This study included 33 male Division I collegiate football players, four of whom suffered from a diagnosed concussion while enrolled in the study. Details of the study population are provided in Table 1 and the Methods section. Three data collection timepoints were established: two during the athletic activities (mid- and post-season) and an off-season collection 86 days after the last game of the season (Fig. 1). Additional data collection was performed for the concussed players within 48 hours after diagnosis.
Microbial DNA was extracted from fecal samples, and next-generation 16S rRNA gene sequencing was performed using the Nanopore MinION platform, which has been shown to provide a more accurate taxonomic classification at the species level than Illumina MiSeq[38]. Long-read 16S rRNA amplicons covering the V1-V9 hypervariable region generated reads of approximately 1500 bp. Sequences were then trimmed using Porechop and classified with Kraken [36], which yielded 23,249,965 classified reads and led to 1,021 species being identified after quality filtering. For the longitudinal analysis of gut microbiota changes across the football season, we excluded any samples collected following a concussion and subsequently grouped all remaining samples according to the three collection timepoints (mid-season, n=29; post-season, n=22, and off-season, n=21). Then, we investigated the differences in gut microbiota community structure across timepoints by calculating beta and alpha diversity metrics. Plotting of weighted UniFrac distances by principal coordinate analysis (PCoA) revealed overlap between the three timepoints, with no statistically significant differences evaluated by analysis of similarity (ANOSIM) (R=0.0064, p=0.3622; Fig. 2a). Further, the alpha diversity at the species level based on the Simpson and Shannon metrics did not significantly differ between the time points (p=0.178 and p=0.425, respectively) (Fig. 2b). Next, we evaluated the gut microbiota composition at the phylum, family, and genus level for all three-time points. Overall, the relative abundance of prevalent taxa was similar between the mid- and post-seasons. However, some shifts were observed in the off-season compared to the other timepoints (Fig. 2c). A random-effects multivariate analysis that only included data from non-concussed athletes who had provided a sample at all three timepoints (n=17) was performed using MaAsLin2. Mid- to post-season and post- to off-season were compared in two independent analyses (Supplemental Table 1). This model identified 8 bacteria species with a statistically significant change in abundance between the post- and off-season timepoints (i.e., having an FDR-corrected p-value <0.05). Supplemental Table 1 shows these species but also extends to show all species with an FRD-corrected p-value as high as 0.125. It is noteworthy that of the 17 species shown, all of them had decreased in relative abundance during the off-season (Fig. 2d). Microbial taxa at the family, genus, and species level meeting both statistical significance criteria and expressing prevalent abundance (mean > 1% in at least one group at family or genus level, >0.1% at species level) are represented in Fig. 2e. Of note, the species Anthrobacter sp. YN and Desulfirispirillum indicum decrease in relative abundance during the off-season compared to the post-season (q=0.027 and 0.039, respectively).
We next inferred the functional composition of the gut microbiota based on the 16S data using PICRUSt2[37]. Changes in abundance of MetaCyc pathways throughout the football season in non-concussed football athletes (n=17) were identified by two independent MaAsLin2 random-effects multivariate analyses comparing the mid- to post-season and post- to the off-season. Results revealed 251 pathways that were differentially abundant between groups, with all significant changes (q<0.05) occurring between the post- and off-season (Supplemental Table 3). Identification of the parent classes for the 251 differentially abundant pathways revealed that they are predominantly related to aromatic compound degradation, amino acid biosynthesis and degradation, fatty acid biosynthesis, and biosynthesis of electron carriers, such as quinol and quinone. In particular, the pathways with the highest fold changes between the post- and off-season are most notably involved in the degradation of sugars and aromatic compounds (Fig. 3).
Relationship between the gut microbiota and a single concussion
We next evaluated the changes in the gut microbiome associated with concussion. Since mid- and post-season samples proved similar in the longitudinal analysis, both timepoints were combined to form an “in-season” group. Samples were then divided into four groups: in-season (n=51), in-season concussion (n=4), off-season (n=21), and off-season concussion (n=4) (Fig. 1b). The in-season concussion group consists of the samples from the four athletes who suffered a concussion collected within 24 to 48 hours following the diagnosed concussion. In contrast, the off-season concussion group includes the samples from the same four athletes who received a concussion collected in the later off-season timepoint.
Beta diversity analysis showed significant differences in the gut microbiota structure between groups (R=0.2309, p<0.001). Pairwise ANOSIM analysis indicated that samples from concussed and non-concussed players were significantly different for both in- (R=0.3959, p=0.0127) and off-season (R=0.5711, p=0.0024) comparisons (Fig. 4a). Additionally, greater alpha diversity was observed in the concussed athletes when compared to their non-concussed teammates, for both the in-season (Simpson, p=0.0082; Shannon p=0.0043) and off-season (Simpson, p=0.0435; Shannon p=0.0329) comparison (Fig. 4b). The abundance of the most prevalent genera and families from the gut microbiota was similar within each group of participants (concussed and non-concussed athletes) between the in- and the off-season timepoints (Fig. 4c). At the phylum level, fecal samples collected post-concussion during the in-season exhibited an overall distinct distribution of the most abundant phyla compared to the rest of the groups represented (Fig. 4c).
A fixed-effects model in MaAsLin2 was used to identify specific microbial taxa with statistically significant differences in abundance between the concussed and non-concussed players at the in- and off-season timepoints (Supplemental Table 2). The model identified 9 species with a significant difference in the off-season model, while none showed significance in the in-season analysis. The relative abundance of the family Lachnospiraceae was decreased in the concussed athletes compared to their non-concussed teammates (q=0.003), whereas the families Ruminococcaceae (q=0.028) and Oscillospiraceae (q=0.028) were increased (Fig. 4d). At the genus level, Bacillus and Bacteroidetes were increased in the concussion group for the off-season analysis (q=0.025 and q=0.023, respectively). The species Eubacterium rectale and Anaerostipes hadrus, both belonging to the Lachnospiraceae family, significantly reduced relative abundance in concussed athletes (q=0.014 and q=0.034, respectively). On the contrary, seven bacterial species, including Ruminococcaceae bacterium CPB6 (q=0.007), Mageeibacillus indolicus (q=0.021), Ethanoligenes harbinense (q=0.007), Anaerococcus prevotii (q=0.032), Bacillus thuringiensis (q=0.014), Flavonifractor plautii (q=0.014) and Bacillus cereus (q=0.017), showed a greater relative abundance in the concussion group.
To investigate the functional changes in the gut microbiota following concussion, we examined pathway abundances generated with PICRUSt2. MaAsLin2 multivariate analysis between the concussed and non-concussed groups identified two and eight differentially abundant pathways in the in-season and off-season, respectively (Fig. 5, Supplemental Table 3). These results indicate that the major changes between off-season concussion and non-concussion included: acetylene degradation, N10-formyl-tetrahydrofolate biosynthesis, NAD salvage pathway, thiamin salvage, flavin biosynthesis, D-galacturonate degradation, beta D-glucuronide and De-glucoronate degradation, hexuronide and hexuronate degradation, and adenosine nucleotides degradation. On the other hand, the major changes between the in-season concussion and in-season included: coumarins biosynthesis and 1, 3-propanediol biosynthesis (Fig. 5). Notably, among the pathways altered in the off-season, sugar acid degradation pathways had significantly reduced relative abundances in the concussion group compared to the non-concussed athletes.
No alterations in the oral microbiota during the longitudinal study or after a concussion.
To study the changes in the oral microbiome taking place as the football season progressed, we analyzed 16S rRNA gene sequencing data from the saliva of non-concussed players (mid-season, n=11; post-season, n=8; and off-season, n=5) (Fig. 1b). Nanopore MinION long-read 16S rRNA sequencing yielded 6,967,795 classified reads with 793 species identified after quality filtering. The structure of the saliva microbial community was similar between groups, as indicated by the beta diversity analysis (R=-0.0178, p=0.5415) (Fig. 6a). Similarly, no significant differences in alpha diversity were observed across the three timepoints analyzed (Simpson, p=0.779; Shannon, p=0.975) (Fig. 6b). The overall shifts in the oral microbiota composition at the phylum, family, and genus level do not appear to be consistent across the three timepoints analyzed (Fig. 6c).
We next assessed whether the oral microbiome was altered in concussed athletes (in-season, n=4; off-season, n=3) compared to their non-concussed teammates (in-season, n=19; off-season n=5) (Fig. 1b). No significant differences were found in either beta diversity (R=-0.0464; p=0.6387; Fig. 6d, e) or alpha diversity (Simpson, p>0.999; Shannon p>0.999 in-season, p=0.989 off-season). At the phylum, family, and genus level, the most abundant taxa are overall comparable in abundance between the in- and the off-season within the same group of athletes (concussed and non-concussed) (Fig. 6f). In conclusion, saliva microbiota biomarkers remained close to their baseline values through all timepoints and after a concussion.
No changes in the optic nerve sheath diameter or cerebral blood flow measurements.
There were no significant differences in the optic nerve sheath diameters (ONSD) or transcranial Doppler (TCD) parameters in the right (R) and left (L) cerebral blood flow velocities (CBFV) (velocity without breath-holding (BH), the velocity with BH, holding time, anterior cerebral artery (ACA), posterior cerebral artery (PCA), internal carotid artery (ICA), L and R PI without BH, and with BH, and diameters of the optic nerve sheath (ONSD) for R and L (diameter and BHI). A single sonographer assessed TCD measurements at the post-season and off-season timepoints (Table 2). No clinical concussion participants underwent clinical TCD testing, and no meaningful differences in velocity or diameter were observed across the timepoints.
Table 2
Transcranial doppler imaging optic nerve sheath diameter statistics. Right (R) and Left (L), breath-holding (BH), anterior cerebral artery (ACA), posterior cerebral artery (PCA), internal carotid artery (ICA), optic nerve sheath (ONSD).
R Velocities |
Group | Off-season | Post-season | P-Value |
R Velocity without Holding (Mean, SD) | 56.6 ± 12.8 | 57.6 ± 15.1 | 0.88 |
R Velocity with Holding (Mean, SD) | 70.7 ± 16.7 | 76.4 ± 20.6 | 0.51 |
R Holding Time (Median, IQR) | 20 [20–20] | 20 [20–20] | 1 |
R ACA (Mean, SD) | 51.6 ± 11.7 | 55.1 ± 8.2 | 0.5 |
R PCA (Mean, SD) | 36.9 ± 5.6 | 31.4 ± 5.4 | 0.04 |
R ICA (Mean, SD) | 31 ± 4.9 | 33.3 ± 4.2 | 0.25 |
RPI without holding (Mean, SD) | 1 ± 0.1 | 1 ± 0.1 | 0.26 |
RPI with holding (Mean, SD) | 0.9 ± 0.1 | 0.8 ± 0.2 | 0.08 |
L Velocities |
Group | Off-season | Post-season | P-Value |
L Velocity without Holding (Mean, SD) | 57.1 ± 12.6 | 59.1 ± 10.2 | 0.71 |
L Velocity with Holding (Mean, SD) | 74.4 ± 13.1 | 73.3 ± 16.4 | 0.88 |
L Holding Time (Median, IQR) | 20 [20–20] | 20 [20–20] | 1 |
L ACA (Mean, SD) | 49 ± 9.2 | 47.6 ± 6.9 | 0.72 |
L PCA (Mean, SD) | 35.2 ± 5.5 | 34.2 ± 5.7 | 0.68 |
L ICA (Mean, SD) | 31.9 ± 3.8 | 32.8 ± 5.1 | 0.66 |
LPI without holding (Mean, SD) | 1 ± 0.1 | 1 ± 0.2 | 0.9 |
LPI with holding (Mean, SD) | 0.8 ± 0.1 | 0.8 ± 0.1 | 0.98 |
Diameters ONSD |
Group | Off-season | Post-season | P-Value |
R Diameter (Mean, SD) | 3.6 ± 0.2 | 3.6 ± 0.3 | 0.68 |
R BHI (Mean, SD) | 1.2 ± 0.5 | 1.4 ± 0.5 | 0.5 |
L Diameter (Mean, SD) | 3.6 ± 0.3 | 3.7 ± 0.4 | 0.88 |
L BHI (Mean, SD) | 1.6 ± 0.7 | 1.5 ± 0.5 | 0.59 |
Changes in blood biomarkers across the football season.
Eleven participants provided blood samples at each of the three timepoints, enabling us to perform a longitudinal analysis of blood serum biomarkers. Specifically, repeated measures ANOVAs were used to determine if the levels of several biomarkers in serum changed across the football season. Results revealed no significant differences for S100β (F(2, 16)=1.087, p=0.361), serum amyloid A (SAA) (F(2, 16)=1.433, p=0.268) and neurofilament light chain (NF-L) (F(2,16)=1.262, p=0.310) (Fig. 7). However, serum glial fibrillary acidic protein (GFAP) was significantly decreased in both the mid-season and the post-season compared to the off-season (p=0.00813 and p=0.0437, respectively; F(2,16)=6.597, p=0.0081) (Fig. 7c). We also compared the biomarker levels in serum in the four samples collected in-season after a concussion to the samples from the rest of the athletes. There was no statistical difference between the groups tested for S100β, SAA, GFAP, and NF-L (p=0.312, p=0.496, p=0.431, and p=0.616, respectively) (Fig. 7e-h). However, we did detect a significant correlation between the concentration of the serum biomarkers S100β and SAA in the longitudinal samples of non-concussed athletes with the abundance of the bacterium species Eubacterium rectale (Fig. 7i).