4 | Discussion
Adaptive introgression is known in some plant taxa, often resulting in divergent ecotypes colonizing novel habitats (Suarez-Gonzalez et al. 2018c). In European and American oaks, divergence and introgression of adaptive traits between interfertile species have been demonstrated. In drought-averse Q. rubra and drought-tolerant Q. ellipsoidalis (section Lobatae ), parapatric populations of both species are fixed for alternate alleles at a CONSTANS -like gene, which affected seedling survival in common gardens, and the Q. ellipsoidalis allele is introgressed into the Q. rubra genome depending on soil moisture conditions in their sympatric populations (Lind-Riehl et al. 2014; Lind-Riehl and Gailing 2016; Khodwekar and Gailing 2017). In Q. robur and Q. petraea (sectionQuercus ), introgression from Q. robur contributes to divergence among locally adapted Q. petraea populations along a gradient of climatic conditions (Leroy et al. 2020). In Asian oaks, introgressed genomic regions were investigated in admixture betweenQ. acutissima and Q. variabillis (section Cerris ), and admixed populations occupying similar habitats tended to share the same introgressed regions (Fu et al. 2022). We also found a candidate system for adaptive introgression, which consists of Q. mongolicavar. crispula (Qc ) in inland habitats and Q. dentata (Qd ) in coastal habitats in northern Japan (Nagamitsu et al. 2019; Nagamitsu et al. 2020). In this system, we tried to verify intensive introgression of Qd alleles at loci associated with traits adaptive to coastal environment.
Our previous study demonstrated environment-dependent introgression fromQd to Qc in coastal habitats, resulting in Q. ×angustilepidota (Qa ), which are hybrids after the first generation of backcross to Qc (Nagamitsu et al. 2020). Our present study using a comprehensive tree collection in common gardens revealed a hybrid zone around the northern distributional limit ofQd , including both northern-edge Qd populations admixed with Qc and coastal Qa populations backcrossed toQc . The observed genomic compositions indicated random admixture between Qc and Qd , suggesting frequent gene flow and weak genetic drift. The genotypic PCA indicated that some northern-edgeQd populations had unique genetic properties in some PCs with minor contributions to genetic variations, suggesting the presence of weak genetic structure in coastal habitats, probably due to fragmentation of sand dunes and coastal cliffs in the distributional range margins. This hybrid zone is likely to be maintained by the tension zone process (Barton and Hewitt 1985). Near the northern distributional limit of Qd , gene flow between Qc andQd seems relatively frequent, because their flowering synchrony increases at higher latitudes probably due to shorter periods of growing season (Shimizu et al. 1995). Lower fitness of Qd at higher latitudes in comparison to fitness of Qc coupled with both lower fitness of Qc in coastal habitats and lower fitness of Qdin inland habitats may be responsible for natural selection that maintains the tension zone.
Common garden experiments have revealed local adaptation in various traits of oaks (Cavender-Bares and Ramírez-Valiente 2017). Local adaptation of Qc to climatic conditions was evident at regional scales (Nagamitsu and Shuri 2021). The size of 30-year-old trees in our common gardens indicated that the size was larger in Qc trees than in Qd trees in the inland site but larger in Qd trees than in Qc trees in the coastal site. This result demonstrates both adaptation of Qc to inland habitats and adaptation ofQd to coastal habitats. Although the plantation density and forest management differed between the sites, smaller tree size in the coastal site than in the inland site suggests higher stress in coastal habitats than in inland habitats. In the coastal site, which was located in north to the northern distributional limit of Qd , the size ofQd trees seemed to be suppressed and to become similar to the size of Qa trees. This result implies that environmental conditions at higher latitudes are maladaptive for Qd , which is distributed at lower latitudes than Qc . These findings of local adaptation are consistent with natural selection expected from the tension zone process.
Morphological traits of leaves and shoots observed in each site indicated more Qd -like phenotypes as the Qd ancestry increased, suggesting that these traits have the genetic basis derived from Qd . Artificially crossed F1 hybrids betweenQc and Qd exhibited pubescent shoots and intermediate phenotypes of morphological traits of leaves and fruits (Ubukata et al. 1996). These findings are consistent with our observed phenotypes of admixed individuals. We categorized into observed traits into taxon-specific and habitat-specific traits based on the results of phenotypic PCA. The habitat-specific traits, of which phenotypes ofQa and Qd trees in coastal habitats mainly differed from those of Qc trees in inland habitats, included the density of stellate hairs (trichomes) on the lower leaf surface, leaf mass per area (LMA), shoot diameter, and the number of axillary buds at bud-scale scars. These traits seem relevant to the stress tolerance in coastal habitats. Higher water absorption capacity of oak leaves with dense trichomes is advantageous in soils with lower water storage capacity (Fernández et al. 2014). Increased LMA in oaks are commonly observed in response to higher intensity of drought and light (Sancho-Knapik et al. 2021). More buds at the lower part of shoots, such as bud-scale scars, can compensate for bud mortality at the upper part of shoots caused by salt spray and harsh wind in winter (Asai et al. 1986). These lines of evidence suggest that Qd -like phenotypes of the habitat-specific traits are more adaptive in coastal environment. Thus, we expected thatQd alleles at loci associated with these traits are more intensively introgressed than those in the genomic background.
The observed shift in phenotypic variation to more Qd -like phenotypes from inland to coastal habitats suggests phenotypic plasticity in response to coastal stress. This phenotypic shift was larger in Qa trees than in Qd trees, indicating lower plasticity in more stress-tolerant Qd trees. This result is consistent with our previous findings that phenotypic plasticity in natural populations in response to coastal stress, which was measured by bud mortality, decreased with higher Qd ancestry estimated from nuclear microsatellite genotypes (Nagamitsu et al. 2019). Similar features in phenotypic plasticity were evident in Mediterranean oaks,Q. faginea , which showed lower morphological and physiological plasticity and more conservative resource-use strategy in populations from drier and colder provenances than in those from mesic and milder provenances (Solé-Medina et al. 2022). The phenotypic shift to moreQd -like phenotypes of Qa trees may result from induced expression of alleles introgressed from stress-tolerant Qd in response to coastal stress.
Introgression from Qd to Qc was evident from a positive genome-wide estimate of the Patterson’s D statistic in our previous study (Nagamitsu et al. 2020), and genome-wide patterns of introgression in our present study using more loci indicated that the extent of introgression were heterogeneous along chromosomes. However, extensively introgressed regions were not always consistent with local ancestry, probably due to incomplete linage sorting. To determine introgressed genomic regions precisely, denser SNP loci mapped to reference sequences of both parental taxa are necessary (Ai et al. 2022; Wang et al. 2023). In spite of the low resolution in our present study, the heterogeneity along chromosomes and various admixture patterns among individuals may make our GWA analysis feasible owing to a diverse array of recombinants.
We detected few loci consistently associated with traits in both SNP genotypes and local ancestry, probably due to sparse distribution of the available ddRAD loci, polygenic regulation of the examined traits, and confounding factors with the measured phenotypes. In a model plant,Oryza sativa , for example, major genes of some agronomic traits with alleles introgressed between rice varieties were clearly detected from both association mapping using SNP genotypes and admixture mapping using local ancestry (Zhao et al. 2010). In poplar, various traits were measured in a common garden of Populus trichocarpa , and associations of SNP genotypes with these traits were examined using 29k genome-wide loci (Mckown et al. 2014a; Mckown et al. 2014b). Introgression of P. balsamifera alleles to P. trichocarpagenome at some candidate genes around these trait-associated loci were detected using denser loci around these candidate genes (Suarez-Gonzalez et al. 2016). Furthermore, using 1169k loci, additional introgressed regions were found from local ancestry inference and admixture mapping (Suarez-Gonzalez et al. 2018a; Suarez-Gonzalez et al. 2018b). These studies imply that more loci are necessary to detect genes responsible for adaptive traits in our oak system. In spite of the limited power of our present study using 27k loci, we found several loci potentially associated with one taxon-specific and three habitat-specific traits. Two of these loci were located at intensively introgressed regions in chromosomes 9 and 11.
GO terms of genes around the loci potentially associated with the four traits suggest that these genes are related to signaling processes and stress responses with functions of DNA binding and protein binding. Some proteins of those genes were related to catabolic pathways of abscisic acid (ABA) in response to drought stress (Kushiro et al. 2004) and ABA transduction cascades in dormant seeds (Lorenzo et al. 2002), immunophilin responses to bacterial invasion (Mokryakova et al. 2014) and homoserine accumulation triggering pathogen resistance (Van Damme et al. 2009), and catabolism of lysine during leaf senescence (Engqvist et al. 2011). These findings imply that these genes can be related to the response to coastal stress. In introgression between Q. acutissima and Q. variabilis in China, not only genes involved in signaling processes such as responses to abiotic stimulus, stress, hormone, and pathogens but also cis -regulatory elements in untranslated regions were enriched in introgressed genomic regions (Fu et al. 2022). To detect candidate genes responsible for theQd -like phenotypes in response to coastal stress, not only morphological and physiological traits but also gene expression patterns should be investigated to select essential transcription factors and regulatory elements underlying adaptive phenotypes.
In summary, we revealed a hybrid zone with random admixture betweenQc and Qd in the northernmost area of Hokkaido. This hybrid zone is located in coastal habitats around the northern distributional limit of Qd , where Qd -like phenotypes are adaptive. We found heterogeneity of introgression along chromosomes and various admixture patterns among individuals, but failed to detect loci associated with the Qd -like phenotypes. To elucidate adaptive introgression suggested in our present study, more SNP genotypes at denser loci in admixed populations and deeper knowledge of genetic basis of adaptive traits are necessary.