Jump to the article here. DOI: 10.13140/RG.2.2.20919.89765
J.Konstapel,Leiden,27-10-2026
Cause: Leidse onderzoeksprojecten ontvangen Open Competitie financiering voor vernieuwend onderzoek
Tuberculosis remains the deadliest infectious disease
attributable to a single pathogen.
Our only licensed vaccine, BCG, is a century old.
It protects infants against severe, disseminated forms of childhood disease.
It does not reliably protect adolescents and adults
against pulmonary tuberculosis —the very form that keeps the epidemic alive.
The field’s own literature states the obstacle plainly.
Vaccine development is held backbecause we do not know what protection against
Mycobacterium tuberculosis (Mtb) actually requires
(Dijkman, Coler and Joosten, 2022).
That is a remarkable admission from a mature discipline.
Modern immunology can enumerate immune cells,
sequence their receptors, measure their cytokines,
and track them over time with spectral flow cytometry.
What it cannot do is say
which of these measurements means “protected”.
The problem is not a shortage of numbers.
The problem is that the numbers do not converge
on a quantity that behaves like protection.
A second anomaly sits beside the first.
Roughly ninety-five per cent of people infected with Mtb
never develop the disease.
About five per cent do — sometimes decades later.
The protected majority are not cured;
the organism persists.
They are not ill either.
They occupy a state that is stable for a lifetime
and yet is not resolution.
The field names this state awkwardly: “latency”.
The clinical boundary between latent infection and active disease
is notoriously difficult to draw.
A system that holds an unresolved condition stable for forty years
is doing something
that the language of presence and absence
does not describe well.
2. The empirical material: the Leiden programme
The source text builds its argument on a concrete body of work:
the research programme on human anti-mycobacterial immunity
at the Leiden University Center for Infectious Diseases,
associated in particular with Simone A. Joosten.
Four features of that programme matter.
None of them is unusual.
Each is a move away from component thinking,
made for empirical reasons,
inside a field still organised around components.
2.1 A reading channel that is the same for everyone
Classical antigen presentation runs through
HLA class Ia and class II molecules —
among the most polymorphic loci in the human genome.
Any vaccine that depends on them
depends on a lock that differs from person to person.
HLA-E is different.
It is virtually monomorphic.
It presents both self-derived and pathogen-derived peptides.
It is not downregulated in HIV co-infection.
T cells that recognise HLA-E-presented peptides
are therefore called donor-unrestricted:
the restricting element barely varies across the population
(Ruibal et al., 2021; Voogd et al., 2024b).
The significance is architectural, not merely immunological.
Beside the individual-specific reading layer
runs a second reading layer invariant across the species.
Vaccine design aimed at HLA-E
is design aimed at the invariant layer.
Design aimed at classical HLA
is design aimed at the variable one.
2.2 Many paths, one connection point
When CD8+ T cells recognising HLA-E tetramers
folded with Mtb-derived peptides
were sorted from twenty-one individuals
and their receptors sequenced,
the repertoire proved highly diverse
(Voogd et al., 2024a).
One nearly invariant presenting molecule;
an open-ended variety of receptors arriving at it.
Earlier work had already shown
that these cells’ functional phenotype
is unorthodox relative to classical class I-restricted T cells:
they can secrete cytokines
associated with both conventional helper programmes.
Read as a diagram rather than a mechanism,
this is a connection point
with many curves terminating on it.
The point is fixed; the approaches are not.
The field calls this a repertoire-characterisation problem.
It can equally be described as the ordinary situation
when a system is organised around closure:
what has to be the same is the junction,
not the route taken to reach it.
2.3 Function rather than quantity
In tuberculosis, antibody titres
have long been a poor guide to anything.
A recent consensus review from the same group
reframes the question around functional capacity —
what antibodies do rather than how much of them is present —
with glycosylation and effector function
carrying the discriminating information
(van Meijgaarden et al., 2026;
see also Rijnink, Ottenhoff and Joosten, 2021).
Joosten’s own statement of method is explicit:
evaluate the sum of immune responses
rather than isolated components,
and assess that sum from a functional perspective.
This is the methodological core of the whole argument,
stated by a practitioner on empirical grounds.
A quantity measured at a point
is being replaced by a property of a completed relation.
The move is forced by the data:
the component measurements did not behave.
The functional ones do.
2.4 The granuloma as a held state
Within the granuloma, the local architecture of cellular interaction
tracks with outcome:
interactions between natural killer cells and macrophages
correlate with limited pathology
(Niewold et al., 2025).
The granuloma is not a storage container for bacteria.
It is a configuration under tension
that neither resolves into clearance
nor collapses into disease.
It is maintained by continuous activity,
and it can persist for decades
before doing one or the other.
3. The reading: protection as closure
The source text borrows a small, positional vocabulary
from the Vacuum.Net framework:
points in a position space,
curves running between them,
and connection points where curves meet
and a path can close.
No physical claim is required.
The vocabulary is used as a geometry of relation.
In that vocabulary:
- A component measurement is the measurement of a point.
- A functional assay is the measurement of a segment of curve.
- Protection — if the reading is right — is neither.
It is the property that a particular path closes.
Closure is not located anywhere on the path.
It is not the sum of the properties of the points,
and it cannot be recovered
by measuring the points more precisely,
because precision at a point
says nothing about whether the circuit returns.
Four consequences, all observed
If that is what protection is,
four things follow immediately.
All four are observed.
First, no single correlate of protection will be found,
because the thing being correlated
is not a quantity carried by any element.
A century of negative results
is then not a record of insufficient resolution
but a correct empirical answer to a malformed question.
Second, the same intervention
will protect in one host state and not in another.
Closure is a property of the path as a whole,
and the path includes the state of the system
at the moment of the attempt.
BCG in an infant and BCG in an adolescent
are not the same path with the same agent.
They are different paths.
The field’s central vaccine anomaly
becomes, on this reading, not an anomaly at all.
The related observation —
that repeated exposure to Mtb and to environmental mycobacteria
shapes subsequent host responses
and may shift vaccine efficacy in either direction —
belongs to the same point:
the history of the path is part of the path.
Third, invariant presentation elements
will be unusually productive targets,
because they are the connection points
that every individual shares.
This is the design logic of the HLA-E programme,
arrived at independently.
Fourth, latency is neither a weak form of disease
nor a cured state.
It is a path held open under tension that has not discharged.
Disease is the discharge.
This removes the need to treat
the latent-active boundary as a diagnostic defect.
The two are not two states
but two positions on one curve —
which is why no test cleanly separates them.
4. What this is, and what it is not
Nothing in the Leiden work refers to any of this.
No part of the argument is evidence for the Vacuum.Net framework.
What is shown is narrower,
and, in the framework’s own terms,
is the only thing that was being asked:
a well-developed experimental programme
can be restated without remainder
in the language of points, curves and connection points;
and when it is restated,
its two standing anomalies —
the missing correlate and the age-dependence of BCG —
stop being anomalies and become consequences.
Restatement is cheap if it only relabels.
The test of whether it does more
is whether it forbids anything.
This one does.
5. A consequence that can fail
The reading issues a prohibition
rather than a prediction of the usual kind.
It holds that no component measured at a single point —
no titre, no cell frequency,
no transcriptional signature evaluated as a quantity —
will ever function as a general correlate of protection against Mtb
across host states.
Every apparent success of this kind
will turn out, on extension,
to be a correlate of a particular path
in a particular population.
Conversely, measures constructed as closure measures —
whether a given effector relation completes,
assayed functionally,
across the whole response rather than within one arm of it —
will generalise where the component measures do not.
This is falsifiable in the ordinary way.
A single robustly transferable component correlate,
holding across ages, co-infections,
prior exposure histories and populations,
would refute it.
The field has searched for exactly that
for a hundred years without finding it —
weak evidence, admittedly.
The argument is that the absence is structural
and should be expected to continue.
There is a constructive side to the same statement.
If protection is closure,
then the design question for a vaccine
is not which antigen to present
but which junction to make reachable
from the greatest number of starting states.
An invariant, population-wide presentation element
with a diverse set of routes converging on it
is precisely such a junction.
That is an argument for the HLA-E direction
that does not depend on any mechanism specific to HLA-E —
and it would carry over
to any other invariant element that is found.
6. Summary for the practitioner
Three sentences carry the essay’s practical weight.
- A century of failed searching for a correlate of protection
is better read as a correct answer
to a malformed question
than as a failure of measurement resolution. - Functional, relation-level assays
should be expected to generalise
where point-level quantities do not;
the field’s own methodology is already moving this way. - Vaccine design should target invariant junctions
reachable from many starting states —
a design principle, not a mechanism claim.
Annotated references
The following list reproduces the sources cited in the original text,
with a short annotation on each.
Where a source underwrites a specific claim above,
the annotation says which one.
Dijkman, K., Coler, R. N., and Joosten, S. A. (2022).
Editorial: Beyond Th1 — novel concepts in tuberculosis vaccine immunology.
Frontiers in Immunology, 13, 1059011.
An editorial statement of the field’s central admission:
progress in tuberculosis vaccine development
is constrained by the absence of knowledge
about what protection actually requires.
It supplies the opening premise of the argument —
that the gap is conceptual, not merely technical.
van Meijgaarden, K. E., Grace, P. S., Freund, N. T., Achkar, J. M.,
Lindenstrøm, T., Tan, J., Chan, J., King, C. G., Ottenhoff, T. H. M.,
and Joosten, S. A., on behalf of the CTVD B cell group (2026).
Antibodies in tuberculosis: functional capacity as key determinant.
Infection and Immunity, 94, e00394-25.
A consensus review reframing humoral immunity in tuberculosis
around what antibodies do rather than how much is present.
It is the clearest published instance
of quantity being replaced by function —
the methodological move the essay generalises.
Niewold, P., IJsselsteijn, M. E., Dijkman, K., de Miranda, N.,
Ottenhoff, T. H. M., Verreck, F. A. W., and Joosten, S. A. (2025).
NK cell–macrophage interactions in granulomas correlate with limited
tuberculosis pathology. bioRxiv, 2025.02.14.638234.
Evidence that the granuloma’s cellular architecture —
specifically NK–macrophage interaction —
tracks with clinical outcome.
It grounds the reading of the granuloma
as a held configuration under tension
rather than a passive containment vessel.
Rijnink, W. F., Ottenhoff, T. H. M., and Joosten, S. A. (2021).
B-cells and antibodies as contributors to effector immune responses
in tuberculosis. Frontiers in Immunology, 12, 640168.
An earlier statement of the position
that antibodies contribute functionally to tuberculosis immunity,
predating and supporting the 2026 consensus review.
Ruibal, P., Voogd, L., Joosten, S. A., and Ottenhoff, T. H. M. (2021).
The role of donor-unrestricted T-cells, innate lymphoid cells,
and NK cells in anti-mycobacterial immunity.
Immunological Reviews, 301(1), 30–47.
The programme’s overview of donor-unrestricted T-cell biology.
It establishes the biological facts behind
the “invariant reading layer”:
HLA-E’s near-monomorphism and its stability in HIV co-infection.
Voogd, L., Drittij, A. M. H. F., Dingenouts, C. K. E.,
Franken, K. L. M. C., van Unen, V., van Meijgaarden, K. E., Ruibal, P.,
Hagedoorn, R. S., Leitner, J. A., Heemskerk, M. H. M., Davis, M. M.,
Ottenhoff, T. H. M., and Joosten, S. A. (2024a).
Mtb HLA-E-tetramer-sorted CD8+ T cells have a diverse TCR repertoire.
iScience.
The tetramer-sorting and sequencing study
showing one nearly invariant presenting molecule
served by a highly diverse receptor repertoire.
The empirical basis for the
“many paths, one connection point” observation.
Voogd, L., Riou, C., Scriba, T. J., van Wolfswinkel, M.,
Ottenhoff, T. H. M., and Joosten, S. A. (2024b).
HLA-E/Mtb specific CD4+ and CD8+ T cells have a memory phenotype
in individuals with TB infection.
Frontiers in Immunology, 15, 1505329.
Further characterisation of HLA-E-restricted, Mtb-specific T cells,
including their memory phenotype in infected individuals.
Supports the invariant-channel reading
together with Ruibal et al. (2021).
Konstapel, J. (2026). Curves in the Vacuum.Net.
constable.blog, 24 October 2026.
The author’s prior exposition of the framework vocabulary
of points, curves and connection points
used as the essay’s geometry of relation.
Konstapel, J. (2026). Where Illness Comes From:
The Net from First Chemistry to the Human. Eighth edition.
The longer work whose invariant-versus-variable distinction
the essay notes was reached from an entirely different direction
than immunology —
and which the Leiden programme’s architecture
independently mirrors.