C 4 photosynthesis evolved multiple times independently in angiosperms, but most origins are relatively old so that the early events linked to photosynthetic diversification are blurred. The grass Alloteropsis semialata is an exception, as this species encompasses C 4 and non-C 4 populations. Using phylogenomics and population genomics, we infer the history of dispersal and secondary gene flow before, during and after photosynthetic divergence in A. semialata. We further analyse the genome composition of individuals with varied ploidy levels to establish the origins of polyploids in this species. Detailed organelle phylogenies indicate limited seed dispersal within the mountainous region of origin and the emergence of a C 4 lineage after dispersal to warmer areas of lower elevation. Nuclear genome analyses highlight repeated secondary gene flow. In particular, the nuclear genome associated with the C 4 phenotype was swept into a distantly related maternal lineage probably via unidirectional pollen flow. Multiple intraspecific allopolyploidy events mediated additional secondary genetic exchanges between photosynthetic types. Overall, our results show that limited dispersal and isolation allowed lineage divergence, with photosynthetic innovation happening after migration to new environments, and pollen-mediated gene flow led to the rapid spread of the derived C 4 physiology away from its region of origin.

Abstract 1. Introduction 2. Materials and methods (a) Sampling, sequencing and data filtering (b) Genome sizing and carbon isotope analyses (c) Assembly of organelle genomes and molecular dating (d) Phylogenetic analyses of the nuclear genome (e) Genetic structure (f) Genome composition 3. Results (a) Genome sizes (b) Time-calibrated organelle phylogenies (c) Nuclear phylogeny (d) Population structure and genome composition 4. Discussion (a) Limited seed dispersal in the region of origin (b) Widespread pollen flow and sweep of the C4 nuclear genome (c) Recurrent hybridization and polyploidization 5. Concluding remarks Data accessibility Authors' contributions Competing interests Funding Acknowledgements Footnotes

This study was funded by the European Research Council (grant no. ERC-2014-STG-638333), the Royal Society (grant no. RGF\EA\181050) and has benefited from ‘Investissements d'Avenir' grants managed by the Agence Nationale de la Recherche (CEBA, ref. ANR-10-LABX-25-01 and TULIP, ref. ANR-10-LABX-41). Edinburgh Genomics, which contributed to the sequencing, is partly supported through core grants from the NERC (grant no. R8/H10/ 56), MRC (grant no. MR/K001744/1) and BBSRC (grant no. BB/ J004243/1). P.A.C. is funded by a Royal Society University Research Fellowship (grant no. URF\R\180022).