The reserve
Naïve / resting
Hasn't joined the fight. Sits quietly in the blood, the dimmest grip of all — pure potential.
CCR7IL7RTCF7
cytotoxic
baseline · all groups
It startswith one bite.
Dusk. A single mosquito, lighter than an eyelash. You never see it coming.
The reach
Every year, the virus in that bite finds about
0
~390M infections/yr (95% CrI 284–528M); ~96M with symptoms — Bhatt et al. High
Bhatt, Nature 2013 · WHO · at-risk
Bhatt, Nature 2013 · WHO · at-risk
The four
There isn't one dengue.
There are four.
1DENV-1
2DENV-2
3DENV-3
4DENV-4
Beat one of them, and you're immune to that one — for life.
The rule
So the first dengue is rarely the one to fear.
It's the second.
A different strain — recognised, and mishandled, by a body that thinks it has seen this before.
Years later
The fever climbs for days. Then, around the fifth — just as it breaks —
the blood vessels
can begin to leak.
It happens quietly — and no one quite knows why.
Severe dengue = plasma leakage from rising vascular permeability, in a critical window as the fever breaks (≈day 3–7). High
CDC · clinical course
CDC · clinical course
The usual suspect
The textbook blames the antibodies — old ones, left over from the first fight, now helping the new virus in.
It's true, and cruelly precise: at one narrow window of fading immunity they can lift the risk of severe disease several-fold. It even has a name — antibody-dependent enhancement.
But that's only half
the story.
At intermediate pre-existing antibody titres, severe-dengue risk peaked ~7.6× vs naïve children. High
Katzelnick, Science 2017
Katzelnick, Science 2017
The other half
The other half is a cell we were taught to trust.
The killer
T cell.
The immune system's assassin — it finds the cells the virus has hijacked, and kills them. We always assumed more of them, gripping harder, was better.
DENV-specific CD8+ T cells play a dual role — protective and pathogenic. High
Srikor et al., Nat Commun 2026
Srikor et al., Nat Commun 2026
The hunt
To find out whether our own assassins turn the second infection lethal, a team in Bangkok did something painstaking.
From people across the whole spectrum of the disease, they fished out the exact cells that recognise dengue —
0
cells. Read one at a time.2,397 DENV NS3-specific CD8+ T cells; plate-based single-cell RNA-seq (Smart-seq2). High
Srikor et al., Nat Commun 2026
Srikor et al., Nat Commun 2026
A killer T cell is a specialist. It's trained to recognise one tiny scrap of the virus — a fragment of a protein called NS3, barely ten amino acids long.
Catching the few cells that know that scrap, out of billions, takes a clever trick.
Here's how they did it ↓
The method · tap to open
How do you catch
one cell?
Step 01
Bait it
Glowing molecular bait, shaped like the virus.
Open
✦
Step 02
Sort it
One cell per well — and its grip strength logged.
Open
⊞
Step 03
Read it
The full gene programme of every cell.
Open
≋
Step 04
Place it
Match each cell to serotype, history, outcome.
Open
◎
HLA-A*11/A*24 tetramers (GTS, NYA epitopes) · index sort · Smart-seq2 · qPCR + PRNT. High
Srikor et al. 2026 · methods
Srikor et al. 2026 · methods
The people
The cells came from real patients, across the whole range of dengue —
4 people who never felt sick
7 people laid low by fever
5 people who bled
The ones who never felt it are the rarest catch in all of dengue research — in one five-year hunt, only eight were ever found.
Acute cohort: 4 asymptomatic, 7 dengue fever, 5 hemorrhagic fever. High · rarity: 8 viraemic-asymptomatic among 179 household contacts over 5 yrs. context
Srikor et al. 2026
Srikor et al. 2026
The readout
Then they lined all 2,397 cells up by how hard they gripped, and asked which kind of cell showed up in whom.
What fell out was the opposite of what immunology would predict.
Keep going↓
The 2,397 cells weren't one thing.
They sorted into
five kinds.
A spectrum — from calm, to lethal. Meet them, one by one.
Each kind is a different cell doing a different job — and two of them decide everything. Meet the cast.
calm
lethal
Swipe through the five — calm on the left, lethal on the right.
← swipe the cast →
Five transcriptomic subsets; cytotoxic potential highest in CX3CR1+, lowest in Naïve/CM (rank, Fig.2c). High
Srikor et al. 2026 · Fig.2
Srikor et al. 2026 · Fig.2
Two cells
Hold the two ends of that spectrum in your mind.
The calm one drifts in the open, holding the virus loosely. It does enough.
The killer clamps down hard — and presses itself against the vessel wall, where the bleeding will happen.
CX3CR1+ = terminally-differentiated effector/memory CTL; the intravascular, vessel-wall–homing tier. Med
Gerlach, Immunity 2016
Gerlach, Immunity 2016
The reveal
So which cell showed up in whom? Watch them cross.
CX3CR1+ killer (p=0.03)intermediate / calm (p=0.009)
The strongest response was in the sickest patients. The people who never felt sick were full of the calm cell. The people who bled were full of the killer.
Schematic of the reported, statistically-significant direction — exact proportions: Srikor et al., Fig. 3b & Fig. 6b.
See it for yourself
Drag the patient
from well to bleeding.
Watch the same 2,397 cells answer the call — and change who they are.
grip strength
mostly calm
no symptomsfeverbled
← drag the slider →
A model of the reported direction — CX3CR1+ killer cells rise with severity (p=0.03), calm intermediate cells fall (p=0.009). Exact proportions: Srikor et al., Fig. 3b.
The twist
And here is the strangest part.
In the patients who bled, the killer cells weren't even aimed at the strain making them sick. They were tuned to the previous serotype — the one from an infection years before.
The body fought
the last war.
In DHF, CX3CR1+ cells trended toward the previously infecting serotype's epitope; in asymptomatic cases, the calm cells trended toward the current one. Med
Srikor et al. 2026 · Fig.3e, Fig.5
Srikor et al. 2026 · Fig.3e, Fig.5
The mechanism
Recalled from memory, gripping too hard, and already pressed against the vessel wall —
the killer cell unloads. Its cytotoxic payload and the chemical alarms it releases may help pry open the very vessels whose leak defines severe dengue.
So is severe dengue the virus winning —
or your own memory, overreaching?
Clonally-expanded CX3CR1+ cells correlated with severity cytokines (IL-15, MIP-1α, IL-10…); proposed endothelial-injury model. Med · proposed
Srikor et al. 2026 · Discussion
Srikor et al. 2026 · Discussion
So where does this
leave us?
leave us?
For a century, we measured immunity by its strength.
This paper says, quietly — strength was the wrong question.
Not how hard you respond.
Which cell answers.
Which cell answers.
"Disease severity may not be solely dictated by T cell reactivity but by the qualitative features of specific subsets." — Srikor et al., Nat Commun 2026 · Discussion
Why it matters
We've been burned by this before. The first dengue vaccine, Dengvaxia, raised the risk of severe disease in children who'd never had dengue — it taught their bodies the wrong memory.
CYD-TDV increased severe dengue in seronegative recipients → WHO pre-vaccination screening. High
WHO · GACVS
WHO · GACVS
The dream
A vaccine that summons the calm cell — and leaves the over-eager veteran asleep.
The honest part
Hold the certainty, though. This was sixteen people — four, seven, five. A photograph of one moment, not a film.
The killer cell travels with severe dengue. Whether it causes the bleeding, or just rides along, is the next war — one the authors are careful to say they haven't won yet.
Small cohort; limited HLAs/epitopes; cross-sectional association, not proven causation (authors' stated limitations). High
Srikor et al. 2026 · Discussion
Srikor et al. 2026 · Discussion
And so
Somewhere in the blood of everyone who's had dengue once, the cells that remember it are still circulating.
Some of them remember
too well.
Waiting, quietly, for the next bite.