Bigger is Faster in the Adaptive Immune Response

TL;DR

The paper addresses why the initiation time of the adaptive immune response, $\tau_{init}$, is nearly invariant across mammals spanning vast body masses. It combines empirical scaling data for lymph node number and size with a diffusion-based analysis of T cell–DC search within LN and supporting agent-based simulations to derive how $\tau_{init}$ scales with mass, $M$, showing $\tau_{init} \propto M^{v-(t+d)}$ under broad assumptions. Key contributions include empirical estimates $N_{LN} \propto M^{1/2}$ and $V_{LN}$ scaling as $M^{1/2}\ln(M)$ or $M^{2/3}$, and the demonstration that larger LN contain more T cells and DCs, reducing first-contact times and enabling near-constant or faster IFCT in larger animals, especially for systemic infections. The results highlight the advantage of a distributed lymphatic network in maintaining rapid adaptive immunity across body sizes, with implications for cross-species disease dynamics and immunological design principles in large organisms. $\tau_{init}$, $N_{LN}$, $V_{LN}$, and the diffusion-based search framework are central to the analysis, linking organ scaling to functional invariance in immune timing.

Abstract

Zoonotic pathogens represent a growing global risk, yet the speed of adaptive immune activation across mammalian species remains poorly understood. Despite orders-of-magnitude differences in size and metabolic rate, we show that the time to initiate adaptive immunity is remarkably consistent across species. To understand this invariance, we analyse empirical data showing how the numbers and sizes of lymph nodes scale with body mass, finding that larger animals have both more and larger lymph nodes. Using scaling theory and our mathematical model, we show that larger lymph nodes enable faster search times, conferring an advantage to larger animals that otherwise face slower biological times. This enables mammals to maintain, or even accelerate, the time to initiate the adaptive immune response as body size increases. We validate our analysis in simulations and compare it to empirical data.

Figures (3)

• Figure 1: Simplified schematic of T cell activation by Dendritic Cells. 1) A pathogen infects tissue, in this case for illustration, the lung. 2) DC deliver the captured antigens from tissue through lymphatic vessels to draining LN. 3) DC display antigens in the LN. 4) Naïve T cells search for cognate antigens presented on the surface of DC. 5) T cell receptors recognize the cognate antigens upon encountering the antigen-bearing DC and get activated upon receiving the activation signal from antigen-bearing DC. 6) Activated T cells proliferate exponentially, and CD8+ T cells transform into cytotoxic T cells (CTLs) that travel through the bloodstream to the inflamed, infected area. 7) CTLs kill the infected cells that display cognate antigens. We model the timing of search and activation in steps 4 and 5, where the adaptive immune response is initiated; the timing of this process depends on LN size.
• Figure 2: Lymphoid organ scaling with mass. Each data point represents a species. Both axes are on a log scale. The dashed lines show the reported regression fits. (A) Spleen volume of 38 species is best fit by the regression $cM^\mu$ with $c=1.1609$ and exponent $\hat{\mu}=1.05$ (95% CI [0.95, 1.19]). (B) Number of lymph nodes for 10 species is best fit by $cM^\mu$ with $c=3.98$ and $\hat{\mu}=0.523$ (95% CI[0.4, 0.64]). (C) Lymph node volume for 16 species is equally well fit in two ways: (1) a theoretically motivated fit including a logarithmic term $c_1M^\mu\ln(c_2M)$ (red line) with $c_1=1$, $c_2=1$, and $\hat{\mu}=0.56$ (95% CI [0.18, 0.94]) and by (2) by a simpler scaling fit, $c_1M^\mu$ (green line) with $c_1=1$, and $\hat{\mu}=0.68$ (95% CI [0.51, 0.85]). The p-value of the exponents is significant at the 0.01 level.
• Figure 3: Theoretical predictions vs. agent‐based simulations for initial T cell–DC contact time ($\tau_{init})$. Simulation data (circles) are compared to predictions (dashed) and best‐fit curves (solid) under two volume hypotheses: (A) Systemic infection assuming $V_{LN}\propto M^{2/3}$: predicted $\tau_{\rm init}\propto M^{-2/3}$ (dashed red), fitted exponent $\hat{\mu}=-0.69\pm0.08$ (solid red); (B) Localized infection assuming $V_{LN}\propto M^{2/3}$: predicted $\tau_{\rm init}\propto M^{0}$ (dashed green), fitted $\hat{\mu}=-0.02\pm0.07$ (solid green); (C) Systemic infection assuming $V_{LN}\propto M^{1/2}\ln(cM)$: predicted $\tau_{\rm init}\propto M^{-1/2}\ln(cM)$ (dashed red), fitted $\hat{\mu}=-0.48\pm0.10$ (solid red); (D) Localized infection assuming $V_{LN}\propto M^{1/2}\ln(cM)$: predicted $\tau_{\rm init}\propto M^{0}\ln(cM)$ (dashed green), fitted $\hat{\mu}=0.04\pm0.10$ (solid green). Each circle represents the simulation result of 20 simulation replicates in panels (A,B) and 100 replicates in panels (C,D) at each estimated LN volume.