A test for meiotic drive in hybrids between Australian and Timor zebra finches

Abstract Meiotic drivers have been proposed as a potent evolutionary force underlying genetic and phenotypic variation, genome structure, and also speciation. Due to their strong selective advantage, they are expected to rapidly spread through a population despite potentially detrimental effects on organismal fitness. Once fixed, autosomal drivers are cryptic within populations and only become visible in between‐population crosses lacking the driver or corresponding suppressor. However, the assumed ubiquity of meiotic drivers has rarely been assessed in crosses between populations or species. Here we test for meiotic drive in hybrid embryos and offspring of Timor and Australian zebra finches—subspecies that have evolved in isolation for about two million years—using 38,541 informative transmissions of 56 markers linked to either centromeres or distal chromosome ends. We did not find evidence for meiotic driver loci on specific chromosomes. However, we observed a weak overall transmission bias toward Timor alleles at centromeres in females (transmission probability of Australian alleles of 47%, nominal p = 6 × 10–5). While this is in line with the centromere drive theory, it goes against the expectation that the subspecies with the larger effective population size (i.e., the Australian zebra finch) should have evolved the more potent meiotic drivers. We thus caution against interpreting our finding as definite evidence for centromeric drive. Yet, weak centromeric meiotic drivers may be more common than generally anticipated and we encourage further studies that are designed to detect also small effect meiotic drivers.


| 13465
KNIEF Et al. segregation have been observed in a range of organisms (Lindholm et al., 2016). Alleles that lead to deviations from Mendelian segregation are collectively referred to as meiotic drivers (Lindholm et al., 2016). Here, we define them as entire chromatids, chromosomes, or parts thereof, which outcompete their homologs (or corresponding sex chromosome) within a parental individual over access to the next generation. Meiotic drivers are often difficult to distinguish from other pre-or postzygotic processes that might appear to bias fair Mendelian segregation. These usually result from interactions between parental genotypes, such as a genetic conflict among parental individuals (e.g., female control over paternity) or early viability selection (summarized in Knief, Schielzeth, et al., 2015).
Whether meiotic drive can be observed typically depends on the fitness consequences of the driver. If a meiotic driver has no detrimental effects on the fitness of the carrier, the drive allele is expected to spread rapidly to fixation within a population (Traulsen & Reed, 2012). The distortion caused by such a drive allele will then no longer be visible because the nondriving ancestral allele has been driven to extinction. Given the short evolutionary trajectory of such past events, most of the known drivers impose a major cost on individual fitness which prevents rapid elimination of the competing ancestral allele (Burt & Trivers, 2006). For instance, drive alleles are often linked to recessive deleterious mutations causing the death of homozygous offspring in heterozygote-heterozygote pairings (Fishman & Kelly, 2015), leading to retention of the ancestral allele. Some drivers also reduce the number of gametes produced by heterozygous carriers (Sutter & Lindholm, 2015), such that a disadvantage in sperm competition further slows down the spread of the drive allele. The resulting decline in fitness may then favor the evolution of unlinked suppressors of drive, which might be able to restore fair Mendelian segregation, making the previous distortion again invisible ("cryptic drive system"; Frank, 1991;Hurst & Pomiankowski, 1991;Sandler & Novitski, 1957). Overall, meiotic drivers might thus only be transiently active within a population (Meyer et al., 2012).
Cryptic meiotic drivers can be detected by crossing individuals from diverged populations. If a meiotic driver evolved in one, but not in the other population, the driver will be reestablished in the naïve genetic background of hybrid individuals (Fishman & Willis, 2005;Hurst & Werren, 2001). A complicating factor is that other processes may also lead to apparent deviations from fair Mendelian segregation in these crosses. For example, Bateson-Dobzhansky-Muller (BDM) incompatibilities can arise when two or more genes independently accumulate mutations that spread to fixation in isolated populations.
Although these mutated genes function perfectly within the population they first occurred in, they might be malfunctioning in hybrids because of epistatic interactions (Bateson, 1909;Dobzhansky, 1937;Muller, 1942). If these detrimental interactions are additive or dominant, their effects are observable in the F1 hybrids. If they are recessive, these effects will only appear in the second generation after hybridization, not in the F1 hybrids themselves. The heterogametic sex (in birds the female, being ZW) is usually more affected (Haldane's rule; Haldane, 1922), which may lead to biased sex ratios.
Importantly, any such detrimental effect on gamete or offspring viability will result in apparent deviations from Mendelian inheritance when considering only surviving offspring. Hence, to distinguish meiotic drive from BDM incompatibilities, it is essential to monitor the genotypes of all offspring, including embryos that failed to develop, and also sample apparently infertile eggs.
To distinguish between meiotic drivers and other processes that bias transmission ratios, we can make use of the fact that meiotic drivers generally act in a sex-specific manner whereas most of the other processes take place in both sexes (Lindholm et al., 2016).
In females, a single oogonium leads to the formation of one oocyte and three polar bodies. Because the latter represent an evolutionary dead end, a meiotic driver may act by outcompeting its homologous chromosome for inclusion into the oocyte (Axelrod & Hamilton, 1981;Pardo-Manuel de Villena & Sapienza, 2001).
Following Sandler and Novitski (1957), we call this "chromosomal meiotic drive." In contrast, male meiosis is symmetric, because a single primary spermatocyte gives rise to four functional spermatozoa. Consequently, meiotic drivers in males must act postmeiotically, causing disruption or out-performance of those sperm that do not carry the driving allele ("genic meiotic drive"; Pardo-Manuel de Villena & Sapienza, 2001;Sandler & Novitski, 1957).
Genic meiotic drivers depend on specific genes (a drive and a target locus) that are often linked by an inversion and can be localized anywhere in the genome. Well-known examples include the t-complex in mice (Mus musculus; reviewed in Lyttle, 1991), the SD locus in Drosophila melanogaster (reviewed in Lyttle, 1991), the wtf genes in the yeast Schizosaccharomyces pombe (Eickbush et al., 2019), and the recently discovered ORF-system in rice (Oryza meridionalis ;Yu et al., 2018). These systems involve "poison" and "antidote" alleles that first disable all sperm or pollen and subsequently resurrect only those of their own genotype.
In contrast, chromosomal meiotic drive arises as a consequence of specific chromosomal structural elements. During meiosis, the spindle apparatus attaches to all chromosomes at their centromeres and separates the homologous chromosomes (meiosis I) and subsequently sister chromatids (meiosis II) into daughter cells. Provided that the spindle apparatus exhibits a functional polarity that differentiates the oocyte from the polar bodies, length and sequence polymorphisms at or near the centromere might enable some centromeres to preferentially attach to the spindle leading to the oocyte (Pardo-Manuel de Villena & Sapienza, 2001). In birds, the oocyte spindle in meiosis I is positioned close to the egg cortex, perpendicular to the egg surface. The first polar body forms toward the cortical side (Yoshimura et al., 1993), such that the distance to the egg cortex provides spatial information for a driving chromosome (Rutkowska & Badyaev, 2008). Recently, it was shown in mice that centromeres with more minor satellite repeats attract more spindle microtubules (Chmátal et al., 2014;Iwata-Otsubo et al., 2017). It was further shown that egg and polar body spindle are differentially tyrosinated as a result of their distance to the egg cortex, which allows those centromeres with more minor satellite repeats to preferentially attach to the egg spindle . As a result, chromosomes with more repeats were transmitted to around 62%-81% of the oocytes (Akera et al., 2019;Iwata-Otsubo et al., 2017), confirming the idea of female chromosomal meiotic drive at centromeres (Henikoff et al., 2001). Interestingly, the mice strains used by Iwata-Otsubo et al. (2017) and Akera et al. (2019) differed consistently in their number of minor satellite repeats at centromeres across all chromosomes, resulting in genome-wide meiotic drive.
Besides meiotic drive that acts on the centromeres, some organisms like maize (Burt & Trivers, 2006) have also evolved "neocentromeres" that are typically located about 50 cM away from the centromere. Their driving mechanism requires a single crossover between the centromere and the driving neocentromere (Dawe & Hiatt, 2004), such that each chromosome contains one neocentromere in meiosis I. Neocentromeres are then pulled toward the spindle poles ahead of the centromeres by a specialized kinesin (Dawe et al., 2018), such that the neocentromeres end up in the outer cells of a linear tetrad and one of them forms the egg cell (Rhoades, 1952). The kinesin gene is tightly linked to the neocentromere, and both are passed on together (Dawe et al., 2018). Hence, drive may also be expected at sites that are ~50 cM away from the centromere.
Birds are an ideal model to study meiotic drive directly, because all embryos resulting from a specific pairing, that is, every egg that a female produces, can be sampled and investigated, including infertile eggs. Despite this, tests for drive have been conducted in only two bird species so far. Intriguingly, both studies  (Shang et al., 2010). In Australian zebra finches, the driver was located on chromosome Tgu2 and acted in both sexes (Knief, Schielzeth, et al., 2015), but the molecular mechanism requires further study. Many (avian) species show a drastic drop in nucleotide diversity toward putative centromeric regions of almost all chromosomes (Burri et al., 2015;Delmore et al., 2015;Ellegren et al., 2012;Irwin et al., 2016;Laine et al., 2016;Van Doren et al., 2017;Vijay et al., 2017;Weissensteiner et al., 2017). Recombination is usually suppressed 5 to more than 200-fold at centromeres (Rahn & Solari, 1986;reviewed in Talbert & Henikoff, 2010), such that linked selection has a more pronounced effect on nucleotide diversity (Burri, 2017;Cruickshank & Hahn, 2014). Both purifying selection against deleterious alleles (background selection; Charlesworth et al., 1993) and positive selection contribute to linked selection (Cutter & Payseur, 2013). In general, purifying selection is suggested to be more pervasive than positive selection (Burri, 2017), but several flycatcher (Ficedula spp.; Burri et al., 2015) and stonechat species (Saxicola spp.; Van Doren et al., 2017) show an excess of high-frequency derived alleles (low values of Fay & Wu's H) at some of their putative centromeric regions, which indicates that they may be under positive selection (Fay & Wu, 2000). Albeit positive selection on centromeric regions could have many reasons, the signature of positive selection may also be caused by meiotic drive (Cruickshank & Hahn, 2014;Ellegren et al., 2012). Moreover, a second region of very low genetic diversity is found on most Australian zebra finch chromosomes, typically at the "distal end" that is about 50 cM away from the centromere  and these places might potentially evolve neocentromeric function (see above; Meyer et al., 2012).
Here, we investigated the idea that drive systems may evolve but become cryptic due to driver allele fixation or due to suppression of drive. We did this in two steps. First, we crossed two subspecies of zebra finches that have evolved in isolation for about two million years (Balakrishnan & Edwards, 2009): the Australian zebra finch (T. g. castanotis) and Timor zebra finch (T. g. guttata; Figure S1). The former has a remarkably large effective population size, while the latter is genetically much less diverse (Balakrishnan & Edwards, 2009).
Thus, we hypothesized that, in the larger Australian population, selfish de novo mutations will have arisen and will have outcompeted the ancestral nondriving allele more often than in the Timor subspecies because there are more individuals in which such a mutation can arise.
Second, we produced a backcross to the Australian subspecies to monitor the performance of Timor centromeres in a predominantly Australian genetic background. Successful drive might be conditional on the protein machinery that controls the segregation of chromosomes. If the genetic background is essential for whether drive happens, we would expect to see differences between females in the genetically heterogeneous backcross generation (females carry 25% of Timor DNA on average, but this varies among individuals and between chromosomes). Hence, we also tested whether female identity significantly affects segregation ratios. To assay segregation distortion in hybrid females, we traced the genetic ancestry of 56 informative molecular markers in close linkage to centromeres and regions with neocentromeric potential ("distal ends") in the zebra finch genome ) through a four-generations backcross pedigree. As a control, we also estimated transmission ratios in hybrid males at centromeres and distal ends.

| Study populations
In 2013, we obtained four (2 males, 2 females) Timor zebra finches from a local breeder in Germany, two of which were brother and sister. The Australian zebra finches used in this study stemmed from a recently wild-derived population housed at the Max Planck Institute for Ornithology in Seewiesen, Germany. They are descendants of birds from study population "Bielefeld-AUS" described in Forstmeier et al. (2007) and genetically close to wild Australian birds.
The two subspecies differ phenotypically (Clayton et al., 1991, Figure 1). Timor zebra finches weigh less and have shorter wings than Australian zebra finches. Males also differ in that Timors have more orange (less red) beaks, smaller (absolute and relative to their size) black breast bands, and no black barring on the throat and foreneck. All these differences were found between the four Timor and 110 Australian zebra finches used in this study (all birds measured by the same observer, WF, Table S1).
To test whether any admixture between Australian and Timor zebra finches had occurred in captivity prior to our experiment, we examined the genome of the male Timor zebra finch that bred in this study. This individual had been sequenced for another project (data kindly shared by Alexander Suh) with 60× coverage using Illumina paired-end reads (150 bp; details of library preparation, analyses scripts, and data will be deposited along with that project). In brief, reads were quality-and adaptor-trimmed using BBDuk (v38.25; https://sourc eforge.net/proje cts/bbmap/), and then mapped against the reference genome WUSTL 3.2.4 (Warren et al., 2010) using bwa-mem (v0.7.17; ). We called SNPs on this Timor male and also on 100 wild-caught Australian zebra finches that had been sequenced using a pooled-sequencing approach (see Knief, Hemmrich-Stanisak, et al., 2015; for methodological details and the Open Science Framework [https://osf.io/ dkqth/] for data) simultaneously with samtools (v1.6;  using "mpileup" (using a base quality Pfed score and a mapping quality score of more than 20 while removing reads that were unmapped, secondarily mapped, quality filtered or duplicated). This generated a set of 24 million high-quality SNPs, in which we compared the SNPs of the Timor male with those in the pool of 100 wild-caught Australian zebra finches. Specifically, we counted the number of SNPs that were homozygous (nonreference) in the Timor zebra finch and absent from the pooled-sequencing data in 500 kb nonoverlapping sliding windows. We regarded those genomic windows that had less than 100 Timor-specific SNPs as being introgressed from an Australian zebra finch. We identified 16 such regions (covering the centromeres on chromosomes Tgu1, Tgu5, Tgu13 and Tgu15, Figure S2), all of which were heterozygous (i.e., 4.78% of the diploid genome is admixed). Hence, none of the 16 introgressed Australian regions were homozygous for the Australian haplotype, meaning that no potential meiotic drivers had gone to fixation prior to our study.

| Breeding design
Birds were housed and bred in large semi-outdoor aviaries. For a detailed description of the housing conditions, see Ihle et al. (2013).
For breeding, individuals were allowed to freely choose partners.
Individuals from each population were split according to sex and put in aviaries such that individuals with Timor ancestry could only pair with individuals from Australia (and vice versa; 2-4 aviaries with 4-24 birds each).

F I G U R E 1
The breeding design and a description of the hybrid generations employed in this study. Squares represent males, circles females and diamonds a mixture of both sexes. Meioses in the four main groups (bold symbols) were informative for the analyses of meiotic drive. Colors refer to the expected fraction of Australian (gray) and Timor (orange) ancestry in each generation, respectively. Sample sizes refer to the numbers of breeding pairs and the combined numbers of embryos and offspring from the F1-, BC1-, and BC2-hybrid generations. A, Australian zebra finch; T, Timor zebra finch First, we hybridized Australian and Timor zebra finches to produce an F1 generation (Figure 1; N = 11 hybrid individuals that reached sexual maturity, 7 males and 4 females). Only two of the four Timor birds (one of each sex, not siblings) contributed to F1.
Then, the F1 generation was backcrossed to Australian birds (BC1, Figure 1). Each pair was allowed to raise young from two clutches (N = 51 individuals that reached sexual maturity, 26 males and 25 females). Of those birds, 41 (18 males and 23 females) were again backcrossed to Australian birds (BC2, Figure 1).
From the BC1 generation, we sampled DNA from almost all eggs from all clutches, including dead embryos and deceased chicks. A small number of eggs were either broken (N = 3), lost (N = 3) or had no egg yolk (N = 1) but since this was a random subset of all eggs, it should not affect any of our conclusions. Eggs from clutches that were not used for further breeding were collected, put into an incubator and opened prior to hatching in order to obtain DNA samples.  (Griffith et al., 2008;Mariette & Griffith, 2012;Zann, 1996). Studies that have examined the presence of sperm on the perivitelline layer of eggs (Birkhead & Fletcher, 1998;Pei et al., 2020) revealed that it is nearly always the absence of sperm that explains why an egg is not fertilized. However, the risk remains that at least some of the eggs classified as infertile suffered early embryo mortality, such that the development cannot be seen, a problem that can be dealt with in various ways (see Knief, Schielzeth, et al., 2015). To control for such potential early embryo mortality, we here contrasted the transmission ratios in female parents to those in male parents, because we did not expect any drive linked to centromeres or distal ends in the latter (see below).
Our backcross design was asymmetric (F1 hybrids were backcrossed to Australian birds but not to Timor birds) for the following reasons. First, only for Australian birds did we have a large enough and genetically diverse captive population available for backcrossing. This is needed to avoid high rates of early embryo mortality due to inbreeding Forstmeier et al., 2012). Second, it is more likely that a driving allele has gone to fixation in the subspecies with the larger effective population size in the wild (i.e., the Australian, see Introduction). Such drivers might require the segregation "machinery" of the Australian subspecies to be active. Hence, backcrossing hybrids to Australian birds will more likely expose cryptic drivers, if they exist.

Because uncontrolled introgression of Timor DNA into captive
Australian zebra finch populations is undesirable, all hybrids were sacrificed at the end of this study. All procedures on zebra finches (housing, breeding, banding, bleeding for parentage assignment, measuring, observing, and sacrificing) do not qualify as animal experimentation according to the relevant national and regional laws and are fully covered by our housing and breeding permit (# 311.4-si, by Landratsamt Starnberg, Germany).

| Genetic markers
Previously, we used a set of 62 microsatellite markers to infer the positions of centromeres in the current zebra finch genome assembly

| Analysis of parentage & chromosomal abnormalities
We performed genetic parentage analyses for each aviary separately using the SOLOMON package (v1.0-1; Christie et al., 2013) in R (v3.4.3; R Core Team, 2017). We used the "Bayesian parentage analysis with no known parents" option with the recommended default settings for microsatellite genotype data. Using all 56 microsatellite markers, parents were unambiguously assigned to all offspring.
For embryos with chromosomal abnormalities (N = 39 out of a total of 1,359 samples; showing either tetraploidy, triploidy, haploidy, trisomy, or monosomy), we identified the parental origin and the most likely cause of the error (nondisjunction in meiosis I or in meiosis II or polyspermy). Whenever the parent contributing the abnormal chromosome set was a hybrid or a backcrossed individual (whose inheritance of alleles is of interest here), we removed the offspring with the chromosomal abnormality from further analyses for the chromosomes affected.

| Testing for segregation distortion
We followed the Timor and the Australian microsatellite alleles through the backcross pedigree and measured segregation distor-  Table S2 for more details). We counted the transmissions for all markers (centromeres and distal chromosome end), for both females and males, and for the F1 and BC1 generation separately. We expected a chromosomal meiotic driver to act only in female parents at centromeres (drive in meiosis I) or at distal ends (drive in meiosis II).
Throughout the study, we counted the transmission of Australian alleles as 1 (success) and of Timor alleles as 0 (failure). We define the drive parameter k as the proportion of progeny (successful gametes) that inherited an Australian allele (see also Lyttle, 1991). Thus, at fair Mendelian segregation k = 0.5 and when Australian alleles are more often transmitted than Timor alleles k > 0.5.
To obtain estimates of the background transmission rates, we fitted generalized linear mixed-effects models (GLMM) with a binomial error structure using the transmission counts of all markers (1 = Australian allele, 0 = Timor allele transmitted) as the dependent variable, the intercept as the only fixed effect, and marker identity as a random intercept effect. We fitted separate models for females and males and for both sexes combined. We used the k-value estimate across all markers in male parents as the background transmission rate, which makes all estimates in our study more conservative This is equivalent to using a binomial test (as in Knief, Schielzeth, et al., 2015) with a success probability of the background transmission rate. We report drive parameter k estimates from these models, such that k = 0.5 corresponds to the background transmission rate (background k = 0.495, see results) and k < 0.5 describe drive parameters smaller than the background rate. By considering all markers separately, weak genome-wide drive could be missed and we thus fitted the same GLMMs as described above for female par- we report k > 0.5 in the power analyses, but it should be noted that k is symmetric around 0.5. All data can be found in Knief et al. (2019).

| RE SULTS
We determined 74,829 microsatellite allele transmissions in the F1 and BC1 generations, of which 38,541 (51.5%) were informative for estimating transmission ratios of Australian and Timor alleles.
Sample sizes were evenly distributed across sexes and generations (Table S3), but varied between markers depending on their allelic diversity, heterozygosity, and null-allele frequency within the population (all p ≤ .01; Table S4). When pooling all markers within generations and sexes, power to detect even weak deviations from  Table S5: "All markers"). We ruled out that this bias resulted from undetected null alleles by analyzing the subset of trios where both parents were heterozygous without a null-allele (combinations 4, 7, and 13 in Table S2). The effect did not change (combined sexes GLMM: N = 25,685 informative transmissions, k = 0.491, p = .014; Table S5: "Heterozygous parents"). Next, we tested whether inbreeding depression might have caused this deviation from Mendelian segregation by sub-setting the data to those trios in which the parents could not produce offspring homozygous for an Australian microsatellite allele (combinations 11, 12, 13, and 16 in Table S2). We thereby assumed that microsatellites tagged larger haplotypes in our captive population that became identical-by-descent in individuals homozygous for a specific Australian allele. In this dataset, the bias toward an increased transmission of Timor alleles was slightly lower and nonsignificant (GLMM: N = 18,714 informative transmissions, k = 0.496, p = .27; Table S5: "No inbreeding"). Thus, to rule out inbreeding effects, we set the transmission rate estimated from all F I G U R E 2 Transmission ratios of individual markers in the F1 generation (a,b) and in the BC1 generation (c,d) at centromeres (a,c) and distal chromosomal ends (b,d). Transmission ratios in females and males are plotted on the x-and y-axis, respectively. Numbers refer to the chromosome number in the zebra finch genome assembly and a gray background indicates a significant deviation from fair Mendelian segregation in either of the two sexes (no marker deviated significantly in both sexes). Note that only 11 out of 208 tests reached statistical significance at α = 0.05. The dashed line is the identity line  Table S5: "All markers"). The bias in centromeric marker transmissions was only present in the BC1 generation (k = 0.475, p = 9 × 10 -4 , Figure 3,   Table S7). It seems unlikely that the apparent overall drive at centromeric markers in BC1 females was due to undetected early embryo mortality (i.e., the missing Australian alleles being hidden in eggs that were incorrectly judged as infertile), because the overall rate of apparent infertility was low (only 6% of eggs) when BC1 females were crossed with Australian males in comparison to other crosses (Table S8).

| D ISCUSS I ON
Most of the meiotic drivers discovered thus far exhibit drive parameters of k > 0.60, especially in crosses between populations or species (Chmátal et al., 2014;Didion et al., 2015;Fishman & Saunders, 2008;Fishman & Willis, 2005;Rhoades, 1942). Despite having high power to discover even smaller deviations from Mendelian segregation in our study, we found no clear evidence for any active meiotic driver in a cross between Australian and Timor zebra finches. It should be noted, however, that we tagged centromeres of only 27 chromosomes, whereas the somatic zebra finch genome consists of 40 chromosomes (Pigozzi & Solari, 1998), leaving the possibility for drivers on the remaining-as yet mostly unassembled-13 microchromosomes.
We found a weak deviation from fair Mendelian segregation ratios in both female and male parents at both centromeres and distal chromosomal ends, indicating the presence of a selective force other than meiotic drive. Contrary to our expectation that the more efficient drive alleles would have evolved in Australian zebra finches, the bias was in favor of Timor alleles being more often transmitted to the next generation. We used a breeding design in which we backcrossed hybrids to Australian zebra finches.
Because of the low recombination rate in the zebra finch genome (Backström et al., 2010), this could have led to large chromosomal parts becoming identical-by-descent. The resulting inbreeding depression might have manifested itself through increased early embryo mortality without any visible embryonic development Forstmeier et al., 2012). If so, it would appear as if Australian alleles were transmitted less often than Timor alleles to the surviving offspring. In contrast, Bateson-Dobzhansky-Muller (BDM;Bateson, 1909;Dobzhansky, 1937;Muller, 1942) incompatibilities are more likely to remove admixed individuals  Figure 1). Point size reflects sample size, that is, the number of informative meioses (N). All estimates stem from models in which the background transmission rate was not accounted for further expected to cause variation in transmission ratios between parental individuals, which we did not find. Specifically testing inbreeding and BDM incompatibility effects ideally requires a reciprocal backcross design.
After accounting for the above-mentioned small deviation from fair Mendelian segregation that might have been caused by inbreeding, none of the chromosomes exhibited segregation distortion by themselves. However, we still observed a small deviation from Mendelian segregation in females of the BC1 generation at centromeres, again in the direction of an increased transmission of Timor alleles (k = 0.470 without and k = 0.475 with control for the background transmission rate). Tests for meiotic drive are sensitive to genotyping errors (see Meyer et al., 2012), which we ruled out by using the transmission ratios of the same markers in males as our background transmission rate. Given that there was also no such effect at distal chromosomal ends in females, this might indicate weak genome-wide meiotic drive in meiosis I, in which Timor centromeres preferentially enter the oocyte and outcompete the Australian alleles. This could happen if larger centromeres are preferentially transmitted within a specific genetic background. Many small mutations may accumulate in such a species, leading to all or most centromeres being enlarged. Whenever these enlarged centromeres compete with shorter centromeres (e.g., in a female hybrid), this would result in genome-wide meiotic drive, as has been found in mice (Akera et al., 2019;Iwata-Otsubo et al., 2017). The nobs in maize are another example, which can be present on all chromosomes, show neocentromeric activity, and-in the heterozygous state-drive across many chromosomes (Rhoades, 1942).
In pure Australian zebra finches, we had previously observed a potential meiotic driver on chromosome Tgu5 (k = 0.602) that was active in only some generations of an extended pedigree (Knief, Schielzeth, et al., 2015). Together with the current results, this might indicate that some meiotic drivers are environmentally induced and not constantly active, as has been described in other systems (Rhoades, 1942;Shaw & Hewitt, 1984). Alternatively, such cases may be examples of the "winner's curse" (Xiao & Boehnke, 2009), in which a false-positive finding is followed by true-negative results.

| CON CLUS ION
Meiotic drive has been proposed as a potent evolutionary force but its frequency in nature remains unknown and its impact on genetic and phenotypic variation, genome structure, or speciation is difficult to assess. We here specifically tested for deviations from fair Mendelian segregation of chromosomes in a cross between two diverged subspecies, the Australian and Timor zebra finch.
Crossing phylogenetically more distant species to Australian zebra finches does not result in fertile offspring (Forshaw et al., 2012;McCarthy, 2006). We expected the more potent meiotic drivers to evolve in the population with the larger effective population size, which is the Australian zebra finch. However, although the weak genome-wide segregation distortion that we observed in females of the BC1 generation might indeed be attributable to centromeric meiotic drive, the Timor alleles outcompeted the Australian ones.
Thus, we caution against interpreting our finding as definite evidence for centromeric drive. A nominal P-value of 6 × 10 -5 is unlikely a type I error, but not fully compelling either, because one could argue that we tested 56 markers in two generations, two sexes and with and without taking the background transmission rate into account (56 × 2 × 2 × 2 = 448 tests, translating into a Bonferroni corrected p-value ≈ .03).
We failed to find evidence for strong localized drivers. The moderate driver on chromosome Tgu2 that we had observed previously (Knief, Schielzeth, et al., 2015) in a domesticated population of Australian zebra finches (population "Seewiesen- GB" in Forstmeier et al., 2007) was not present in the pedigree we analyzed for the current study (Australian zebra finches were derived from population "Bielefeld-AUS" in Forstmeier et al., 2007).
Assuming an infinite population size and no heterozygote disadvantage, even weak drivers will eventually invade and ultimately reach fixation (Traulsen & Reed, 2012), thereby causing a reduction in genetic diversity at the driving loci and potentially a decrease in organismal fitness ("selfish sweeps"; Didion et al., 2016). Weak meiotic drivers as we might have found here have not been reported yet, but this might in part be due to detection bias and insufficient statistical power. Weak meiotic drive might be a more common phenomenon warranting further investigation in other taxa.

ACK N OWLED G M ENTS
We thank Frank Rößler for providing the Timor zebra finches. We are grateful to K. Martin and M. Ihle for help with breeding and S. Bauer, E. Bodendorfer, A. Grötsch, A. Kortner, K. Martin, P. Neubauer, F. Weigel, and B. Wörle for animal care. We further thank M. Schneider for laboratory work. Alexander Suh provided access to his sequencing data. This study was funded by the Max Planck Society (B.K.).
Open access funding enabled and organized by Projekt DEAL.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.