Ethical approval was obtained from the Northwestern University Institutional Review Board (IRB) (STU00209226) & University of Chicago IRB (IRB22-0595). Patients were consented and enrolled at the Northwestern University Cutaneous Lymphoma clinic and University of Chicago Radiation Oncology clinic between 2019 and 2023 in compliance with the Declaration of Helsinki. Demographic/clinical data and lesional and contralateral non-lesional skin samples were collected from 12 patients with biopsy-confirmed CTCL ( Supplementary Table S1 ). Six participants received local IR, and 6 patients received TSEBT ( Supplementary Table S2 ). Eleven patients were concurrently on topical steroids and 8 on systemic therapy (i.e., acitretin and interferon) but no new treatments were introduced within 6 months prior to or during the study interval and any concurrent therapies were present for at least 6 months prior to IR. Additionally, there were no changes in diet, bathing habits, or non-CTCL medications based on chart review and survey data collected during the study interval. Patients who utilized antibiotics within the 4 weeks prior to collections were excluded. Healthy controls (HC) were comprised of 25 volunteers without CTCL or other active skin disease ( Supplementary Table S3 ).

Skin samples from 12 CTCL patients and 25 HC were obtained through sterile swabs. All specimens were placed immediately in sterile cryovials and stored at -80°C until DNA extraction. Genomic DNA was extracted using a Maxwell 16 LEV Blood DNA Kit (Promega, Madison, WI) implemented on a Maxwell 16 Instrument, following the manufacturer’s instructions with minor modifications: a lysozyme incubation (10 ng/μl lysozyme; Thermo Fisher Scientific, Waltham, MA) for 30 minutes at 37°C and bead beating (40 seconds at 6 min/sec) using a FastPrep-24 System (MP Biomedicals, Irvine, CA). Homogenized samples were transferred to the Maxwell cartridges for final DNA purification.

Genomic DNA was prepared for sequencing using a two-stage amplicon workflow and targeting the V4 variable region of microbial 16S rRNA genes as described previously (10, 17, 18).

Sequencing resulted in a total of 9,257,316 reads with an average of 31,275 reads per sample. Forward (F) and reverse (R) reads were trimmed using cutadapt v3.5 to remove primer sequences (19). Processing, decontamination, and filtering were performed in a matter identical to our previous work examining nbUVB and the CTCL skin microbiome (10, 20–23). Following QC processing, decontamination, and filtering, there were a total of 5,125,206 merged read pairs with an average of 18,982 per sample. To maximize data retention, while removing uninformative patient samples, only samples with minimum 1000 reads following processing were retained. Patient samples were paired across disease status (lesional, non-lesional) and time (pre-IR, post-IR).

Genomic DNA was PCR amplified with primers CS1_tuf2-F (ACACTGACGACATGGTTCTACAACAGGCCGTGTTGA ACGTG) and CS2_tuf2-R (TACGGTAGCAGAGACTTGG TCTACAGTACGTCCACCTTCACG) targeting the Staphylococcus tuf gene (24, 25). Amplicons were generated using a two-stage PCR amplification protocol, as previously described and utilized in our previous work (10, 17).

Tuf2 sequencing generated a total of 10,289,743 reads with an average of 34,880 raw reads per sample. Primer trimming and denoising were accomplished using the same procedure as above for the 16S amplicon data with the following modifications: during filtering and trimming, the maxEE for all reads was set to 3,3 due to their increased length, read merging used default parameters, and the read length cutoff window range was 460-470 nucleotides for merged read pairs. Following processing there were a total of 3,914,382 reads retained for an average sample read count of 13,269. ASVs were taxonomically annotated using BLAST alignments against NCBI prokaryotic (nr) refseq database (online access 10 April 2023). Only sequences that were annotated as Staphylococcus were retained.

The samples in the cleaned ASV table were visually evaluated using phyloseq v1.42.0 (26). α-diversity metrics were generated using the ASV table rarefied to 1000 sequences. Differences in α-diversity between patient sets were calculated using Wilcoxon rank-sum non-paired tests from the stats R package while differences within patient-matched samples were calculated using Pairwise Wilcoxon rank-sum tests. β-diversity metrics were generated using the rarefied ASV table (27). Principal coordinate analysis (PCoA) with Bray-Curtis dissimilarity was performed to identify β-diversity using an ADONIS2 method with default parameters (28).

Differential abundance analysis was conducted by DESeq2 v1.44.0 using the non-rarefied, CLR-transformed ASV table, with ASVs removed if they had less than 4 counts or a prevalence below 10% across the sample set (29). A linear model was implemented within the approach to compare abundance of taxa among different groups. Significant ASVs were only considered if they achieved a false discovery rate (BH-FDR)-adjusted p-value of <0.05 (q-value).

The post-process tuf2 ASV and taxa table was filtered using phyloseq v1.38 (26). To identify the most abundant Staphylococcus species, data were centered log-ratio (CLR) transformed and the final species-level table including only species present at 10% abundance or greater was used for statistical analysis. Differences in abundance between patient groups were calculated using Wilcoxon rank-sum non-paired tests from the stats R package while Pairwise Wilcoxon rank-sum tests were used to calculate differences between patient-matched samples (27). Differential abundance analysis was conducted using the CLR transformed abundance table. A linear model as described above was implemented to compare abundance of the top Staphylococcus species among different groups. Significant ASVs were only considered if they achieved an FDR of <0.05.

The median age of CTCL patients was 56 years (range 21-77) and 9 patients (75%) were male. Ten (83%) patients had mycosis fungoides, and 2 had other CTCL types. Median time between pre-IR sample collection and IR was 20 days; median time between IR and post-IR sample collection was 30 days. Modified Severity Weighted Assessment Tool (mSWAT) decreased by 25.3 points, on average, after IR. Patients receiving local IR received an average of 9.5 Gray over 1-12 fractions; patients receiving TSEBT received an average of 20 Gray over 6-16 fractions. There were no significant differences between age, sex, race, Fitzpatrick skin type (FST), CTCL subtype, stage, or non-IR therapies between patient’s receiving local IR versus TSEBT (p>0.05) ( Supplementary Table S4 ). There were no statistically significant differences between HC and patient age, sex, race, or comorbidities (p>0.05) ( Supplementary Table S5 ).

Across all individuals, Staphylococcus was the most abundant genus pre- and post-IR, followed by Corynebacterium and Streptococcus ( Figure 1A ).

We first analyzed any skin (both lesional and non-lesional) exposed to any form of IR (both local IR and TSEBT). Phylogenetic diversity (α-diversity) was significantly higher with IR exposure (Observed p=0.0019), and community structure (β-diversity) differed significantly between skin with and without IR exposure (p=0.001) ( Figure 1B ). Next, we analyzed only lesional skin exposed to IR. In this analysis, there were no significant differences in α- or β-diversity between lesions with and without exposure ( Figure 1C ). Finally, we analyzed only non-lesional skin exposed to IR. Phylogenetic diversity was significantly higher with IR exposure (p=0.02), and β-diversity differed significantly between non-lesional skin with and without IR exposure (p=0.03) ( Figure 1D ).

When observing changes to taxa, healthy commensal bacteria such as Bifidobacterium and coagulase negative species S. pettenkoferi were significantly more abundant in skin with IR exposure, and known pathogenic bacteria such as Streptococcus and S. aureus were either significantly lower (p<0.01, q<0.05) or trended lower (p<0.05, q>0.05), respectively, in skin with IR exposure. ( Figure 1E ). Comparatively, very few bacteria with pathogenic potential were increased post-IR exposure. These included Trueperella and Porphyromonas which have only been reported to cause skin infections in rare opportunistic cases (30, 31).

Next, we compared α- and β-diversity between pre-TSEBT and post-TSEBT and between lesional and non-lesional skin. α-diversity was significantly higher post-TSEBT compared to pre-TSEBT in both lesional (Observed p=0.024) and non-lesional (p=0.0015) skin ( Figure 2A ). Conversely, α-diversity did not differ between non-lesional and lesional skin at either the pre-TSEBT (p=0.23) or post-TSEBT (p=0.07) time points ( Figure 2A ).

Pre-TSEBT versus post-TSEBT β-diversity differed significantly for lesional (p=0.001) and for non-lesional skin (p=0.005). By comparison, lesional versus non-lesional skin β-diversity did not differ significantly before (p=0.50) or after (p=0.21) TSEBT exposure ( Figure 2B ).

Pre- versus post-local IR did not demonstrate significant differences in α- or β-diversity ( Figures 2C , D ). This was true for both lesional pre- versus post-local IR and non-lesional pre versus post-local IR. Importantly, non-lesional pre- versus post-local IR serves as an internal control as these sites were not exposed to IR.

In lesional skin, Streptococcus, Acinetobacter, and Roseomonas were significantly less abundant post- versus pre-TSEBT (p<0.01, q<0.05) while S. lugdunensis trended lower (p<0.05, q>0.05). Healthy commensals including genera Peptococcus and Cutibacterium, and family Anaerovoracaceae were higher in post-TSEBT lesional skin (p<0.0001, q<0.0001), and genera Moryella, Anaerococcus, and DNF00809, a member of the Eggerthellaceae family, trended higher ( Figure 3A ). Other bacteria that increased post-TSEBT included Trueperella, Olsenella, and Porphyromonas.

In non-lesional skin, Rothia, Xanthomonas, and Streptococcus were increased pre-TSEBT relative to post-TSEBT (p<0.01, q<0.05). In contrast, Bifidobacterium, S. caprae, S. pettenkoferi, S. lugdunensis, Trueperella and several members of the order Lactobacillales were increased post-TSEBT relative to pre-TSEBT (p<0.01, q<0.05) ( Figure 3B ).

At the species level, S. aureus, S. lugdunensis, S. warneri, and S. caprae were significantly more abundant in pre-TSEBT lesional versus non-lesional skin (p<0.001, q<0.05) ( Figure 3C ). Post-TSEBT, there were no significant differences between lesional and non-lesional skin.

In treated lesions, Kytococcus and Turicella were lower post-local IR compared to pre-local IR (p<0.0001, q<0.0001) ( Figure 3D ).

Finally, we compared CTCL skin before and after any form of IR to matched HC skin. HC had significantly higher α-diversity compared to CTCL pre-IR (Observed p<0.0001), and microbial communities differed significantly (p=0.001) ( Figure 4A ). Relative abundances of S. aureus and S. haemolyticus were increased in CTCL patients, while S. hominis, Bifidobacterium, and several members of the order Lactobacillales were higher in HC. After IR, α-diversity no longer differed between HC and CTCL patients (p=0.53), whereas microbial communities remained distinct (p=0.001) ( Figure 4B ). DeSeq2 identified increased relative abundance of S. aureus, S. caprae, S. haemolyticus, and S. lugdunensis amongst CTCL patients. Comparison of pre-TSEBT patients versus HC and post-TSEBT versus HC revealed similar results to those seen in pre-IR versus HC and post-IR versus HC ( Figures 4C , D ).