What proportion of patients who undergo tumor resection for TIL therapy ultimately do not receive treatment because TIL fail to expand from their tumor fragments?
Great question. The percentage of patients who never receive TIL because no cells expand from their tumor fragments varies widely by tumor type.
Historically, in early academic TIL programs, expansion success ranged anywhere from 30% to 90%, meaning 10–70% of patients could not proceed simply because their tumor fragments did not yield sufficient TIL. So manufacturing failure was a major barrier.
In melanoma, where TIL therapy is now clinically established, modern standardized manufacturing—like the lifileucel process—has pushed success rates above 90–96%, so only about 4–10% of melanoma patients with a resection fail due to lack of outgrowth.
But in head and neck squamous cell carcinoma, which is the disease model in my dissertation, the biology is very different. Several groups have reported much lower TIL outgrowth success from HNSCC specimens, sometimes only 30–40% of tumors yielding adequate TIL. This means that in HNSCC, a substantial fraction of patients may not reach infusion simply because the tumor fragments do not expand well ex vivo.
So the takeaway is:
In melanoma today, manufacturing failures are rare.
In HNSCC and other non-melanoma solid tumors, TIL outgrowth remains a significant barrier, largely reflecting a more suppressive tumor microenvironment and less predictable ex vivo expansion.
In what scenario would RT on 7 days before tumor harvest be better than 5 days before tumor harvest and how/why?
In our study we chose 5 days before harvest mainly based on practical constraints and on pilot data showing that RT at that time point increased the proportion of tumors from which we could expand TIL. We did not directly compare 5 versus 7 days. But you can definitely imagine scenarios where 7 days could be advantageous.
Mechanistically, after a single ablative RT dose you get a sequence of events: immunogenic cell death and type I IFN within the first 1–3 days, then induction of CXCR3 ligands like CXCL9 and CXCL10, dendritic cell activation and migration to the tumor-draining lymph node, and only then the influx of newly primed effector CD8 T cells back into the tumor. Several preclinical studies show that tumor-specific CD8 T cells and functional CD8⁺ TIL activity often peak around a week or slightly later after RT or chemoradiation. For example, in an HPV⁺ TC-1 chemoradiation model, E7-specific CD8⁺ TILs peaked on day 7 after treatment.
BMJ Journals
Other groups have reported CD8⁺ infiltration increasing from day 3 and peaking between days 7 and 14 after ablative RT, with maximal granzyme-B⁺ CD8⁺ T cells around day 7 in some fractionation schemes.
SpringerLink
+2
AACR Journals
+2
So in a setting where those kinetics hold, taking the tumor 7 days after RT might capture a time window where you have more tumor-resident, recently activated CD8⁺ T cells per fragment, and a more inflamed, chemokine-rich microenvironment than at day 5. In other words, if your limiting factor is the number and activation state of tumor-reactive T cells you can harvest—especially in a relatively ‘cold’ tumor at baseline—RT on day −7 could, in principle, yield a richer starting TIL product than RT on day −5.
The trade-off is that, as time goes on, suppressive populations and stromal changes also evolve, and you’re giving the tumor a bit more time to grow. There’s also no direct head-to-head dataset yet comparing −5 versus −7 days specifically for TIL manufacturing. So I would frame it as a hypothesis: I would expect RT at −7 to be preferable in a scenario where we know that peak effector T-cell infiltration and chemokine expression occur around day 7 for the chosen dose and fractionation, and where the clinical team is less constrained by the extra two days before surgery. Testing −5 versus −7 in a kinetics study—looking at TIL yields, phenotype, and function—would be a very natural next preclinical experiment.
Why did you focus on preREP TIL instead of postREP TIL?
Clinically, TIL manufacturing is usually described as a two-step process:
a pre-REP phase, where tumor fragments are cultured in high-dose IL-2 for 2–3 weeks and TIL grow out of the fragments, and
a rapid expansion protocol (REP), a 14-day bulk expansion with anti-CD3, irradiated feeder cells, and IL-2.
Our murine system was designed to model exactly that first bottleneck, the pre-REP outgrowth step: we irradiated tumors, harvested them five days later, and asked whether more fragments from more mice would successfully yield pure TIL cultures and whether those TIL were more functional. RT increased the proportion of mice from which we could expand TIL (96% vs 74%) and increased the fraction of expanding fragments, with improved TNFα production and better in-vivo activity after adoptive transfer.
Methodologically, adding a formal REP on top of this would have layered a very strong, non-physiologic stimulus on all TIL (anti-CD3 + feeder cells), which tends to partly homogenize phenotype and can drive further differentiation. Several clinical/technical papers point out that REP can push TIL toward more terminal effector states, and that many groups are now trying to shorten or modify pre-REP/REP specifically to preserve less-differentiated, stem-like or TRM-like cells.
For our mechanistic question—“does RT change how many and what kind of T cells you can harvest from the tumor in the first place?”—it is cleaner to read out effects before applying that massive in-vitro expansion step.
would RT preconditioning also matter at the post-REP level?
The REP product still reflects its starting pool.
Even though REP massively expands TIL, the final infusion product is still seeded by whichever clonotypes and phenotypes were present pre-REP. Studies of “TIL 3.0”, NeoExpand, and IL-2/IL-15/IL-21 cocktails all show that when you start with a pool that is enriched for tumor-reactive, less-differentiated, or PD-1⁺/CD39⁺ cells, the post-REP product is more functional and sometimes more stem-like than with conventional protocols.
BMJ Journals
+4
iovance.com
+4
BMJ Journals
+4
That implies that anything which changes the composition and activation state of tumor-resident T cells at harvest—such as RT preconditioning—should influence the biology of the eventual post-REP product as well.
RT selectively modulates intratumoral T cells rather than just ‘wiping them out’.
Work from Arina et al. and others shows that tumor-resident, antigen-experienced T cells can survive clinically relevant RT doses and actually become more motile and more IFNγ-producing, whereas naïve or peripheral T cells are more radiosensitive.
Nature
+2
ResearchGate
+2
In our own data, pre-RT tumors harvested five days after 8 Gy showed increased chemokine expression (e.g., CCL21, CXCL10) and higher chemokine-receptor expression on TIL, as well as increased cytokine production.
Manuscript - anonymized
So RT is not just changing tumor cells; it is re-shaping the TIL compartment that ultimately seeds expansion. It is reasonable to expect those advantages to propagate, at least in part, through a subsequent REP.
Clinically, the main manufacturing bottleneck is often pre-REP outgrowth, not REP itself.
Reviews of TIL manufacturing emphasize that many patients are lost because their fragments simply do not yield adequate TIL during the pre-REP phase; once you have a robust pre-REP culture, REP usually succeeds.
What are “feeder cells”?
In TIL therapy, feeder cells are irradiated accessory cells whose job is to provide:
Costimulation
Cytokines
Surface ligands
A safe activation platform during REP (rapid expansion protocol)
They do NOT proliferate because they are irradiated, but they provide the signals that allow TIL to undergo massive (1,000–5,000×) expansion.
In humans:
Clinical TIL manufacturing uses irradiated allogeneic PBMCs from 2–3 donors, typically at a 200:1 feeder:TIL ratio, stimulated with anti-CD3 (OKT3) + IL-2.
These human PBMC feeders provide CD80/CD86, 4-1BBL, OX40L, ICAM-1, cytokines, and other costimulatory ligands necessary for explosive T cell proliferation.
what are young TIL?
Young TIL are TIL products that skip the REP phase and undergo only limited expansion. They’re called ‘young’ because they are less differentiated and retain more stem-like, early-memory features. Compared to conventional REP-expanded TIL, young TIL show better persistence and often contain a higher proportion of neoantigen-reactive clones. The idea is to avoid pushing TIL into a terminally differentiated state. They are promising but not yet the clinical standard.
Q: Do feeder cells in REP need to be MHC/HLA matched?
A: No. In REP, TCR stimulation comes from anti-CD3, not peptide–MHC. Feeder cells are irradiated and only provide costimulation and cytokines, so matching is unnecessary.
Q: Why do target cells in co-culture need to be MHC/HLA matched?
A: Because TCR recognition is MHC restricted. T cells only recognize antigen on self MHC; without matching, they cannot see the target, so you lose killing and cytokine responses even if the T cells are functional.
how do they make sure its not turning into a wildfire?
Fire crews prevent a prescribed burn from turning into a wildfire by controlling fuel, boundaries, weather, and ignition patterns. Specifically:
Firebreaks are created first—bare soil, mowed strips, or plowed lines that stop the fire from spreading beyond the planned area.
Weather conditions are chosen carefully: low wind, stable humidity, and predictable wind direction. Burns are canceled if conditions shift.
Fuel loads are assessed and reduced ahead of time so the fire intensity stays within safe limits.
Ignition patterns (like backing fires or narrow lines of flame) are used to make the fire move slowly and predictably.
Trained crews and equipment—including water tanks, rakes, blowers, and sometimes engines—surround the burn unit and monitor containment continuously.
In short: they control when, where, and how the fire burns, and they don’t light it unless conditions are safe and crews are in place to contain it.
how do controlled fires help Florida Scrub Jay?
Scrub jays need low, open scrub. Without fire, oaks and pines grow too tall and dense, and jays disappear.
For the scrub jay, fire keeps the scrub short, patchy, and full of acorns and insects; need for nesting; spot predators easily
how do controlled fires increase insects for florida scrub jay and food for black bears?
Flash-card version (short, punchy):
Why fire boosts insects
• Resets vegetation.
Fire removes litter and triggers fresh, nutritious regrowth → more herb-eating insects.
• Spikes flowering.
More blooms and seeds after burns → more pollinators and seed feeders.
• Creates patches.
Burned + unburned mosaics support many different insect niches → higher diversity.
• Opens the ground.
Exposed, warmer soil helps ground-dwelling insects move, nest, and thrive.
• Supports fire specialists.
Some insects depend on burned wood, smoke cues, or fire-stimulated plants.
• Predators lag behind.
Short term: insects rebound faster than predators → brief population boom.
If you want, I can compress this even further to 2–3 lines.
How do controlled fires help florida black bear?
For Florida black bears, fire thins out the dense brush and then brings back a flush of new growth—berries, nuts, and tender shoots—that they feed on, especially a few years after a burn
How do controlled fires help gopher tortoise?
For gopher tortoises, fire keeps the canopy open so sunlight can hit the sand and grow the grasses and wildflowers they eat, and it maintains the open, sandy ground they need for burrows and nesting.
what are the main findings of our ABM-GRID paper?
RT alone isn’t enough.
Cytotoxic RT (2 Gy × 35) shrinks tumors but fails to cure because effector T cells stay suppressed and tumors regrow.
• Immunogenic RT changes the game.
When RT-killed cells recruit T cells, whole-tumor RT can clear tumors — but only in some TIME states. Early control = RT; final clearance = immune.
• GRID can outperform whole-tumor RT (when immunity helps).
With immunogenic death, GRID pushes even “cold” tumors toward immune-driven clearance — despite a lower average dose.
• Different dominant mechanisms.
Whole-tumor RT = mostly radiation log-kill (immune matters late).
GRID RT = mainly T-cell killing, especially at the end.
• Context matters.
Outcomes depend on the starting TIME. Same RT → different results. Personalize RT to the immune context, not one-size-fits-all.
how would you design an ABM of RT+ACT?
I’d build a spatial ABM that explicitly represents tumor cells, endogenous T cells, and adoptively transferred TIL on a 2D or 3D grid. Tumor cells would proliferate and die stochastically; T cells would move, recognize tumor cells, kill, become exhausted, or die, based on local rules.
Radiation would enter the model as:
a direct cytotoxic effect on tumor (and to some extent immune) cells, and
an immunogenic effect that increases antigen release, chemokines, and recruitment/activation of new T cells from the draining lymph node.
I’d first calibrate basic parameters (tumor growth, TIL killing, RT log-kill, simple immune recruitment) to existing in vivo data from our RT + TIL experiments. Then, once the model reproduces key readouts (tumor growth curves, rough TIL numbers, and timing of responses), I’d run simulation “experiments” where I systematically vary:
Total RT dose (e.g., 4, 8, 12 Gy)
Fractionation (single vs 4 Gy × 2 vs 2 Gy × 4)
Timing of RT relative to tumor harvest and/or TIL infusion
Different initial TIME states (T-cell–rich vs T-cell–poor, high vs low Tregs, etc.).
For each scenario, I’d track not just final tumor size but also which mechanism dominates at different times (RT log-kill vs TIL killing, changes in TIL phenotype, recruitment vs exhaustion). Comparing these runs would let us identify dose and timing “windows” where RT most strongly boosts TIL efficacy, and how that window shifts with different baseline TIME compositions. That’s how the ABM would give us both candidate personalized RT schedules and mechanistic hypotheses to test back in the mouse
why didn’t the dewan et al results translate to the clinics very well?
Different targets: CTLA-4 = priming in tdLNs; PD-1 = exhausted T cells in tumors
• Model vs clinic: naïve mice vs heavily pretreated patients
• RT scope: usually only one lesion → tiny “in-situ vaccine”
• Timing: mice tightly synchronized; clinics not
• tdLNs: sometimes irradiated, sometimes not → priming disrupted
Bottom line: 8 Gy × 3 was optimized for CTLA-4 priming, not PD-1, so the abscopal effect is smaller and inconsistent.
What role does aCTLA4 play in T cell tdLN priming?
Naive T cells require signal 1 (TCR–MHC) and signal 2 (CD28–CD80/86) to be primed.
CTLA-4 competes with CD28 for CD80/86 with much higher affinity.
When CTLA-4 is blocked:
CD28 costimulation is restored and amplified.
More naive T cells cross the activation threshold.
Tregs lose part of their suppressive dominance in lymph nodes (CTLA-4 is highly expressed on Tregs).
for what indication is RT+IO standard of care or guidline?
Unresectable stage III NSCLC
Standard management includes concurrent chemoradiation followed by consolidation durvalumab (PD-L1 blockade). This is sequential with RT, but it is widely recognized as standard-of-care based on PACIFIC and follow-up data.
Locally advanced (high-risk) cervical cancer (FIGO 2014 stage III–IVA)
Pembrolizumab (anti–PD-1) with chemoradiotherapy is FDA-approved and is an example of immunotherapy integrated with definitive RT (given with CRT).
why is it always CRT and not just RT
Short answer (what you can say in one sentence)
“Because chemotherapy improves local control and radiosensitizes tumors, RT alone is often insufficient for curative intent in locally advanced disease, so immunotherapy is added on top of CRT rather than replacing chemotherapy.”
Deeper explanation (why CRT is the backbone)
1. Historical efficacy: CRT works better than RT alone
For many locally advanced solid tumors (e.g. NSCLC, cervical, HNSCC):
RT alone → inferior local control and survival
Concurrent chemotherapy + RT → improved outcomes
This is well established from decades of randomized trials.
So guidelines start from:
“What already cures or controls disease best?”
And that answer is usually CRT, not RT alone.
- Chemotherapy acts as a radiosensitizer
Most CRT regimens use relatively low-dose chemo (e.g. cisplatin):
Enhances DNA damage
Impairs repair mechanisms
Improves tumor kill per Gy
This allows:
Lower RT doses than RT-alone regimens
Better control of bulky or hypoxic disease
Importantly, this is local synergy, not immune synergy.
- Immunotherapy was added after CRT for safety reasons
When ICIs first entered the clinic:
Combining them with RT alone had little historical precedent
Combining them with CRT preserved the known curative backbone
- RT alone is rarely definitive in advanced disease
RT alone is typically used for:
Early-stage disease
Palliation
Medically inoperable patients
Those populations:
Are heterogeneous
Are often excluded from large immunotherapy trials
Make immune endpoints harder to interpret
Thus, RT-alone + IO has remained mostly investigational.
- Chemo complicates immune effects — but clinicians accept that
From an immunology perspective:
Chemotherapy can be immunosuppressive
It can blunt lymphocyte responses
But clinically:
Local tumor control still dominates decision-making
Immunotherapy is layered on top rather than replacing proven components
This creates exactly the tension your work addresses:
RT can be immunostimulatory or immunosuppressive depending on context.
what are the most common side effects in TIL therapy (HNC study)
Regarding safety, almost all patients (98%) experienced treatment-emergent adverse events (TEAEs). The most frequent AEs reported were chills (60%), hypotension (53%), fever (47%), hypophosphatemia (42%), febrile neutropenia (42%), and hypocalcemia (33%).
Q: Why do HPVE7⁺ CD8 T cells increase in the tumor but not in the tumor-draining lymph node (tdLN)?
A:
The data show two separable effects:
Tumor: Increased HPVE7⁺ CD8 frequency and increased tetramer MFI (stronger signal per cell).
tdLN: HPVE7⁺ CD8 frequency and tetramer MFI remain flat or too variable to detect change.
This pattern can be explained by:
Kinetics (tdLN as a transient waypoint):
HPV-specific CD8 priming/expansion in the tdLN likely peaks early (days 2–4). By day 5, these cells may have already exited the node and accumulated in the tumor.
Trafficking-driven enrichment:
Preferential homing and retention in tumor (e.g., CXCR3–CXCL9/10 axis, adhesion signals, antigen encounter) can concentrate E7-specific CD8 in tumor without increasing their abundance in the tdLN, which is continuously exporting cells.
Local antigen encounter in tumor:
Once E7-specific CD8 enter the tumor, repeated antigen stimulation can promote local proliferation and/or longer dwell time, increasing HPVE7⁺ frequency in tumor even if the tdLN pool does not expand at that timepoint.
Selection for high-avidity clones:
Increased tetramer MFI in tumor suggests enrichment for higher-avidity/activated E7-specific CD8. Tumors can “filter” for these clones, while the tdLN contains a broader mixture that does not shift enough to change the average.
Measurement sensitivity:
HPVE7⁺ cells are rare in tdLN (sub-1%). With small n and variability, true changes may be hard to detect there, whereas tumors allow detectable enrichment of rare antigen-specific cells.
