Male Age and Wolbachia Dynamics: Investigating How Fast and Why Bacterial Densities and Cytoplasmic Incompatibility Strengths Vary

ABSTRACT Endosymbionts can influence host reproduction and fitness to favor their maternal transmission. For example, endosymbiotic Wolbachia bacteria often cause cytoplasmic incompatibility (CI) that kills uninfected embryos fertilized by Wolbachia-modified sperm. Infected females can rescue CI, providing them a relative fitness advantage. Wolbachia-induced CI strength varies widely and tends to decrease as host males age. Since strong CI drives Wolbachia to high equilibrium frequencies, understanding how fast and why CI strength declines with male age is crucial to explaining age-dependent CI’s influence on Wolbachia prevalence. Here, we investigate if Wolbachia densities and/or CI gene (cif) expression covary with CI-strength variation and explore covariates of age-dependent Wolbachia-density variation in two classic CI systems. wRi CI strength decreases slowly with Drosophila simulans male age (6%/day), but wMel CI strength decreases very rapidly (19%/day), yielding statistically insignificant CI after only 3 days of Drosophila melanogaster adult emergence. Wolbachia densities and cif expression in testes decrease as wRi-infected males age, but both surprisingly increase as wMel-infected males age, and CI strength declines. We then tested if phage lysis, Octomom copy number (which impacts wMel density), or host immune expression covary with age-dependent wMel densities. Only host immune expression correlated with density. Together, our results identify how fast CI strength declines with male age in two model systems and reveal unique relationships between male age, Wolbachia densities, cif expression, and host immunity. We discuss new hypotheses about the basis of age-dependent CI strength and its contributions to Wolbachia prevalence.

. Lytic phages form particles that burst through the bacterial cell membrane, killing the bacterial host. The phage density model proposes that as phage densities increase, Wolbachia densities decrease (Fig. 1B) (53). Temperature-induced phage lysis covaries with lower Wolbachia densities and CI strength in some parasitoid wasps (53,66), though it is unknown if phage lysis influences Wolbachia densities in any other systems. Second, wMel Wolbachia have a unique ampliconic gene region composed of eight genes termed "Octomom" (71)(72)(73)(74)(75). Octomom copy number varies among wMelCS and wMelPop Wolbachia between host generations and positively covaries with Wolbachia densities (Fig. 1B), but the effects of Octomom-dependent Wolbachia densities on CI have not been investigated. Third, theory predicts that selection favors the evolution of host suppressors (6), as observed for male killing (76)(77)(78)(79). Indeed, CI strength varies considerably across host backgrounds (24,29,39,(80)(81)(82), supporting a role for host genotype in CI-strength variation. The genetic underpinnings and mechanistic consequences of host suppression remain unknown, but two models have been proposed (2). The defensive model suggests that host CI targets diverge to prevent interaction with cif products, and the offensive model suggests that host products directly interfere with Wolbachia density or the proper expression of cif products (e.g., through immune regulation) (Fig. 1B). Only a taxon-restricted gene of Nasonia wasps and host transcriptional activity in Drosophila have been functionally determined to contribute to Wolbachia-density variation (83,84); thus, considerable work is necessary to uncover host determinants of variation in Wolbachia density. Since Wolbachia densities significantly contribute to several phenotypes (54,85), investigation of the causes of Wolbachia-density variation is sorely needed.
Here, we investigate rates of CI decline with male age and its mechanistic underpinnings in two classic Wolbachia CI systems, wRi and wMel (25,28,32). These Wolbachia bacteria diverged up to 6 million years ago and have unique cif repertoires (60,63). We demonstrate that relative to wRi, wMel-induced CI strength declines more than three times faster, disappearing in a matter of days. We provide the first direct test of the cifexpression hypothesis in either system and the highest-resolution investigation of Wolbachia-density variation across ages to date. Our results suggest that Wolbachia density and cif expression in full-testes extracts cannot explain age-dependent CIstrength relationships across Wolbachia-host associations and motivate future work to investigate how host immunity could contribute to age-dependent Wolbachia densities. We discuss how these data inform our understanding of the causes of CIstrength variation, Wolbachia-density variation, and the consequences for Wolbachia prevalence in nature.

RESULTS
How much does CI strength vary with age? CI manifests as embryonic lethality (Fig. 1A). As such, we measured CI strength as the percentage of embryos that hatch from a mating pair's clutch of offspring; high compatibility corresponds with high hatching. Our experiments used males of different ages to test the impact of male age on CI strength. Here, we defined age as days since eclosion, where males paired with females the day they eclosed were considered 0 days old. For wMel, we measured CI strength daily across the first 3 days of male age ( Fig. 2A) and separately every 2 days Male Age and Wolbachia Dynamics ® across the first 8 days of male age (Fig. 2B). This design enabled us to determine the rate of CI decline and the ages where males no longer cause significant CI. Crossing uninfected D. melanogaster females and males yielded high levels of compatibility ( Fig. 2A; 95% confidence interval of the mean = 74 to 93%). Young 0-day-old wMelinfected males induced strong CI when mated with uninfected females (95% interval = 9 to 27%). wMel-infected females significantly rescued CI caused by infected 0-day-old males (95% interval = 87 to 92%, P = 1.74E-12). Crosses using older 1-(95% interval = 31 to 51%), 2-(95% interval = 53 to 73%), and 3-day-old (95% interval = 69 to 83%) Male age is displayed to the right of the corresponding sex symbol. CI crosses are colored red, rescue crosses are purple, and uninfected crosses are green. Boxplots represent median and interquartile ranges. Letters to the right represent statistically significant differences based on a = 0.05 calculated by Dunn's test with correction for multiple comparisons between all groups; crosses that do not share a letter are significantly different. P values are reported in Table S1. These data demonstrate that CI strength decreases with age in two Wolbachia-host associations and more slowly in wRi-infected D. simulans.
What causes CI strength to vary with age? The bacterial-density and CI-gene-expression hypotheses are both proposed to explain CI-strength variation. These hypotheses predict that Wolbachia density and/or cif expression positively covary with CI strength. To elucidate the causes of declining CI strength with male age, we tested both hypotheses in the context of rapidly declining wMel CI strength and more slowly declining wRi CI strength.
Bacterial density differentially covaries with age between species. We tested the bacterial density hypothesis by dissecting testes from siblings of flies used in our CI assays described above, extracting DNA, and measuring the relative abundance of a single-copy Wolbachia gene (ftsZ) relative to a single-copy ultraconserved element (UCE) (96) of Drosophila via quantitative PCR (qPCR). We selected a random infected sample from the 0-day-old age group as the reference for all fold change analyses within each experiment. We report all qPCR data as fold change relative to this control. Surprisingly, 0-day-old D. melanogaster testes had low wMel density ( Fig. 3A; 95% interval = 0.53-to 1.01-fold change), and older 2-(95% interval = 0.92 to 1.11), 4-(95% interval = 0.96 to 1.72), 6-(95% interval = 1.17 to 1.49), and 8-day-old (95% interval = 1.19 to 1.51) infected testes had progressively higher wMel densities (Fig. 3A). wMel densities were significantly different among age groups according to a Kruskal-Wallis test ( Fig. 3A; P = 1.1E-03). To test for a correlation between wMel densities and CI strength, we performed Pearson (r p ) and Spearman (r s ) correlations on the relationship between wMel fold change against median hatch rates from the associated age groups. wMel densities were significantly positively correlated with increasing compatibility (Table S3; r p = 0.75, P = 5.5E-06; r s = 0.77, P = 2.3E-06). wMel densities also covaried with age ( Fig. S1A; P = 0.02) and correlated with increasing compatibility (Table S3; r p = 0.64, P = 7.7E-04; r s = 0.64, P = 7.4E-04) in the younger 0-, 1-, 2-, and 3-day-old D. melanogaster age group. This result was contrary to our prediction that higher wMel densities would be correlated with stronger CI and lower compatibility.
In conclusion, these data fail to support the bacterial density hypothesis for age-dependent CI-strength variation in wMel-infected D. melanogaster but support the hypothesis in wRi-infected D. simulans. Thus, Wolbachia densities from full-testis extracts cannot explain age-dependent CI across Wolbachia-host associations, suggesting that other factors contribute to these patterns.
cif expression varies with age, but the direction differs between strains. cif expression is hypothesized to control CI-strength variation within Wolbachia-host associations (2,60). cif loci are classified into five different phylogenetic clades called "types" (60,(97)(98)(99). wMel has a single pair of type I cifs, and wRi has two identical pairs closely related to the wMel copy plus a divergent type 2 pair (60). We investigated three questions regarding cif expression. First, does cif expression change relative to the host as males age? We expected that cif expression per host cell would be the key determinant of CI-strength variation. To test this, we used reverse transcriptase quantitative PCR (RT-qPCR) to measure the transcript abundance of cifA and cifB and compared their expression to b spectrin (bspec), a Drosophila membrane protein with invariable expression with age (see Materials and Methods for details). Second, does cif expression decrease relative to Wolbachia as males age? Since wMel densities increase with male age, wMel would need to express cif wMel[T1] at lower levels in older males to allow cif wMel[T1] to decrease relative to the host. Finally, does cifA expression change relative to cifB as males age? Evidence of differential localization of cif loci that covaries with age might indicate more complex determinants of age-dependent CI based on the relative abundance of these products.
We started by investigating these questions in wMel-infected D. melanogaster. Contrary to our first prediction, the relative expression of cifA wMel[T1] to D. melanogaster bspec was lowest in 0-day-old infected males (95% = 1.1 to 1.6) and consistently increased in 2-(95% interval = 1.5 to 3.2), 4-(95% interval = 1.9 to 2.3), 6-(95% interval = 2.1 to 2.8), and 8-day-old (95% interval = 0.9 to 3.8) testes (Fig. 4A). The relative FIG 3 Testing the bacterial density model for CI-strength variation. (A and B) Fold change in testes across male age for the relative expression of (A) wMel ftsZ to D. melanogaster UCE and (B) wRi ftsZ to D. simulans UCE. Letters above the data represent statistically significant differences based on a = 0.05 calculated by Dunn's test with correction for multiple comparisons between all groups; crosses that do not share a letter are significantly different. The fold change was calculated as 2 -DDCq . We selected a random infected sample from the youngest 0-day-old age group as the reference for all fold change analyses within each experiment. P values are reported in Table S1. These data demonstrate that Wolbachia density differentially covaries with age between Wolbachia-host associations.
Next, we investigated our second question. Does cif wMel[T1] expression vary relative to Wolbachia as males age? Indeed, relative expression of cifA wMel[T1] to wMel ftsZ was highest in 0-day-old infected D. melanogaster testes (95% interval = 0.9 to 1.1) and consistently decreased in 2-(95% interval = 0.7 to 0.8), 4-(95% interval = 0.7 to 0.9), 6-(95% interval = 0.6 to 0.7), and 8-day-old (95% interval = 0.4 to 0.9) testes (Fig. 4B). Relative expression of cifA wMel[T1] to wMel ftsZ significantly varied with age (P = 2.9E-03) and was significantly correlated with increasing compatibility (Table S3; r p = 20.8, P = 4.0E-07; r s = 20.7, P = 3.5E-05). Relative expression of cifB wMel[T1] to wMel ftsZ did not significantly covary with age ( Fig. S2B; P = 0.3) but was significantly correlated with increasing compatibility (Table S3; r p = 20.42, P = 3.7E-02; r s = 20.46, P = 2.2E-02). In summary, cif wMel[T1] expression decreased relative to a Wolbachia housekeeping gene . Letters above the data represent statistically significant differences based on a = 0.05 calculated by Dunn's test with correction for multiple comparisons between all groups; crosses that do not share a letter are significantly different. The fold change was calculated as 2 -DDCq . We selected a random infected sample from the youngest 0-day-old age group as the reference for all fold change analyses within each experiment. P values are reported in Table S1. These data demonstrate that age-dependent cif expression is variably related to host expression, cif wMel[T1] expression decreases per Wolbachia with age, and cifA/B relative expression only marginally decreases with age in both systems.
Male Age and Wolbachia Dynamics ® with age, consistent with prior reports that wMel expression of cifA wMel[T1] and cifB wMel[T1] decrease as males age (60). However, since cif wMel[T1] expression did not decrease relative to the host with age, we conclude that the decrease in cif wMel[T1] expression per Wolbachia is insufficient to overcome the increase in cif wMel[T1] expression caused by increased wMel density in full-testes extracts.
Finally, we tested if the relative expression of cifA wMel[T1] to cifB wMel[T1] varied with age. Intriguingly, cifA/B wMel[T1] relative expression did not significantly covary with age ( Fig. 4C; P = 0.09) but was positively correlated with decreasing CI strength (Table S3; r p = 20.61, P = 1.3E-03; r s = 20.46, P = 0.021). In summary, these data suggest that cif wMel[T1] expression per wMel decreases as males age, that cifA wMel[T1] expression decreases marginally faster than cifB wMel [T1] , and that overall cif wMel[T1] expression increases relative to the host as males age and CI strength decreases. This is the first report that CI strength is decoupled from Wolbachia densities and cif expression in testes.
Next, we investigated the cif-expression hypotheses in wRi. We predicted that cif wRi[T1] and/or cif wRi[T2] expression would decrease relative to host expression. Since wRi density decreased with age, cif expression per wRi would not need to change to accomplish this shift in relative expression. As predicted, relative expression of cifA wRi[T1] to D. simulans bspec was highest in infected 0-day-old (95% interval = 0.7 to 1.7) testes and declined in 4-(95% interval = 0.1 to 0.4), 8-(95% interval = 0.3 to 0.7), and 12-day-old (95% interval = 0.2 to 0.3) testes (Fig. 4D). Relative expression of cifA wRi[T1] to D. simulans bspec significantly covaried with age (P = 1.2E-03) and was significantly correlated with decreasing CI strength (Table S3;  to D. simulans bspec also decreased with age, and each was significantly correlated with decreasing CI strength (Table S3). These results support the cif-expression hypothesis for age-dependent CI in wRi.
Finally, as with wMel, we investigated the relationship between cifA and cifB expression in wRi across age and found similar results, where cifA wRi[T1] expression relative to cifB wRi[T1] expression did not significantly vary with male age ( Fig. 4F; P = 0.2) but did significantly correlate with increasing compatibility (Table S3; r p = 20.44, P = 0.045; r s = 20.46, P = 0.035). Relative expression of cifA wRi[T1] to cifA wRi[T2] expression did not covary with age ( Fig. S3G; P = 0.6) or increasing compatibility (Table S3; S3K; P = 1.7E-03) expression with male age; D. simulans bspec C q did not vary with age ( Fig. S3L; P = 0.6), and wRi ftsZ C q significantly increased with age ( Fig. S3M; P = 8.9E-04). In summary, cif wRi expression significantly decreased with age in wRi testes, cifA wRi[T1] expression decreased marginally faster than cifB wRi[T1] expression, and there was a small decrease in cifA wRi[T1] expression relative to wRi, but other cif wRi loci do not follow similar trends.
In conclusion, we found that wMel cif expression did not explain age-dependent CI-strength variation. More specifically, wMel's expression of cif genes decreased with age (60), relative wMel and wRi cifA-to-cifB expression varied marginally with age, and cif expression dynamics varied considerably across male age and differed between wMel-and wRi-infected hosts.
What causes Wolbachia density to vary with age? We found that Wolbachia densities from full-testes extracts significantly increased with male age in wMel-infected D. melanogaster and significantly decreased with male age in wRi-infected D. simulans. The causes of age-dependent Wolbachia-density variation have not been explored. We tested three hypotheses. Namely, that phage lytic activity, Octomom copy number, or host immune expression may govern age-dependent Wolbachia densities.
Phage density does not covary with age-dependent Wolbachia density. The phage density hypothesis predicts that Wolbachia density negatively covaries with phage lytic activity (53). Since phage lysis corresponds with increased phage copy number (53,66), we tested the phage density model by measuring the relative abundance of phage to Wolbachia ftsZ using qPCR. wMel and wRi each harbor a unique set of phage haplotypes; wMel has two phages (WOMelA and WOMelB), and wRi has four (WORiA to -C, WORiB is duplicated) (100). At least one of wMel's phages is capable of particle production, but it is unknown if wRi's phages yield viral particles (70). We monitored WOMelA and WOMelB of wMel simultaneously using primers that target homologs present in a single copy in each phage. Conversely, we monitored WORiA, WORiB, and WORiC separately since shared homologs are too diverged to make suitable qPCR primers that match multiple phage haplotypes.
First, we evaluated the phage density model for wMel. We predicted the relative abundance of WOMelA/B to decrease with D. melanogaster male age since wMel density increases with age. However, there was no change in WOMelA/B abundance relative to wMel ftsZ as males aged ( Fig. 5A; P = 0.3), while WOMelA/B abundance relative to FIG 5 Testing the phage density, Octomom, and host immunity hypotheses for age-dependent Wolbachia-density variation. (A to F) Fold change in testes across male age for the relative abundance or expression of (A) WOMelA/B to wMel ftsZ, (B) WORiB to wRi ftsZ, (C) Octomom gene WD0509 to wMel ftsZ, (D) D. melanogaster Rel to Bspec, and (F) D. simulans Rel to Bspec. Correlation between the relative expression of Rel to Bspec and ftsZ to Bspec for (E) wMel and (F) wRi. Letters above the data represent statistically significant differences based on a = 0.05 calculated by Dunn's test with correction for multiple comparisons between all groups; crosses that do not share a letter are significantly different. (E and G) Pearson (top) and Spearman (bottom) correlations are reported. Linear regressions are plotted with 95% confidence intervals. The fold change was calculated as 2 2DDCq . We selected a random infected sample from the youngest 0-day-old age group as the reference for all fold change analyses within each experiment. P values are reported in Table S1. These data demonstrate that age-dependent Wolbachia densities are not controlled by phage WO lysis or Octomom copy number but are correlated with Rel expression in D. melanogaster and less so in D. simulans. D. melanogaster UCE increased similar to wMel density ( Fig. S4A; P = 3.0E-04). Relative phage abundance was not significantly correlated with increasing compatibility (Table S3; r p = 20.065, P = 0.75; r s = 0.17, P = 0.39). Similarly, WOMelA/B significantly varied with age relative to UCE ( Fig. S4B; P = 0.049) but not wMel ftsZ ( Fig. S4C; P = 0.15) in the 0-, 1-, 2-, and 3-day-old age experiment.
Next, we predicted that WORi phage abundance would increase with decreasing wRi densities across D. simulans male age if governed by the phage density model. As with wMel in D. melanogaster, relative WORiB to wRi ftsZ abundance did not significantly covary with male age (Fig. 5B; P = 0.053) or correlate with increasing compatibility (Table S3; r p = 0.032, P = 0.88; r s = 0.12, P = 0.58). Relative WORiB to D. simulans UCE abundance increased with age, similar to wRi density ( Fig. S4D; P = 4.4E-04). Comparably, WORiA ( Fig. S4E; P = 0.3) and WORiC ( Fig. S4F; P = 0.4) abundance relative to wRi did not vary with male age. These data suggest that phage WO is unrelated to age-dependent Wolbachia-density variation in wMel and wRi.
Octomom does not vary with age-dependent wMel density. Only very closely related wMel variants encode all eight Octomom genes (e.g., wMel, wMelCS, wMelPop). The relative abundance of Octomom to Wolbachia genes positively covaries with wMelCS and wMelPop density (71)(72)(73)(74)(75), commonly changing between host generations. A pair of repeat regions flank the Octomom genes and are hypothesized to be involved in Octomom amplification. In wMel, the 39 repeat region has a transposon insertion that likely prevents Octomom amplification (71). As such, we predicted that Octomom copy number would be invariable with age. We tested if Octomom copy number variation correlated with age-dependent wMel density variation using qPCR. Indeed, the relative abundance of an Octomom gene (WD0509) to wMel ftsZ did not covary with male age ( Fig. 5C; P = 0.53) or correlate with increasing compatibility (Table S3; r s = 20.19, P = 0.36; r s = 0.1, P = 0.61). Similar results were observed in 0-, 1-, 2-, and 3-day-old wMel-infected males ( Fig. S1B; Table S3). We conclude that Octomom copy number is unrelated to the age-dependent increase in wMel densities.
Relish expression is positively correlated with age-dependent wMel, but not wRi, densities. Theory predicts that natural selection favors the evolution of host genes that suppress CI (6). Manipulation of Wolbachia densities is one mechanism that may drive CI suppression (2). Since the immune system is designed to control bacterial loads, we investigated the role of the host immune system in Wolbachia-density variation across male age. The immune deficiency (Imd) pathway is broadly involved in defense against Gram-negative bacteria such as Wolbachia (101). Bacteria activate the Imd pathway by interacting with peptidoglycan (PG) recognition proteins, which start a signal cascade that results in the expression of the NF-k B transcription factor Relish (Rel). Relish then activates antimicrobial peptide production. Wolbachia lacks the full suite of genes needed to synthesize PG (102-104) but can express the PG precursor lipid II, which is sufficient to activate the Imd pathway (104,105).
Conversely, relative expression of D. simulans Relish to bspec did not significantly covary with age ( Fig. 5F; P = 0.7) but remained positively correlated with the relative expression of wRi ftsZ to bspec within testes samples according to Pearson, but not Spearman, analyses ( Fig. 5G; r p = 0.55, P = 0.012; r s = 0.13, P = 0.59). In summary, Relish expression is positively correlated with age-dependent wMel densities in D. melanogaster, but less so in wRi-infected D. simulans, supporting a role for the Imd pathway in the regulation of at least wMel density variation. However, more work is necessary to determine if the correlation between age-dependent immune expression and Wolbachia density in testes are causatively associated.

DISCUSSION
Within Wolbachia-host systems, several factors influence CI strength (29,30,37,38,53,66,(86)(87)(88)(89), but male age can be particularly impactful (3,18,27,29). Our results determine how fast and investigate why CI strength declines as males age. First, we estimate that CI strength decreases rapidly for wMel-infected D. melanogaster (19%/day), becoming statistically insignificant when males reach 3 days old. In contrast, wRi causes intense CI that declines more slowly (6%/day), resulting in statistically significant CI through at least the first 12 days of D. simulans male life. Second, Wolbachia densities and cif expression from full-testes extracts increase in wMel-infected D. melanogaster and decrease in wRi-infected D. simulans as males age and CI weakens. These results indicate that bacterial density and CI gene expression in full-testes extracts cannot fully account for age-dependent CI strength across host-Wolbachia associations. Third, while WO phage activity and Octomom copy number cannot explain Wolbachia-density variation, D. melanogaster immune expression covaries with wMel densities, suggesting the host immune system may contribute to age-dependent Wolbachia density in D. melanogaster, but much less so in D. simulans. Notably, the transcript-based data (e.g., cif and Relish) described here are subject to the caveat that mRNA levels may not correlate perfectly with protein expression or activity. Future proteomics analyses will be needed to confirm that these trends hold at the protein level. We discuss how our discoveries inform the basis of age-dependent CI-strength variation, how multiple mechanistic underpinnings may govern age-dependent Wolbachia densities, and how age-dependent CI may contribute to Wolbachia frequency variation observed in nature.
Wolbachia density and CI-gene expression in full-testes extracts do not fully explain age-dependent CI-strength variation. Despite support that CI strength is linked to Wolbachia density and cif expression across and within systems (37, 38, 51-54, 60, 66), our observations add to a growing body of literature suggesting Wolbachia densities in adult testes (30,88) and, for the first time, cif expression, are insufficient to explain CI-strength variation broadly. We discuss three hypotheses to explain the disconnect between Wolbachia density and cif expression in full-testes extracts and CI strength with male age. Note, however, that these results may also be explained by a decoupling of cif transcription and protein translation, which will require future proteomics analyses to investigate.
First, the localization and density of Wolbachia and cif products within specific cells in testes may more accurately predict CI strength. Indeed, the proportion of infected spermatocyte cysts covaries with CI strength in natural and transinfected combinations of CI-inducing Wolbachia and D. melanogaster, D. simulans, D. yakuba, D. teissieri, and D. santomea (51,52). Intriguingly, two wRi-infected D. simulans strains whose Wolbachia cause variable CI did not have different Wolbachia densities according to qPCR, but the number of infected sperm cysts covaried with CI between strains (106). Thus, Wolbachia densities in full-testes extracts may not reflect the cyst infection frequency, but it is unknown how generalizable this discrepancy is across or within Wolbachia-host associations with variable CI strengths. It seems plausible that while wMel densities increase in the testes as males age, the proportion of infected spermatocytes could decrease. Notably, since wMel infections increase drastically as males age, a considerable shift in localization and density dynamics would be necessary. Microscopy assays will be required to test if Wolbachia and cif localization explains wMel age-dependent CI-strength variation.
Second, age-dependent CI may be governed by developmental constraints of CI susceptibility. For instance, the paternal grandmother age effect, where Wolbachiainfected sons of older virgin females cause stronger CI than sons of younger females, covaries with Wolbachia densities in embryos but not in adult males (30). Intriguingly, temperature-sensitive CI-strength variation in Cardinium-infected Encarsia wasps is also decoupled from symbiont densities, but CI strongly correlates with pupal development time (107,108). Cardinium CI effectors likely have more time to interact with host targets at critical stages of pupal development when slowed by cool temperatures, despite lower Cardinium density (107,108). These studies suggest that sperm are modified in spermatogenesis before adult eclosion and that variation in symbiont densities during early development can contribute to CI-strength variation. If modified sperm are primarily produced during pupal or larval development, younger adult males would have a higher proportion of CI-modified sperm in their seminal vesicle than older males since older males continue to produce sperm as adults. Intriguingly, remating seems to weaken CI (86,87), supporting this hypothesis. However, since CI strength decreases faster in D. melanogaster than in D. simulans, this hypothesis predicts that adult D. simulans sperm production is slower and/or CI modification occurs for an extended time. Functional work is necessary to determine if CI modification is developmentally restricted.
Finally, age-dependent CI may be related to the availability of CI-effector targets with male age and not the abundance of cif products. Indeed, the number of genes transcribed by D. melanogaster increases from 7,000 in embryos to over 12,000 in adult males, and nearly a third of genes are not expressed until the 3rd larval instar (109). As adult males age, the number of transcribed genes continues to vary, though less so than during metamorphosis (109). These data support the possibility that host targets of CI may vary in abundance as males age. However, since transgenic cif expression can significantly enhance CI strength above wild-type levels (60), there are circumstances when natural cif expression is not high enough to saturate all targets. It is unknown if similar experimental approaches can strengthen age-dependent CI. More work will be necessary to determine the host genes that modify CI and how those factors vary in expression relative to CI strength.
Age-dependent bacterial density covaries with immune expression, not phage or Octomom. We report a strong relationship between male age and Wolbachia densities that differ between systems; densities decrease in wRi-infected D. simulans and increase in wMel-infected D. melanogaster. Reports of age-dependent variation in Wolbachia densities across age in different tissues and sexes are common (51,71,90,91,(110)(111)(112), but the basis of this variation remains unexplored. We investigated the cause(s) of this variation for the first time. We predicted that genes that covary with age-dependent densities might be causatively linked, although additional experiments will be necessary to confirm this. First, we tested whether phage or Octomom covary with age-dependent Wolbachia densities. Despite prior reports that phage WO of Nasonia and Habrobracon Wolbachia can regulate temperature-dependent Wolbachia densities (53,66) and that Octomom copy number correlates with wMelCS and wMelPop densities (72, 73), we found that neither covaries with age-dependent Wolbachia densities in testes.
Next, we asked whether host immune gene expression correlates with age-dependent Wolbachia densities. We report that Relish expression, which activates antimicrobial peptide (AMP) production in the Imd pathway (101), strongly correlates with wMel densities and is highest when wMel densities are high. This result was surprising since we predicted that immune expression would hinder Wolbachia proliferation if it were correlated. Conversely, Relish does not vary with D. simulans male age and is only very weakly correlated with wRi densities. It is plausible that the correlation between Relish expression and wMel density represents a spurious and noncausative association. Additionally, Relish transcription does not necessarily equate to increased Relish activity and AMP production since endoproteolytic cleavage is necessary to activate the Relish protein (113), but future analysis of AMP expression will elucidate this. However, this correlation may represent a causative link between age-dependent wMel densities and immune expression. We propose two nonexclusive hypotheses to explain this relationship.
First, wMel rapidly proliferates as males age, elicit an immune response proportional to their infection density, but evade the effects of immune activation. Wolbachia synthesize lipid II (102)(103)(104), which is sufficient to activate the Imd pathway (104,105), and increase AMP gene expression when transinfection into novel host backgrounds occurs (114)(115)(116)(117), suggesting that Wolbachia can trigger Imd activity. However, Relish and AMP expression do not vary with Wolbachia infection state (118)(119)(120)(121)(122)(123)(124) or density (84,121) in natural Wolbachia-host associations. It has been proposed that Wolbachia evade the host immune system by residing in host-derived membranes or bacteriocyte-like cells (125,126). Thus, the correlation between Relish expression and wMel density may indicate that wMel triggers the immune system but evades the immune response, preventing its densities from decreasing. Notably, since this hypothesis assumes that wMel densities increase independently of Imd expression, it does not explain why wMel densities increase with age or why age-dependent wMel and wRi densities differ.
Second, age-dependent Imd expression increases independently of Wolbachia but impacts Wolbachia densities. Aging in D. melanogaster is associated with increased expression of AMPs, Relish, and other immune genes (127)(128)(129)(130)(131)(132)(133). Counterintuitively, age also covaries with increased gut microbial loads and Imd activation in D. melanogaster (127)(128)(129)(134)(135)(136). Why gut bacterial loads increase with D. melanogaster age and immune expression remains unknown. However, age-dependent immune expression may damage the epithelium, lead to dysbiosis through differential effects on gut microbial members, alter gut tissue renewal and differentiation, and/or cause cellular inflammation (101,137). In other words, the positive correlation between Relish expression and wMel density may be caused by off-target effects of immune expression on the cellular environment. To our knowledge, we report the first case where endosymbiont densities increase with age-dependent immune expression, suggesting that the cause(s) of age-dependent bacterial proliferation may apply to more than gut microbes. Functional assays, such as Relish knockdowns, will be necessary to causatively link male age-dependent Wolbachia densities and immune expression.
Age-dependent CI strength could contribute to Wolbachia frequency variation in nature. We can consider our estimates of age-dependent CI strength in the context of an idealized discrete-generation model of Wolbachia frequency dynamics first proposed by Hoffmann et al. (3). This model incorporates imperfect maternal transmission (m), Wolbachia effects on host fitness (F), and the proportion of embryos that hatch in a CI cross relative to compatible crosses (H) (3). Across all experiments, CI strength (s h = 1 2 H) progressively decreases as males age (Table S2); wMel CI strength decreases quickly (day 0 s h = 0.860; day 8 s h = 20.007), and wRi CI strength decreases relatively slowly (day 0 s h = 0.991; day 8 s h = 0.244). Small negative values of s h indicate that the CI cross has a slightly higher egg hatch than the compatible crosses.
wRi occurs globally at high and relatively stable infection frequencies, consistent with generally strong CI (4,26), while wMel varies in frequency on several continents. In eastern Australia, wMel frequencies range from ;90% in the tropical north to ;30% in the temperate south (34). While transmission rate variation contributes significantly to clinal wMel frequencies, mathematical modeling suggests clinal differences in CI strength are also likely to contribute (34,152). For example, CI must be essentially nonexistent (s h ,, 0.05) to explain relatively low wMel frequencies observed in temperate Australia, assuming little imperfect transmission (m = 0.01 to 0.026) (138). Conversely, with m = 0.026 and similarly low-to-nonexistent CI (s h # 0.055), large and positive wMel effects on host fitness (F ; 1.3) are required to explain higher wMel frequencies observed in the tropics. Explaining higher tropical frequencies becomes easier with stronger CI (s h . 0.05) or more reliable wMel maternal transmission (m , 0.026) (34).
So, what is wMel CI strength in nature? Field-collected males from near the middle of the Australian cline to the northern tropics cause very weak (s h ; 0.05) to no CI (138). These and other data from the middle of the cline (29) led Kriesner et al. (34) to conjecture that the plausible range of s h in subtropical/tropical Australian populations is s h = 0 to 0.05 but ,0.1. In our study, only 6-(s h = 20.006) and 8-day-old (s h = 20.007) wMel-infected males exhibited CI weaker than s h = 0.1, suggesting that fieldcollected males causing little or no CI (138) are older than 4 days. Interactions among male age, temperature, remating, and other factors likely contribute to weaker CI in younger males (29,37,38,53,66,86,87). Future analyses to disentangle the contributions of male age and other factors to CI-strength variation are sorely needed. These estimates, along with estimates of Wolbachia transmission rate variation across genetic and abiotic contexts (22), are ultimately required to better understand Wolbachia frequency variation in host populations (7,22,24,34,139).
Conclusions. Our results highlight that Wolbachia densities and cif expression from full-testes extracts are insufficient to explain age-dependent CI strength. While age-dependent CI strength in wRi aligns with the bacterial density and CI gene expression hypotheses without the need to consider other factors, wMel CI strength cannot be explained by either of these hypotheses. We propose that localization, development, and/or host genetic variation contribute to this relationship. Moreover, wMel densities increase, and wRi densities decrease, as their respective hosts age. Neither phage WO nor Octomom explain age-dependent Wolbachia density, but variation in these systems covaries with the expression of the immune gene Relish. This represents the first report that the host immune system may contribute to variation in Wolbachia density in a natural Wolbachia-host association. This work motivates an extensive analysis of Wolbachia and cif expression in the context of localization and development and a thorough investigation of the relationship between host genes and Wolbachia density and CI phenotypes. Finally, incorporating the age dependency of CI into future modeling efforts may help improve our ability to explain temporally and spatially variable Wolbachia infection frequencies, as incorporating temperature effects on wMel-like Wolbachia transmission has (22,24,140). Ultimately, this will help explain Wolbachia's status as the most prevalent endosymbiont in nature.

MATERIALS AND METHODS
Fly lines. All fly lines used in this study are listed in Table S4. Uninfected flies were derived via tetracycline treatment in prior studies (16,60). Tetracycline cleared lines were used in experiments over a year after treatment, mitigating the effects of antibiotics on mitochondria (141). We regularly confirmed infection status by using PCR to amplify the Wolbachia surface protein (wsp). An arthropod-specific 28S rDNA was amplified in a separate reaction and served as a control for DNA quality and PCR inhibitors (24,142). The y 1 w 1 D. melanogaster line was confirmed to be wMel infected, as opposed to wMelCS, using IS5-WD1310 primers (143). DNA was extracted for infection checks using a squish buffer protocol. Briefly, flies were homogenized in 50 mL squish buffer per fly (100 mL 1 M Tris-HCl, 0.0372 g EDTA, 0.1461 g NaCl, 90 mL H 2 O, 150 mL proteinase K), incubated at 65°C for 45 min, incubated at 94°C for 4 min, and centrifuged for 2 min, and the supernatant was used immediately for PCR.
Fly care and maintenance. Flies were reared in vials with 10 mL of food made of cornmeal (32.6%), dry corn syrup (32%), malt extract (20.6%), inactive yeast (7.8%), soy flour (4.5%), and agar (2.6%). Fly stocks were maintained at 23°C between experiments. Flies used for virgin collections were reared at 25°C, virgin flies were stored at 25°C, and experiments were performed at 25°C. Flies were always kept on a 12:12 light:dark cycle. Flies were anesthetized using CO 2 for virgin collections and dissections. During hatch-rate assays, flies were mouth aspirated between vials.
Hatch-rate assays. CI manifests as embryonic death. We measured CI as the percentage of embryos that failed to hatch into larvae. Flies used in hatch rates were derived from vials where flies were given ;24 h to lay to control for rearing density (88). In the morning, virgin 6-to 8-day-old females were added individually to vials containing a small ice cream spoon filled with fly food. Spoon fly food was prepared as described above, but with blue food coloring added, 0.1 g extra agar per 100 mL of food, and fresh yeast smeared on top. After 4 to 5 h of acclimation, a single virgin male was added to each vial. The age of virgin males varied by experiment and cross. Paternal grandmother age was not controlled, but paternal grandmothers were nonvirgin when setting up vials for fathers. Since Wolbachia densities associated with older paternal grandmothers are reduced upon mating (30), we do not expect variation in paternal grandmother Wolbachia densities across experiments or conditions. Vials with paired flies were incubated overnight at 25°C. Flies were then aspirated into new vials with a fresh spoon. Vials were incubated for another 24 h before flies were removed via aspirating. Embryos were counted on spoons immediately after flies were removed. After 48 h, the number of remaining unhatched eggs were counted. The percentage of embryos that hatched was then calculated.
Relative abundance assays. Siblings from hatch-rate assays were collected for DNA extractions. Virgin males were anesthetized, and testes were dissected in chilled phosphate-buffered saline (PBS). Five pairs of testes were placed into a single 1.5-mL Eppendorf tube and stored at 280°C until processing. All tissue was collected the day after the hatch-rate setup. Tissue was homogenized using a pestle, and the DNeasy blood and tissue kit (Qiagen) was used to extract and purify DNA.
qPCR was used to measure the relative abundance of host, Wolbachia, phage WO, and Octomom DNA. Samples were tested in triplicate using Powerup SYBR green master mix (Applied Biosystems), which contains a ROX passive reference dye. Primers were designed using Primer3 v2.3.7 in Geneious Prime (144). Host primers target an ultraconserved element (UCE), Mid1, identified previously (96). Phage genes were also identified from prior work (100). Primers for wMel's phages target both WOMelA (WD0288) and WOMelB (WD0634), while those for wRi are unique to a single phage haplotype. WORiA, WORiB, and WORiC were measured with wRi_012460, wRi_005590/wRi_010250, and wRi_006880 primers, respectively. Only wMel has all eight Octomom genes (WD0507 to WD0514) (71). We measured the wMel Octomom copy number using primers targeting WD0509. Primer sequences and PCR conditions are listed in Table S5. The fold difference was calculated as 2 2DDCT for each comparison. A random sample in the youngest age group was selected as the reference.
Gene expression assays. Siblings from hatch-rate assays were collected for RNA extractions. Virgin males were anesthetized, and testes were dissected in chilled RNase-free PBS. Then, 15 pairs of testes were placed into a single 2-mL tube with 200 mL of TRIzol and four 3-mm glass beads. Tissue was kept on ice between dissections. Samples were then homogenized using a TissueLyser II (Qiagen) at 25 Hz for 2 min, centrifuged, and stored at 280°C until processing. All tissue was collected the day after the hatch-rate setup.
Samples were thawed, 200 mL of additional TRIzol was added, and tissue was further homogenized using a TissueLyser II at 25 Hz for 2 min. RNA was extracted using the Direct-Zol RNA miniprep kit (Zymo Research) following the manufacturer's recommendations, but with an extra wash step. On-column DNase treatment was not performed. Instead, the "rigorous" treatment protocol from the DNA-free kit (Ambion) was used to degrade DNA in RNA samples. Samples were confirmed DNA-free using PCR and gel electrophoresis for an arthropod-specific 28S rDNA (24,142). The Qubit RNA HS assay kit (Invitrogen) was used to measure the RNA concentration. Samples within an experiment were diluted to the same concentration. RNA was converted to cDNA using SuperScript IV VILO master mix (Invitrogen) with either 200 ng or 500 ng of total RNA per reaction, depending on the experiment. qRT-PCR was performed using 1 ng of cDNA per reaction using Powerup SYBR green master mix (Applied Biosystems). All samples were tested in triplicate.
Primers for expression included the host reference, Wolbachia reference, cif, and host immune genes. Primers to Drosophila genes for qRT-PCR were selected from FlyPrimerBank (145). Since Drosophila expression patterns change with age (109), a host gene that is invariable with male age was selected to act as a reference gene for relative expression analyses. We selected an invariable gene using the Drosophila Gene Expression Tool (DGET) to retrieve modENCODE gene expression data for ribosome and cytoskeletal genes (146). DGET reports expression as reads per kilobase of transcript, per million mapped reads (RPKM), and included data for adult males 1, 5, and 30 days after eclosion. b-spec (1 day = 81 RPKM, 5 day = 80, 30 day = 79) was selected because it is largely invariable across age. Our results confirm invariable expression across male age ( Fig. S2E and S3L). D. melanogaster and D. simulans are identical across bspec primer binding sequences. All other primers were designed using Primer3 in Geneious Prime (144) and are listed in Table S5. The fold difference was calculated as 2 2DDCT for each comparison. A random sample in the youngest age group was selected as the reference.
Statistical analyses. All statistics were performed in R (147). Hatch-rate, relative-abundance, and expression assays were analyzed using a Kruskal-Wallis test followed by Dunn's test with corrections for multiple comparisons. Kruskal-Wallis and Dunn's P values are reported in Table S1. Correlations between hatch rate and expression or relative abundance measures were performed using Pearson and Spearman correlations in ggpubr (148). Correlation statistics are reported in Table S3. The 95% confidence intervals were calculated using the classic MeanCI function in DescTools (149). The 95% bias-corrected and accelerated (BCa) intervals were calculated using boot.ci in boot (150). Samples with fewer than 10 embryos laid were excluded from hatch-rate analyses. Samples with a C q standard deviation exceeding 0.4 between triplicate measures were excluded from qPCR and qRT-PCR analyses. Figures were created using ggplot2 (151), and figure aesthetics were edited in Affinity Designer v1.8 (Serif Europe, Nottingham, UK).
Data availability. All data are made publicly available in the supplemental material.

ACKNOWLEDGMENTS
We thank Michael Turelli for helpful feedback on our experimental design. We also thank Will Conner for help identifying phage gene targets for qPCR, Mike Hague for