventriculus. We have therefore developed a semiautomatic method that estimates (i) the percentage of the proventriculus infected, (ii) the percentage of the proventriculus surface occupied by bacteria, and (iii) the change over time in bacterial density (i.e., growth or not) during the proventriculus infection. We used the new method to gain insights into the mechanisms responsible for the colonization and blockage of the flea proventriculus by Y. pestis.

RESULTS AND DISCUSSION

Choice of the filter set used to acquire fluorescence microscopy images of the proventriculus.

The proventriculus is covered with inward-pointing spines that auto-fluoresce at all wavelengths of the visible spectrum (Fig. 1A). Despite the interference from auto-fluorescence, it is possible to discern Y. pestis expressing the green fluorescent protein (GFP) from Aequorea coerulescens (excitation max = 475nm, emission max = 505 nm) located in the proventriculus (Fig. 1B). We used a raster graphics editor to make the bacteria appear in blue while conserving the green color of the proventricular spines (4, 15). To avoid this image processing step (which is a source of bias), it is possible to use a long-pass B2-A cube filter (Nikon) to expand the absorption window for fluorophores that are excited at 470/40 nm (Fig. 1B) (17). This filter transmitted all wavelengths above 500 nm (i.e., green, yellow, and red light), whereas the bandpass GFP cube filter transmitted only wavelengths between 500 and 560 nm (i.e., green). By combining this technical change with a dichromic mirror of 505 nm for B2-A (instead of 500 nm for GFP), almost all of the autofluorescence in the proventriculus appears in orange and the GFP bacteria still fluoresce in green. We therefore selected the B2-A filter and GFP-Y. pestis for our quantification of the surface area occupied by the bacteria in the proventriculus and as the relative bacterial load.

FIG 1.

(A) A fluorescence emission spectrum of an uninfected proventriculus. λ_exc = 488 nm, Δλ_em = 9 nm. (B) Uninfected and Y. pestis-infected proventriculi, observed by confocal light microscopy and a GFP filter or a B2-A filter. Objective: 20×.

Semiautomated quantitative analysis of the proventriculus infected by Y. pestis.

After infection, Y. pestis remains confined to the midgut and/or the proventriculus (4, 8). Therefore, bacteria located in the midgut must be excluded from the acquired image before the signal from proventriculus can be processed. In other words, it is necessary to manually select the proventriculus, which we delineated as the zone between the lower spines (a marker of the proventriculus/midgut interface) and the top of the spines or the edge of a bacterial mass overflowing into the esophagus (if present) (Fig. 2).

FIG 2.

A sketch of the procedure for measuring the surface area of the proventriculus colonized by bacteria and for estimating the bacterial load from the fluorescence intensity (using ImageJ software). Created with BioRender.com.

To determine the surface area occupied by the bacteria (i.e., the area in green), a color threshold is applied in order to remove the orange color (corresponding to the spines of the proventriculus) (Fig. 2). To this end, we defined a hue range (47 to 255, on our microscope and with our settings) and applied it throughout the automatic analysis. Next, a binary white mask (representing the presence of bacteria in green) was generated. Lastly, the surface area of the white mask was calculated as a proportion of the total surface area of the proventriculus. This process was also used to estimate the number of proventriculi infected with bacteria.

We also sought to estimate the bacterial load within the lumen of the proventriculus. To this end, we selected the elements within the chosen hue range, split the three RGB channels of the image, and analyzed the fluorescence intensity on a 32-bit, floating-point grayscale mask (Fig. 2). Lastly, the mean fluorescence intensity (relative to the surface area occupied by the bacteria) was calculated.

The method described above is time-consuming if performed manually: 4 h are needed to analyze 20 proventriculi. We therefore generated an ImageJ macro (see supplemental material) that performed the analysis automatically. ImageJ is a free and public program developed by the National Institutes of Health and can be run on Mac, Windows, and Linux. This automation reduced the time required by a factor of 16 (i.e., 15 min, rather than 4 h). After launching ImageJ (Fig. 2), the user must click on “Plugins” menu, then select “Run.” The software asks to select the macro file then to select the folder in which the images of the proventriculus to be analyzed are stored. Next, the user will be requested to manually select the area of interest (i.e., the area occupied by the spines of the proventriculus, as described above). Once the region of interest has been selected, the percentage of the surface area occupied by the bacteria in the proventriculus and the mean fluorescence intensity were determined automatically. Lastly, all the processed images and measured values are saved in a folder called “mask.”

Colonization by a strain fully competent for flea blockage.

In a proof-of-principle experiment, we studied the course of proventriculus colonization in fleas fed on blood contaminated with a Y. pestis strain known to be fully competent for flea blockage. However, in order to define the limits of detection (i.e., the noise), we first analyzed 20 proventriculi from uninfected fleas. The median (range) values of the noise for “colonized surface area” and “fluorescence intensity” were 0.3% (0.014% to 0.807%) and 1.17 × 10⁷ arbitrary units (AU) (1.04 × 10⁷ AU to 1.31 × 10⁷ AU), respectively (Fig. 3C and D). We then arbitrarily set the thresholds (above which we consider a signal to be positive) to 1% for “colonized surface area” and 1.32 × 10⁷ AU for “fluorescence intensity.”

FIG 3.

Colonization of the X. cheopis flea proventriculus by the wild-type Y. pestis strain, the ΔhmsHFRS strain, and the ΔompR-envZ strain. (A) Fluorescence images of the proventriculus (in orange) infected with the different strains (in green) acquired at 1, 2, 6, 13, and 27 days postinfection (B2-A filter, objective 20×). (B) The percentage of proventriculi colonized (in green) or not (in orange) by Y. pestis. (C) Changes over time in the surface area colonization (%) of the proventriculus and (D) the mean fluorescence intensity (arbitrary units AU) in the proventriculus of fleas that had fed on blood infected with the wild-type strain (in pink), the ΔhmsHFRS strain (in blue) or the ΔompR-envZ strain (in green). The data were obtained from the images shown in (A); each dot corresponds to an individual flea (n = 16 to 20). The data obtained from uninfected proventriculi are also shown (black circles). The dotted line represents the threshold value above which the fluorescence value (AU) indicates the presence of bacteria. The thresholds were set to 1% for the surface area and 1.32 × 10⁷ AU for the fluorescence intensity. Open circles (regardless of the color) represent values below the threshold. The black line represents the median of all the values above the threshold.

Next, we applied our method to images of proventriculi acquired 1, 2, 6, 13, and 27 days (D) after infection (Fig. 3). A surface area analysis showed that all fleas were colonized on D1, D2, and D6 but that one was no longer colonized on D13 and D27 (Fig. 3B). The data revealed that approximately half of the surface area of the proventriculus was colonized by bacteria within 2 days of the infection (median: 51% on D1, and 65% on D2) (Fig. 3A and C). At the end of the first week of infection and on D13 postinfection, ~90% of the surface area of the proventriculus had been colonized by Y. pestis. After 27 days of infection, the colonized surface area had fallen slightly but was still high in absolute terms (median: ~70%). As expected, the pattern of surface area occupation by Y. pestis followed that of the fluorescence intensity (Fig. 3C and 3D). The median fluorescence intensity rose from D1 to a maximum value on D6 (5.62 × 10⁷ AU), which persisted until at least D13 and then started to decrease slightly. However, the fluorescence intensity was still high on D27 postinfection. This decrease in fluorescence intensity was not related to a loss of the GFP-expressing plasmid. Indeed, the plasmid pAcGFP1 is highly stable for at least 4 weeks in the absence of any antibiotic selection pressure (B. J. Hinnebusch, personal communication).

The results of the semiautomatic analysis were strongly correlation with those obtained by eye (R² = 0.9882), except for D1 postinfection (Table 1). In fact, the fluorescence intensity strongly underestimated the number of infected proventriculi, which was not the case when considering the surface area. Therefore, we could only use surface area values to measure the number of infected proventriculi throughout the experiment. However, it should be borne in mind that the surface area analysis might have slightly overestimated the number of infected proventriculi (16 out of 16), relative to a by-eye analysis (14 out of 16). This overestimation might conceivably have been due to background noise from the proventriculus spines. However, the automatic analysis of 20 additional uninfected proventriculi showed that none was colonized. Hence, the computer-based image analysis might have detected a few bacteria that could not be detected by eye, due to their particular orientation. Taken as a whole, these results indicate that (i) our semiautomatic method provides reliable data throughout the infection, and (ii) a quick comparison of the data obtained automatically and those validated by eye is nevertheless advisable on D1 postinfection.

TABLE 1.

The number of fleas colonized, according to each of the three approaches

Day postinfection	Fleas studied (n)	No. of fleas colonized, according to:
		By eye	Surface area occupied	Fluorescence intensity
1	16	14	16	10
2	20	20	20	19
6	20	20	20	20
13	20	19	19	19
27	20	19	19	18

We found that the bacterial load in the proventriculus peaked at D6 postinfection (Fig. 3). Interestingly, we had reported previously that the bacterial load in the flea as a whole (i.e., in the proventriculus and the midgut) peaked on D1 postinfection (4, 5)—suggesting that Y. pestis colonizes the proventriculus more slowly than the midgut. In turn, slower colonization might indicate that the proventriculus is a more hostile environment than the midgut. Alternatively, the slower colonization might be due to a “flushing” effect: during the first week of infection, the bacteria located in the proventriculus would be more easily flushed out by the act of feeding than those located in the midgut; feeding occurred once (on D2 postinfection) in both experiments. This “flushing” hypothesis is consistent with the fact that (i) the Y. pestis-containing proventricular cast anchored to the proventriculus is partially or totally flushed out by the act of feeding, and (ii) all or most of the cast released during the act of feeding accumulates in the midgut because it too large to be excreted in the feces (4).

Although the proventriculus appears to be colonized more slowly than the midgut during the first week of infection, this is not the case for the first 2 days of infection (i.e., between D1 of infection and the first uninfected meal on D2). Indeed, bacterial loads in the whole gut and in the proventriculus were, respectively, lower and greater just before the meal on D2 than immediately after infection (16) (Fig. 3D). This difference may be due to the digestion process, which involves the release of enzymes, nutrient-sequestering molecules, and bactericidal molecules into the midgut (8, 15, 16, 18). In other words, digestion might generate a bactericidal environment in the midgut. Furthermore, only Y. pestis present in the midgut will be excreted in the feces. This is possible because (i) Y. pestis does not bind to cells in the midgut, and (ii) although fleas are able to defecate for 2 days postinfection, the Y. pestis entrapped in the cast remained firmly attached unless blood is drawn.

Further analysis of images of infected proventriculi collected at the onset of infection (on D1) revealed that the few bacteria present are consistently located in the anterior half region of the proventriculus and not in the posterior half (Fig. 3A). Once bacteria are present in both the anterior and posterior halves of the proventriculus (from D2 onwards), the former region always contains a much larger number of bacteria (Fig. 3A). The apparent anterior-to-posterior progression of bacterial colonization of the proventriculus, the rising bacterial load in the proventriculus, and the decreasing bacterial load in the midgut suggest that the proventriculus is colonized by bacteria trapped in this organ soon after infection rather than those returning from the midgut. This hypothesis is consistent with a previous model in which the anterior half of the anterior region of the proventriculus is the initial site of infection; this leads to recurrent colonization and, ultimately, blockage of the proventriculus (4). However, we cannot yet fully rule out a midgut origin for the bacteria present in the anterior half of the proventriculus at the start of the infection and there intend to carry out further (nontrivial) experiments addressing this question.

Colonization by the ΔhmsHFRS strain of Y. pestis.

Although the hmsHFRS operon is essential for flea blockage (9), the literature data indicate that a ΔhmsHFRS mutant can persist in the proventriculus for at least 13 days postinfection (4). We therefore used our method to study the progression of the infection and the status of proventriculi infected with a ΔhmsHFRS strain (relative to the wild-type strain), in order to confirm the literature data, extend the observation beyond D13 postinfection, and thus, better characterize the hmsHFRS operon’s role in the colonization of the proventriculus.

We found that the ΔhmsHFRS mutant colonizes and persists in the proventriculus up to 27 days postinfection (Fig. 3). However, the median colonized surface area of the proventriculus was about half that seen in the wild-type strain (~45%) on D6 postinfection and remained so until D27 postinfection (Fig. 3C). The median fluorescence signal intensity remained very low throughout the experiment (Fig. 3D). Furthermore, 21% and 55% of the proventriculi collected, respectively, on D2 and D27 postinfection apparently contained no bacteria (Fig. 3A and B), suggesting that the mutant is slowly flushed out of the proventriculus. Taken as a whole, our results show that the hmsHFRS operon is not essential for infecting the proventriculus but enables the maintenance of Y. pestis in the proventriculus for a long period in a large population of insects. This is because the hmsHFRS⁺ strain of Y. pestis is firmly anchored to the proventriculus, thanks to the synthesis of a β-N-acetylglucosamine polymer; the latter consolidates the mass and traps the bacteria in the proventriculus (4, 9). Therefore, the hmsHFRS operon’s role is to increase the incidence of both proventriculus infection and flea blockage. In this context, one can hypothesize that hmsHFRS enables Y. pestis to use the flea as a long-term reservoir for transmission.

Interestingly, several proventriculi were colonized by the hmsHFRS mutant to much the same extent (in terms of the percentage surface area and fluorescence intensity) as by the wild-type strain on D6, D13, and D27 postinfection (Fig. 3). This observation raises the question of whether the mutant could block a very low proportion and/or be transmitted more than a month after infection. To answer this question, one would need to study a very large population and measure (i) the blockage rate; (ii) the colonization of the proventriculus (using the fluorescence method); (iii) the bacterial load in the proventriculus before and after feeding; and (iv) the regurgitation of bacteria. However, this type of study is far from trivial and requires the sacrifice of fleas; this might bias the results by the removal of nonblocked insects that might have become blocked later in the experiment.

Colonization by the ΔompR-envZ strain of Y. pestis.

After applying our method to a mutant unable to block fleas, we studied a strain that has a partial impact on the flea blockage rate. We focused on the OmpR-EnvZ phosphorylated regulatory system because it activates the transcription of genes required for flea blockage—presumably after having sensed nutrient depletion (15). Therefore, we analyzed the proventriculi of fleas infected with the ΔompR-envZ and the wild-type strains in the same infection experiment (Fig. 3).

The results showed that the median surface area of the proventriculus colonized by the ΔompR-envZ mutant was 20% on D1 postinfection and peaked at 40% on D13 postinfection (Fig. 3C). Furthermore, the maximum surface area occupied by the mutant in some proventriculi was consistently lower than that seen for the wild-type strain (~70% versus ~95%, respectively) at the time points when several fleas are diagnosed as blocked (i.e., D6, D13, and D27 postinfection). Consistently, the median fluorescence intensity was very low at all postinfection time points (Fig. 3D). It is noteworthy that the fluorescence intensity values in two proventriculi collected on D6, D13, and D27 postinfection were similar to those measured with the wild-type strain, although the surface area of the proventriculus occupied by the mutant strain did not reach that observed with the wild-type strain (Fig. 3C and D). Taken as a whole, our present data and our previous findings (showing that the proportion of fleas blocked by the ΔompR-envZ mutant is half that of the wild-type strain) suggest that colonization of the entire surface area of the proventriculus and a high bacterial load in the proventriculus are not prerequisites for flea gut blockage. Therefore, the consolidation of the mass that entraps Y. pestis soon after infection in the proventriculus might be more important for flea blockage (and therefore the transmission of Y. pestis) than having a high bacterial load in the entire proventriculus. Furthermore, the anterior part of the proventriculus (i.e., close to the esophagus) is certainly the most important site for consolidation and blockage, because the ΔompR-envZ mutant is mostly located there (Fig. 3A). Lastly, given that the wild-type strain blocks more fleas, colonizes a greater surface area of the proventriculus, and results in a high bacterial load in the proventriculus (relative to the ΔompR-envZ mutant), one can hypothesize that full colonization of the proventriculus maximize Y. pestis’ chances of blocking the flea.

Conclusion.

We developed a user-friendly, semiautomated method for studying the infection of the flea proventriculus by Y. pestis. The method is based on the analysis of fluorescent images but does not require any empirical analyses and/or manipulation of the images. It quantifies the number of proventriculi colonized by Y. pestis, the colonized surface area in each individual proventriculus, and the estimated bacterial growth. Previous studies showed that images of proventriculus of Ctenocephalides felis, Oropsylla montana, and Oropsylla hirsuta infected with Y. pestis GFP strain, using a B2-A filter, display the same color as with X. cheopis (17, 19, 20). In other words, our method will be useful to compare colonization by Y. pestis of different flea species. We also expect this tool to greatly facilitate the characterization of the roles of genes required for flea blockage and to better bridge in vitro experiments with in vivo flea infection experiments.

The implementation of our method with X. cheopis fed on mouse blood contaminated with the wild-type strain, a ΔhmsHFRS strain, or a ΔompR-envZ strain provided further insights into how Y. pestis produces a transmissible infection in fleas. In particular, our results suggest that the infectious process leading to flea blockage results primarily from Y. pestis entrapment and replication in a cast in the anterior part of the proventriculus, in conjunction with consolidation of this mass early after infection. The flea blockage rate appears to be increased by extension of the colonization from the anterior part of the proventriculus to the posterior part, with the lumen being completely filled by the consolidated mass.

Lastly, our data indicate that Y. pestis’ ability to consolidate the mass in the proventriculus and thus block fleas is not a prerequisite for colonization of this organ. However, this consolidation allows Y. pestis to maintain itself in a larger population of fleas for a longer period of time. In other words, the consolidation of the proventricular mass by Y. pestis not only increases the likelihood of bacterial transmission via gut blockage but also constitutes a long-term reservoir for the bacteria.

MATERIALS AND METHODS

Ethics statement.

All experiments complied with the national and institutional regulations and ethical guidelines, and all procedures were approved by the regional animal care and use committee.

Strains.

The bacterial strains used in this study were the wild-type Kim 6+ strain (21), the ΔompR-envZ::dfrB_#2 mutant (15), and the ΔhmsHFRS::aphA3’ mutant (4) harboring the pAcGFP1 plasmid (TaKaRa Bio Compagny, Mountain View, CA). The pAcGFP1 plasmid is stable in Y. pestis regardless of the genetic background according the following protocol: populations were propagated by 1% serial transfer every 24 h for a total of three transfers. After dilution and plating onto LB Agar, 60 single colonies per populations were patched on LB supplemented with and without carbenicillin. All the colonies grew on both media. Bacteria were cultured without shaking in brain heart infusion broth (211059, BD Difco, Franklin Lakes, NJ) supplemented with carbenicillin (100 μg/mL, to ensure the retention of the pAcGFP1 plasmid).

Flea rearing and infection.

Oriental rat fleas (Xenopsylla cheopis) were allowed to feed on heparinized blood from OF1 female mice (Charles River Laboratories, Saint Germain Nuelles, France) containing 5 × 10⁸ bacteria/mL (grown overnight first at 28°C and then at 37°C in brain heart infusion broth), using an artificial feeding apparatus (13, 22). One hour later, cohorts of fleas were collected and disposed on a chill table under a dissecting microscope. Only those that had taken a blood meal (i.e., confirmed by the presence of fresh, red blood in the flea’s foregut) were selected for use in the study. Fleas were maintained at 21°C and 75% relative humidity and were fed on D2, D6, D9, D13, D16, D20, and D23 postinfection.

Imaging of GFP-tagged bacteria during proventriculus colonization.

The guts from 16 to 20 randomly selected fleas were immediately mounted in dH₂O between a glass slide and a glass coverslip (thickness: 1.5 mm) after dissection, and then observed under an Eclipse CiS microscope through a 20× objective (Plan fluor 20×, NA 0.5). The GFP (Ex. 470/40, Em. 535/50), and B2-A (Ex. 470/40, Em. 500) fluorescent filter cubes (Nikon, Melville, NY) were used. Bright-field and fluorescence images of the gut were acquired with a Sight DS-F1c camera (Nikon). Fluorescence images were analyzed using ImageJ software (23).

Spectra.

The spectrum was scanned with a laser scanning confocal microscope (Zeiss AxioobserverZ1, LSM880) with a 40× objective (EC Plan Neofluar 40×, NA 1.3, oil immersion) and the Quasar Detector. In lambda mode, we acquired 22 successive intensity images within a spectral bandwidth of 9 nm each, ranging from 504 nm to 690 nm. Excitation at 488 nm was provided by an argon ion laser.

Statistical analysis.

One-way analyses of variance with Tukey’s posttest for multiple comparisons were performed with Prism 9 software (GraphPad Software, La Jolla, CA).