Microbial symbionts exhibit broad genotypic variation in their fitness effects on hosts, leaving hosts vulnerable to costly partnerships. Interspecific conflict and partner-maladaptation are frameworks to explain this variation, with different implications for mutualism stability. We investigated the mutualist service of nitrogen fixation in a metapopulation of root-nodule forming Bradyrhizobium symbionts in Acmispon hosts. We uncovered Bradyrhizobium genotypes that provide negligible mutualist services to hosts and had superior in planta fitness during clonal infections, consistent with cheater strains that destabilise mutualisms. Interspecific conflict was also confirmed at the metapopulation level - by a significant negative association between the fitness benefits provided by Bradyrhizobium genotypes and their local genotype frequencies - indicating that selection favours cheating rhizobia. Legumes have mechanisms to defend against rhizobia that fail to fix sufficient nitrogen, but these data support predictions that rhizobia can subvert plant defenses and evolve to exploit hosts.

Efficient host control predicts the extirpation of ineffective symbionts, but they are nonetheless widespread in nature. We tested three hypotheses for the maintenance of symbiotic variation in rhizobia that associate with a native legume: partner mismatch between host and symbiont, such that symbiont effectiveness varies with host genotype; resource satiation, whereby extrinsic sources of nutrients relax host control; and variation in host control among host genotypes. We inoculated Acmispon strigosus from six populations with three Bradyrhizobium strains that vary in symbiotic effectiveness on sympatric hosts. We measured proxies of host and symbiont fitness in single- and co-inoculations under fertilization treatments of zero added nitrogen (N) and near-growth-saturating N. We examined two components of host control: 'host investment' into nodule size during single- and co-inoculations, and 'host sanctions' against less effective strains during co-inoculations. The Bradyrhizobium strains displayed conserved growth effects on hosts, and host control did not decline under experimental fertilization. Host sanctions were robust in all hosts, but host lines from different populations varied significantly in measures of host investment in both single- and co-inoculation experiments. Variation in host investment could promote variation in symbiotic effectiveness and prevent the extinction of ineffective Bradyrhizobium from natural populations.