*4.5. Putative Protease Identification*

Annotations obtained from Prokka, PANNZER2, and eggNOG were consolidated in Table S4. Putative proteases were identified by matching annotations with the following keywords: Peptidase, protease, proteinase, sortase, caspase, penicillin-binding protein, insulinase, snapalysin, and mycosin. Each coding sequence that presented a keyword in at least one of the predictions was manually checked using BLASTp v.2. against the NR database, to confirm the proteolytic nature of the enzyme. In this work, the terms "protease" and "peptidase" have been used indistinctly [82].

#### *4.6. Classification of Protease Families and Identification of Protease Orthogroups*

For the classification into protease families, a reciprocal best hit search (E-value threshold, 1 × <sup>10</sup>−<sup>20</sup> of the three-strain dataset was performed against MEROPS, the peptidase database, release 11, using the HMMER-based web-server utility [83]. This database classifies peptidases into seven superfamilies based on the catalytic residue (Aspartic (A), Cysteine (C), Glutamic (G), Metallo (M), Asparagine (N), Serine (S), Threonine (T)) along with two superfamilies of Mixed (P) and Unknown (U) catalytic types, and further divides these superfamilies into 255 proteolytic families based on similarities in amino acid sequences [84]. Additionally, the curated list of proteases for the three strains was grouped into orthogroups using Orthofinder [42] with the default E-value 1 × <sup>10</sup>−<sup>40</sup> These were

named protease orthogroup or p-orthogroup. Venn diagram visualization of common p-orthogroups between strains was done using the package eulerr in R [85].

#### *4.7. Creation of Custom Databases: Functional Keratinases and Putative Non-Keratinases*

To identify putative keratinases in the genomes, two custom databases with sequences of functional keratinases and putative non-keratinases were constructed by retrieving sequences from the NCBI NR database (query date: March 2020). The functional keratinase database was built according to literature and contains 61 sequences from *Actinomadura*, *Amycolatopsis*, *Bacillus*, *Bifidobacterium*, *Brevibacillus*, *Deinococcus*, *Geobacillus*, *Lactobacillus*, *Meiothermus*, *Nocardiopsis*, *Streptomyces*, and *Thermus* (Table S7). In addition, to differentiate and discard those sequences that do not encode possible keratinolytic proteases, a putative non-keratinase database was created. Several reports indicate that keratin has high stability against common proteases, such as pepsin, papain, and trypsin [1,2]. Based on this information, a hypothetical non-keratinase database was constructed (Table S9). This database contains 50 protein sequences belonging either to the trypsin, papain, or pepsin families. All sequences are derived from Gram-positive bacteria belonging to the genera *Bacillus*, *Bifidobacterium*, *Brevibacterium*, *Chloroflexi*, *Clostridium*, *Corynebacterium*, *Enterococcus*, *Lactobacillus*, *Listeria*, *Staphylococcus*, *Streptococcus*, and *Streptomyces*.
