Aldo-keto reductases (AKRs) are distributed in three families and multiple subfamilies in mammals. The mouse Akr1b3 gene is clearly orthologous to human AKR1B1, both coding for aldose reductase, and their gene products show similar tissue distribution, regulation by osmotic stress and kinetic properties. In contrast, no unambiguous orthologs of human AKR1B10 and AKR1B15.1 have been identified in rodents. Although two more AKRs, AKR1B7 and AKR1B8, have been identified and characterized in mouse, none of them seems to exhibit properties similar to the human AKRs. Recently, a novel mouse AKR gene, Akr1b16, was annotated and the respective gene product, AKR1B16 (sharing 83% and 80% amino acid sequence identity with AKR1B10 and AKR1B15.1, respectively), was expressed as insoluble and inactive protein in a bacterial expression system. Here we describe the expression and purification of a soluble and enzymatically active AKR1B16 from E. coli using three chaperone systems. A structural model of AKR1B16 allowed the estimation of its active-site pocket volume, which was much wider (402 Å3) than those of AKR1B10 (279 Å3) and AKR1B15.1 (60 Å3). AKR1B16 reduced aliphatic and aromatic carbonyl compounds, using NADPH as a cofactor, with moderate or low activity (highest kcat values around 5 min−1). The best substrate for the enzyme was pyridine-3-aldehyde. AKR1B16 showed poor inhibition with classical AKR inhibitors, tolrestat being the most potent. Kinetics and inhibition properties resemble those of rat AKR1B17 but differ from those of the human enzymes. In addition, AKR1B16 catalyzed the oxidation of 17β-hydroxysteroids in a NADP+-dependent manner. These results, together with a phylogenetic analysis, suggest that mouse AKR1B16 is an ortholog of rat AKR1B17, but not of human AKR1B10 or AKR1B15.1. These human enzymes have no counterpart in the murine species, which is evidenced by forming a separate cluster in the phylogenetic tree and by their unique activity with retinaldehyde.