It should come as no surprise that this reaction is catalyzed by an enzyme called 2-oxoglutarate dehydrogenase (OGDH, also known by its old name: α-ketoglutarate dehydrogenase) (EC 22.214.171.124) that's almost identical to pyruvate dehydrogenase. In fact, both PDC and OGDH evolved from a common ancestral enzyme. We know that the citric acid cycle enzyme is a late comer because many species of bacteria don't have it. Indeed, they don't even have a citric acid cycle.
So we need to look elsewhere if we are going to find the source of PDC. The most primitive enzymatic reaction is almost certainly one that's required in amino acid metabolism.1 In this case it's a reaction involved in the degradation of the branched chain amino acids; leucine, valine, and isoleucine. Look at the pathway below.
The first step in the degradation is the removal of the amino group (-NH3+) and its replacement with an oxygen to form a keto (-C=O) group. This creates three similar 2-oxo acids (α-keto acids) all of which resemble 2-oxoglutarate and pyruvate. All three of the 2-oxo (α-keto) acids are acted upon by the same enzyme called branched chain 2-oxoacid dehygrogenase (OADH, BCOADH, α-ketoacid dehydrogenase) (EC 126.96.36.199) to create an acyl-CoA product. This is the same reaction as that catalyze by the pyruvate dehydrogenase complex except that the R group in pyruvate is -CH3 while in the case of the branched chain dehydrogenase it's a three, four, or five carbon branched structure.
BCOADH is found in all species. It is the most "primitive" enzyme. Like PDC it has a complex structure with three different subunits. E1 catalyzes the decarboxylation reaction. E2 catalyzes the formation of acyl-CoA—it has the lipoamide swinging arm. E3 catalyzes the oxidation of the lipoamide and the reduction of NAD+.
It looks like the "primitive" BCOADH could also catalyze the oxidative decarboxylation of pyruvate. In fact some of the modern enzymes still have residual activity towards the other substrates. Over time, the genes for some of the subunits duplicated and the two enzymes (PDC and BCOADH) diverged as they became more specialized for their modern substrates.
We can see the result if we look at the phylogenetic tree for the E2 subunit (below). This figure is from a paper by Scharrenberger & Martin (2002). They use a slightly different nomenclature (PDH=pyruvate dehydrogenase complex). This is an unrooted tree so you can't really tell which enzyme came first but, as I explained above, there is good reason to believe that the E2 from PDC and the E2 from OGDH evolved from the E2 gene for BCOADH via successive duplications.
Recall that the E2 subunits form the core of the complex (left). They contain the lipoamide swinging arm that carries substrate to three different active sites. The E3 subunits of the three enzymes are identical. There is only one E3 gene and it supplies the dihydrolipoamide dehydrogenase activity for BCOADH, PDC, and OADH.
The situation with the E1 subunit is more complicated. This is the part of the enzyme that recognizes the different types of substrate (e.g. pyruvate, 2-oxo acids, 2-oxoglutarate) so it makes sense that the three enzymes have different E1 subunits. All the eukaryotic versions of the PDC E1 subunit are related to the E1 subunit from BOADH. So are most of the bacterial versions. Other bacterial versions of the PDC E1 subunit are not related to those of the other enzymes (Schreiner et al. 2005).
The conclusion from the molecular data is that the pyruvate dehydrogenase complex evolved from the branched chain 2-oxo acid complex about 2 billion years ago. Subsequently, in some bacterial lineages a different E1 subunit replaced the one that's homologous to the BCOADH subunit. The α-proteobacteria and cyanobacteria lineages that gave rise to mitochondria and chloroplast respectively, retained the PDC E1 subunit that is related to BCOADH enzymes. This explains the eukaryotic versions of PDC.
1. This is a common theme in the evolution of metabolic enzymes. The evidence suggests strongly that amino acid metabolism is more ancient than most carbohydrate metabolism.
Schreiner, M.E., Fiur, D., Holatko, J., Patek, M. and Eikmanns, B.J. (2005) E1 enzyme of the pyruvate dehydrogenase complex in Corynebacterium glutamicum: molecular analysis of the gene and phylogenetic aspects. J Bacteriol. 187:6005-18.
Schnarrenberger, C. and Martin, W.. (2002) Evolution of the enzymes of the citric acid cycle and the glyoxylate cycle of higher plants. A case study of endosymbiotic gene transfer. Eur J Biochem. 269:868-83.