Recent information has revealed the functional diversity and importance of mitochondria in many cellular processes including orchestrating the innate immune response. Intriguingly, several infectious agents, such as Toxoplasma, Legionella, and Chlamydia, have been reported to grow within vacuoles surrounded by host mitochondria. Although many hypotheses have been proposed for the existence of host mitochondrial association (HMA), the causes and biological consequences of HMA have remained unanswered. Here we show that HMA is present in type I and III strains of Toxoplasma but missing in type II strains, both in vitro and in vivo. Analysis of F1 progeny from a type II×III cross revealed that HMA is a Mendelian trait that we could map. We use bioinformatics to select potential candidates and experimentally identify the polymorphic parasite protein involved, mitochondrial association factor 1 (MAF1). We show that introducing the type I (HMA+) MAF1 allele into type II (HMA-) parasites results in conversion to HMA+ and deletion of MAF1 in type I parasites results in a loss of HMA. We observe that the loss and gain of HMA are associated with alterations in the transcription of host cell immune genes and the in vivo cytokine response during murine infection. Lastly, we use exogenous expression of MAF1 to show that it binds host mitochondria and thus MAF1 is the parasite protein directly responsible for HMA. Our findings suggest that association with host mitochondria may represent a novel means by which Toxoplasma tachyzoites manipulate the host. The existence of naturally occurring HMA+ and HMA- strains of Toxoplasma, Legionella, and Chlamydia indicates the existence of evolutionary niches where HMA is either advantageous or disadvantageous, likely reflecting tradeoffs in metabolism, immune regulation, and other functions of mitochondria.