Lactic acid bacteria (LAB) play a key role in many food fermentations. However, some LAB species can also cause food spoilage, e.g., through the formation of biogenic amines. Paucilactobacillus wasatchensis is a LAB that causes late gas production in Cheddar cheese, the molecular causes of which are not fully understood. This study reports on the ability of P. wasatchensis WDC04 to produce cadaverine and putrescine in broth supplemented with lysine and ornithine, as well as in a model cheese. The raclette-type semi-hard cheese produced with P. wasatchensis as an adjunct culture contained 1,085 mg kg−1 of cadaverine and 304 mg kg−1 of putrescine after 120 days of ripening. We identified two ornithine decarboxylase genes (odc) and a putrescine-ornithine antiporter gene (potE) in the genome sequence of P. wasatchensis. We could show that the two odc genes, which are located on two contigs, are contiguous and form the genetic cluster odc2-odc1-potE. Alignment searches showed that similar gene clusters exist in the genomes of Levilactobacillus paucivorans DSMZ22467, Lentilactobacillus kribbianus YH-lac9, Levilactobacillus hunanensis 151-2B, and Levilactobacillus lindianensis 220-4. More amino acid sequence comparisons showed that Odc1 and Odc2 shared 72 and 69% identity with a lysine and ornithine decarboxylase from Ligilactobacillus saerimneri 30a, respectively. To clarify the catalytic activities of both enzymes, the odc-coding genes were cloned and heterologously expressed as His-tagged fusion protein. The purified Odc1 protein decarboxylated lysine into cadaverine, while the recombinant Odc2 protein preferentially produced putrescine from ornithine but also exhibited low lysine decarboxylating activity. Both enzymes were active at pH of 5.5, a value often found in cheese. To our knowledge, this is only the second lysine decarboxylase in LAB whose function has been verified. The tandem arrangement of the genes in a single cluster suggests a gene duplication, evolving the ability to metabolize more amino. Divergent substrate preferences highlight the necessity of verifying the functions of genes, in addition to automatic annotation based on sequence similarity. Acquiring new biochemical data allows better predictive models and, in this case, more accurate biogenic amine production potential for LAB strains and microbiomes.