Negative heterosis for meiotic recombination rate in spermatocytes of the domestic chicken Gallus gallus

Benef its and costs of meiotic recombination are a matter of discussion. Because recombination breaks allele combinations already tested by natural selection and generates new ones of unpredictable f itness, a high recombination rate is generally benef icial for the populations living in a f luctuating or a rapidly changing environment and costly in a stable environment. Besides genetic benef its and costs, there are cytological effects of recombination, both positive and negative. Recombination is necessary for chromosome synapsis and segregation. However, it involves a massive generation of double-strand DNA breaks, erroneous repair of which may lead to germ cell death or various mutations and chromosome rearrangements. Thus, the benef its of recombination (generation of new allele combinations) would prevail over its costs (occurrence of deleterious mutations) as long as the population remains suff iciently heterogeneous. Using immunolocalization of MLH1, a mismatch repair protein, at the synaptonemal complexes, we examined the number and distribution of recombination nodules in spermatocytes of two chicken breeds with high (Pervomai) and low (Russian Crested) recombination rates and their F1 hybrids and backcrosses. We detected negative heterosis for recombination rate in the F1 hybrids. Backcrosses to the Pervomai breed were rather homogenous and showed an intermediate recombination rate. The differences in overall recombination rate between the breeds, hybrids and backcrosses were mainly determined by the differences in the crossing over number in the seven largest macrochromosomes. The decrease in recombination rate in F1 is probably determined by diff iculties in homology matching between the DNA sequences of genetically divergent breeds. The suppression of recombination in the hybrids may impede gene f low between parapatric populations and therefore accelerate their genetic divergence.


Introduction
Benefits and costs of meiotic recombination are a favorite subject of theoretical discussions and mathematical mo dels (Kondrashov, 1993;Otto, Lenormand, 2002;Hartfield, Keightley, 2012;Rybnikov et al., 2020). They are mostly focused on the population genetic effects of recombination, i. e. its contribution to genetic and phenotypic variability. Crossing over reduces linkage disequilibrium by breaking old allele combinations already tested by natural selection and generating new ones of unpredictable fitness. Therefore, a high recombination rate is generally beneficial for popula tions living in fluctuating or rapidly changing environments and costly in a stable environment (Otto, Michalakis, 1998;Lenormand, Otto, 2000). Besides genetic benefits and costs, there are cytological effects of recombination, both positive and negative. Recombination is necessary for chromosome synapsis and segregation. However, it involves a massive generation of doublestrand DNA breaks. Insufficient or er roneous repair of the breaks leads to the death of the affected germ cells or various mutations and chromosome rearrange ments (Zickler, Kleckner, 2015).
Crossing over distribution along the chromosomes is an other important variable affecting both genetic and cytologi cal benefits and costs of recombination. Two crossing overs positioned too close to each other do not affect the linkage phase (Gorlov, Gorlova, 2001;Berchowitz, Copenhaver, 2010). Similarly, crossing overs located too close to a centro mere of an acrocentric chromosome or to telomere do not produce new allele combinations. In these cases, the cost of recombination is paid, but no benefit is gained. Cytological costs of cross ing overs that are too distal or too proximal should also be taken into account. They often lead to incorrect chromosome segregation and generation of chromosomally unbalanced gametes (Koehler et al., 1996;Hassold, Hunt, 2001). Thus, the benefits of recombination (generation of new allele combina tions) would prevail over its costs (occurrence of deleterious mutations) as long as the population remains sufficiently heterogeneous.
The heritability of recombination rate was estimated as 0.30 in humans, 0.22 to 0.26 in cattle and 0.15 in sheep (Kong et al., 2004;Sandor et al., 2012;Johnston et al., 2016). Inter breed variation in recombination rate was detected in rams (Davenport et al., 2018) and roosters (Malinovskaya et al., 2019). The most intriguing finding of the latter study was a correspondence between the age of the breed and its recom bination rate. Relatively young breeds created by crossing several local breeds showed high recombination rates, while ancient local breeds displayed a low recombination rate. The decrease in recombination rate with breed age might be a correlative response to a decrease in genetic heterogeneity within each breed with time due to inbreeding and artificial selection (Lipinski et al., 2008;Gibbs et al., 2009). Early stages of conscious selection for economic traits were probably ac companied by unconscious selection for a high recombination rate. A reduction of genetic variability, an inevitable result of inbreeding and selection, leads to a decrease in recombination efficiency and therefore reduces selective advantages of high recombination rate.
In this paper, we examine the inheritance of the recombina tion rate in male F 1 hybrids and backcrosses of the chicken breeds showing the highest (Pervomai) and lowest (Russian Crested) level of recombination among the six breeds exa mined by L.P. Malinovskaya et al. (2019). The Pervomai breed was produced in 1930-1960 by a complex reproductive crossing of three crossbred breeds: White Wyandotte (derived from crosses between Brahmas and Hamburgs), Rhode Island (derived from crosses between Malays and brown Italian Leg horns) and Yurlov Crower (derived from crosses of Chinese meat chicken, gamecocks and landraces). Russian Crested is an ancient local breed described in the European part of Rus sia in the early XIX century (Paronyan, Yurchenko, 1989).
We estimated the number and distribution of recombination nodules in spermatocytes using immunolocalization of MLH1, a mismatch repair protein of mature recombination nodules, at the synaptonemal complexes (SCs). This method has proved to produce reliable estimates of the overall recombination frequency and the distribution of recombination events along individual chromosomes (Anderson et al., 1999;Froenicke et al., 2002;Segura et al., 2013;Pigozzi, 2016).

Material and methods
Animals. Thirtyfour adult fivemonthold roosters were used in this study. Eight of them were Pervomai breed, nine -Rus sian Crested breed, three -F 1 hybrids between Pervomai dams and Russian Crested sires, fourteen -backcrosses of F 1 sires to Pervomai dams.
The roosters were bred, raised and maintained at the poultry farm of the L.K. Ernst Federal Research Centre for Animal Husbandry under conventional conditions. Maintenance, hand ling and euthanasia of animals were carried out in accordance with the approved national guidelines for the care and use of laboratory animals. All experiments were approved by the Ethics Committee on Animal Care and Use at the Institute of Cytology and Genetics of the Siberian Branch of the Russian Academy of Sciences (approval No. 35 of October 26, 2016 and 45/2 of January 10, 2019).
Antibodies were diluted in PBT (3 % bovine serum albumin and 0.05 % Tween 20 in PBS). A solution of 10 % PBT was used for blocking nonspecific binding of antibodies. Vecta shield antifade mounting medium (Vector Laboratories, Bur lingame, CA, USA) was used to reduce fluorescence fading. The preparations were visualized with an Axioplan 2 micro scope (Carl Zeiss, Germany) equipped with a CCD camera (CV M300, JAI Corporation, Yokohama, Japan), CHROMA filter sets and ISIS4 imageprocessing package (MetaSystems GmbH, Altlußheim, Germany). The location of each imaged immunolabeled SC spread was recorded so that it could be relocated on the slide after FISH.

Image analysis.
We measured the length of each SC and the total SC length in µm, scored the number of MLH1 sig nals localized on SCs and recorded their positions relative to the centromere using MicroMeasure 3.3 software (Reeves, 2001). For the seven largest macroSCs identified by relative lengths and centromeric indices, we visualized the pattern of MLH1 foci distribution. We divided the average length of SC by intervals and plotted the relative number (the proportion) of MLH1 foci within each interval. To make the intervals on chromosomes of different lengths comparable, we set the number of intervals for each SC proportional to the average SC length, being ~1 μm.
The Statistica 6.0 software package (StatSoft) was used for descriptive statistics. Mann-Whitney Utest was used to estimate the differences between the genotypes in the average number of MLH1 foci per cell and each macrochromosome, p < 0.01 was considered to be statistically significant. Values in the text and figures are presented as means ± S.D.

Results
We analyzed the number and distribution of MLH1 foci at 52 650 SC in 1350 spermatocytes of 34 roosters. The rooster pachytene karyotype contained 38 autosomal SCs and a ZZ pair. We identified the seven largest macroSCs by their relative lengths and centromeric indices. SC1, SC2 and SCZZ were large metacentrics. They differed from each other in length and centromeric indices ( p < 0.001). SC3 and SC5 were large and mediumsized acrocentics, while SC4 and SC7 were mediumsized submetacentics, which also differed from each other in their relative lengths and centromeric indices. The macroSCs 6, 8-10 and all microSCs were acrocentric, with gradually decreasing chromosomal sizes (Fig. 1). All chromo somes showed orderly synapsis. No SCs with asynapsis were detected at pachytene spreads of the specimens of the parental breeds and their F 1 hybrids and backcrosses.
In order to test the reliability of the morphological iden tification of macrochromosomes, we performed FISH with universal BAC probes obtained from the CHORY261 library, marking chicken macrochromosomes, on SC preparations after immunolocalization of SYCP3 and centromeric proteins (Fig. 2). Comparison of the FISH results with the results of identification by relative sizes and centromeric indices showed good agreement for all chromosomes. We correctly identified the first seven macrochromosomes and chromosome Z. Chro mosomes 6 and 7 are of similar SC lengths and are acrocentric and subacrocentric, respectively.

List of BAC clones used for FISH BAC clone
Gallus gallus chromosome Arm Negative heterosis for meiotic recombination rate in spermatocytes of the domestic chicken Gallus gallus The numbers in parentheses indicate the number of studied individuals and cells. Average values of genotypes are shown in black, individual values of backcrosses are shown in gray. "*" -differences with Pervomai, Mann-Whitney test, p < 0.01; "+" -differences with Russian Crested, Mann-Whitney test, p < 0.01; "#" -differences with F 1 , Mann-Whitney test, p < 0.01.  The colors represent the proportion of bivalents with 1 to 13 MLH1 foci per chromosome. "*" -differences with Pervomai, Mann-Whitney test, p < 0.01; "+" -differences with Russian Crested, Mann-Whitney test, p < 0.01; "#" -differences with F 1 , Mann-Whitney test, p < 0.01.
Despite these differences in the number of MLH1 foci per particular macrochromosome between the parental breeds, F 1 and backcrosses, each of them showed almost the same chromosomespecific pattern of MLH1 foci distribution along the SC (see Fig. 4). On most chromosomes, an increase in the frequency of recombination was observed in the distal regions.

Discussion
The most important and surprising result of our study is a discovery of overdominance of low recombination rate in F 1 hybrids, measured as the number of MLH1 foci per pachy tene cell. Backcrosses of the F 1 hybrids to the parental breed with high recombination rate were rather homogenous and showed an intermediate recombination rate. Thus, the model of inheritance of recombination rate in roosters can be formally described as negative heterosis in F 1 and additive inheritance in backcrosses.
The differences in overall recombination rate between the breeds, hybrids and backcrosses were mainly determined by the differences in the crossing over number in the large macrochromosomes. They are characterized by a high (up to 13!) and variable number of crossing overs, while small macrochromosomes have one or two chiasmata and each microchromosome contains only a single obligate chiasma necessary for orderly chromosome segregation.
Generally, crossbreeds are expected to show positive hete rosis for productivity traits (hybrid vigor) (Chen, 2013). This expectation contradicts the negative heterosis for the recom bination rate observed in this study. Interestingly, the rate of dilution of heterosis for recombination rate in backcrosses is higher than the rate of dilution of positive heterosis for eco nomic traits, at least in plants (Fridman, 2015). The decrease in recombination rate in F 1 is probably determined by difficul ties in homology matching between the DNA sequences of genetically divergent breeds (which we shall discuss below), rather than by dominant/overdominant genetic effects. With further level of backcrossing, the recombination rate acts like a regular complex trait with additive heritable component and environmental influence.
Our finding poses at least three interesting questions. How common is the negative heterosis for the recombination rate? What might be its molecular mechanism? What are its popula tion genetic implications?
The first question is difficult to answer because we are aware of only a few prior studies in which recombination rates have been compared between parental breeds or species and their hybrids. There were no significant differences in autosomal recombination rate between two species of dwarf hamsters diverged about 1 MYA and their F 1 female and male hybrids (Bikchurina et al., 2018). On the other hand, recombination in female hybrids between Microtus arvalis and M. levis diverged from 0.2 to 0.4 MYA and differing by a series of chromosomal rearrangements was significantly reduced compared to the parental species (Torgasheva, Borodin, 2016). Interspecific hybrids between Saccharomyces cerevisiae and S. paradoxus demonstrated low frequencies of genetic recombination (Hunter et al., 1996). Genomewide introgression between two closely related nematode species Caenorhabditis briggsae and C. nigoni also revealed substantial suppression of recombina tion in the hybrids (Bi et al., 2015).
The molecular mechanism of negative heterosis for re combination rate is probably linked with the initial stages of chromosome synapsis and recombination, which includes scheduled generation of multiple doublestrand DNA breaks (DSB), RAD51mediated strand invasion and sequence ho mology matching (Zickler, Kleckner, 2015). Reduced recom bination in interspecies hybrids may occur due to a significant decrease in homology between parent species accompanied by serious impairments of the chromosome synapsis in meiosis. However, even a minor decrease in homology at the early stages of divergence can apparently affect recombination due to decreased sequence identity. Comparison of recombination boundary sequences suggests that recombination in hybrids may require a region of high sequence identity of several kilobases in length (Ren et al., 2018).
Similarly, the study of recombination rate in hybrids be tween S. cerevisiae strains using highthroughput method showed a positive correlation of its level with sequence similarity between homologs at different scales (Raffoux et al., 2018). This is consistent with the finding that sequence divergence greater than about 1 % leads to the suppression of recombination due to heteroduplex rejection by the mismatch repair machinery (Chen, JinksRobertson, 1999). An anti recombination activity of the mismatch repair system during meiosis might contribute towards a decrease in recombination rate in hybrids between diverging breeds, populations and species (Radman, Wagner, 1993). At relatively low genetic distances it decreases the recombination rate in the hybrids, at greater genetic distances it impairs chromosome synapsis and might lead to hybrid sterility due to meiotic silencing of unpaired chromatin (Turner, 2015).

Conclusion
There might be interesting evolutionary and population ge netic implications of our findings. The negative heterosis for recombination in the hybrids may play an important role in speciation. Suppression of recombination impedes gene flow between parapatric populations and therefore accelerates their genetic divergence (Rieseberg et al., 1999;Baack, Rieseberg, 2007). A possibility of negative heterosis for recombination may also be taken into account in the calculations of the in trogression time based on the size of linkage disequilibrium blocks (Payseur, 2010). They are based on the assumption that global and local recombination rates are constant over the generations. Our data indicate that it might not be the case. We detected a decrease in recombination in the macrochromo somes of the hybrids, while the microchromosomes retained the same recombination rate because it had already been the minimal required for orderly segregation.