Immunity

A key component of this project is the resistance to and tolerance of Borrelia infection in Peromyscus mice. As such, we obtained from NEON ear tissue samples from 976 individuals across three seasons of their collection efforts (2022 - 2024). From these samples, we measured the expression of six important genes thought to be associated with resistance and tolerance (Table 3). We also quantified Borrelia infection and burden, using Droplet Digital Polymerase Chain Reaction (ddPCR) techniques. The ddPCR variant is more sensitive than the traditional PCR that NEON uses to determine infection status3. Gene expression and Borrelia infection information were collected at the University of South Florida (USF). Due to limitations in the amount of tissue available in 2022, ddPCR was not conducted in this year. Additional tissue from each individual was obtained in 2023 and 2024 to facilitate this. Because the expression (and burden) data are highly right-skewed (many low-values), we present and analyze \(log(x + 1)\) transforms of these variables throughout this project.


Table 3: Resistance and tolerance genes, the triplex batch on which they were analyzed, and their expected immune function.
gene triplex group
IL-10 1 resistance
IFN-y 1 resistance
IL-6 1 resistance
TGF-B 2 resistance
TLR-2 2 tolerance
GATA-3 2 tolerance

Genetic correlations

First, we look at how the expression of these genes traits intercorrelated among these individuals. Figure 2 shows that the genetic traits are nearly all significantly intercorrelated (the lone exception being TGF-B and IL-6, which are uncorrelated). Notably, the resistance traits (IL-10, IFN-y, IL-6, TLR-2) are most strongly correlated with each other and, similarly, the tolerance traits (TGF-B and GATA-3) are most strongly correlated with each each other.

Correlations among log-transformed genetic expression traits. IL-10, IFN-y, IL-6, and TLR-2 are thought to be resistance-associated genes and TGF-B and GATA-3 are throught to be tolerance-associated.

Figure 2: Correlations among log-transformed genetic expression traits. IL-10, IFN-y, IL-6, and TLR-2 are thought to be resistance-associated genes and TGF-B and GATA-3 are throught to be tolerance-associated.

PCA

The resistance and tolerance genetic traits also broadly group together in a principal component analysis (PCA; Figure 3), with “resistance” expression aligning with the first principal component (PC) and “tolerance” expression aligning with the second PC. Overall, these first two PCs explain 87% of the variation among the genetic expression variables (Table 4). It also appears that Borrelia infection status is not strongly associated with either resistance or tolerance, but Borrelia burden may be associated with tolerance expression (Figure Figure 3, blue lines).

Principal component analysis of genetic expression traits. Traits were log(x+1) transformed prior to the analysis.

Figure 3: Principal component analysis of genetic expression traits. Traits were log(x+1) transformed prior to the analysis.


Table 4: Importance of genetic expression principal components.
PC1 PC2 PC3 PC4 PC5 PC6
Eigenvalue 3.44 1.75 0.56 0.16 0.06 0.03
Proportion Explained 0.57 0.29 0.09 0.03 0.01 0.01
Cumulative Proportion 0.57 0.86 0.96 0.98 0.99 1.00


Table 5: Association of Borrelia burden and infection status with genetic expression (principal components).
variable PC1 PC2 type perms rsquared pvals
log_burden 0.08 1.00 vector 999 0.02 0.03
Bb_statusNegative 0.02 -0.13 factor 999 0.02 0.00
Bb_statusPositive 0.01 0.00 factor 999 0.02 0.00


Multivariate regression of burden and infection against the first two genetic expression PCs implies that only the “tolerance” genes (PC2) are
significantly associated with the Borrelia infection or burden (6). This approach is admittedly simplistic, since we aren’t accounting for spatial or temporal variation. These relationships will be assessed more vigorously in later sections.


Table 6: Multivariate principal component regression (PCR) of Borrelia infection status and burden.
Response Predictor Coef SE t P r-squared
Bb_infected (Intercept) 0.56 0.03 20.12 0.00 0.05
Bb_infected expr_PC1 -0.03 0.09 -0.37 0.71 0.05
Bb_infected expr_PC2 0.37 0.09 4.04 0.00 0.05
log_burden (Intercept) 1.80 0.12 15.42 0.00 0.02
log_burden expr_PC1 0.08 0.36 0.23 0.82 0.02
log_burden expr_PC2 1.05 0.38 2.74 0.01 0.02

Infection sample sizes

This project measured gene expression, but not Borrelia in 2022. NEON did measure infection during that year, but we can only compare Borrelia-positive samples across methods (we can be confident that a “positive” is correct, but a “negative” is not certain with NEON’s PCR technique). Adding these samples would only net us 5 total samples (7), which is likely not worth the cost of the new “method” variable that would be required to differentiate between the two infection screening methodologies.

Table 7: Tabulation of expression, behavior, and infection samples across NEON plots. Columns correspond to the year, sand plot by which counts are grouped; the number of samples for which we have expression, capture timing, and infection data (either from USF or NEON); and the samples gained by including NEON infection data.
year plotID expression captime USF NEON added
2022 BLAN_009 11 64 0 4 0
2022 HARV_008 36 264 0 4 0
2022 MLBS_006 21 240 0 4 0
2022 ORNL_003 5 36 0 3 1
2022 SCBI_008 17 167 0 3 0
2022 SERC_015 8 64 0 4 2
2022 STEI_012 64 235 0 4 2
2022 UNDE_027 45 184 0 0 0
2023 BLAN_009 19 67 17 0 0
2023 HARV_008 4 5 0 0 0
2023 MLBS_006 44 272 32 0 0
2023 ORNL_003 7 25 8 0 0
2023 SCBI_008 23 121 16 0 0
2023 SERC_015 13 103 6 0 0
2023 STEI_012 99 406 52 0 0
2023 UNDE_027 20 109 14 1 0
2024 BLAN_009 34 161 60 0 0
2024 HARV_008 38 149 44 0 0
2024 MLBS_006 64 270 70 0 0
2024 ORNL_003 19 46 22 0 0
2024 SCBI_008 18 101 28 0 0
2024 SERC_015 9 128 35 0 0
2024 STEI_012 20 278 79 0 0
2024 UNDE_027 12 136 56 0 0

  1. Citation needed↩︎