B cells, antibodies, and vaccination

Antibodies co-evolve with pathogens

B cells determine antibody diversity

Structure of a B cell receptor

Origins of B cell receptor diversity

VDJ recombination

Affinity maturation

VDJ loci

VDJ recombination

Selection on naive repertoire

Stability

Autoreactivity

Germinal centers

Affinity maturation

GC B cells evolve

AID induces somatic hypermutation

Mutation rate varies over the variable region

GCs produce plasma and memory cells

In humans, memory cells evolve further

Evidence from sampling lymph nodes

Clones can be long-lived

Broadly neutralizing antibodies are often highly diverged

Broadly neutralizing antibodies to flu elusive

Stalk might not be "evolution-proof"

Historically conserved sites might not be constrained

Response to 2009 H1N1 vaccine

Stalk responses can fade

Stalk antibodies bind with lower avidity

Stalk antibodies are more polyreactive

Birth year differences in stalk targeting

Early infections affect vaccine response

A general model of sin

In 2013, H1N1 acquired K166Q

Uneven recognition

Homology with childhood strain

Recapitulation in ferrets

What determines Ab immunodominance?

Host genotype

Order (Age?) of infection

Immunogenicity of antigen (e.g., adjuvanted vaccine vs. mild infection)

Immunogenicity of epitopes

Autoreactivity

Measuring the impacts of vaccination

Definitions of "VE"

Vaccine efficacy: derived from clinical trial with randomization

Vaccine effectiveness: derived from observational data (no control over who's vaccinated)

This distinction is critical

Voluntary vaccinees are very different from non-vaccinees

(also, people who enroll in clinical trials are different from those who don't)

Efficacy (or effectiveness) against what?

VE often comes from observational studies

Trials can be unethical if recommendation in place

Trials can be very expensive

Outpatient and hospitalization data are "convenient"

Common observational study approach: test-negative design

Prospective vs. test-negative design

Some implicit assumptions of TND

Vaccine is not "leaky" (all-or-nothing protection)

Decision to vaccinate is random

We can mitigate a few problems by statistically adjusting for or stratifying by comorbidities, age, risky behavior, etc., so we can estimate their effects (purely associational) separately---but we still can’t measure critical things, and we don’t know everything we need to measure.

What is the reference group?

Intuitively, "you" without vaccination, but we can't measure that!

Practically, an unvaccinated person (of the same age, health status, risk appetite...)

... who was recently infected?

... an "average" unvaccinated person?

Influenza vaccine effectiveness is low

H3N2: 33% (95% CI: 26-39)

B: 54% (46-61)

pH1N1: 61% (57-65)

H1N1: 67% (29-85)

What's going on?

Poor strain selection (antigenic mismatch)

Egg adaptations (antigenic mismatch)

Interaction with pre-existing immune responses

Mismatch does not drive low VE

OAS might "hide" strain updates

Another curiosity: low VE in repeat vaccinees

In U.S., flu vaccine coverage is high(-ish)

In U.S., most vaccinees are repeat vaccinees

(Vaccinated this year and last year)

Potential explanations for apparent increase in infection risk in repeat vaccinees

1. Failure to adjust for timing of vaccination + waning: modest effect
2. Failure to account for nonrandom switching: modest effect, wrong direction.

Prior infection influences vaccine responses

Impact of recent infections in Ha Nam

Past infections could explain increased infection risk in repeat vaccinees

Could this explain differences in immunogenicity?

Innate immune signatures of vaccine responses