Treg Depletion and/or Inhibition¶

← back to shieldbreaks

Last updated: 2026-04-27 Rows: 51 across 44 studies Outcomes: 19 succeeded / 9 partial / 12 failed / 11 not-assessed (ratio-shift)

📄 Download PDF report (152 KB — executive summary + ranked interventions + per-trial detail tables)

🔍 Download critique PDF (405 KB — per-paper appraisal + cross-paper synthesis)

📊 Download Pharmacodynamic results (HTML) (213 KB — sortable, filterable trial table; opens locally in any browser)

Research question¶

Show / hide

Which clinical interventions reduce regulatory T cells (Tregs) — in absolute number, frequency, functional dominance, or suppressive capacity — in humans, and how durable / context-dependent is that effect?

Scope summary¶

Show / hide

Scope summary for Treg Depletion and/or Inhibition. Last updated 2026-04-24 from 51 trial-rows across 44 studies (skeptic-critiqued: 44/44, 100%). Evidence-synthesis aid for research planning — not clinical guidance.

Target effect¶

Show / hide

Reduce the abundance, frequency, dominance, or suppressive function of regulatory T cells in humans — via classical absolute depletion, ratio shift in favor of effectors, or destabilization of the Treg phenotype (FoxP3 loss, TSDR demethylation reversal, functional-suppression-assay impairment). Tissue compartments of interest are tumor, tumor-draining lymph node, peripheral blood, bone marrow, and involved tissue (skin, ascites, colon). Durability and compartment concordance are tracked alongside magnitude.

Cross-cutting caveat (read first)¶

Show / hide

The magnitude of Treg depletion is frequently uncoupled from clinical benefit in this dataset, and several of the "positive" intervention classes carry structural confounds that inflate the apparent effect. Three examples motivate the synthesis: (1) the long-running denileukin diftitox (DD) literature is pervasively compromised by CD25-gating during CD25-targeting therapy (Dannull 2005 family; flagged by the skeptic as cd25-gating-confound), (2) the IL-2-variant "favorable-ratio" framing (bempegaldesleukin, nemvaleukin) did not translate into clinical benefit in PIVOT-IO (phase 3 failed), (3) standard anti-CTLA-4 reliably fails to deplete intratumoral Tregs in humans (Sharma 2019, Huang 2011, Penter 2023) despite a coherent preclinical mechanism — the reconciling variable is Fc engineering, not target engagement. Compartment dissociation is also pervasive (e.g., Ager 2026: tumor Tregs ↓, tdLN Tregs ↑). Rankings below are calibrated to this.

Intervention grouping¶

Show / hide

Anti-CCR4 (mogamulizumab) — Fujikawa 2023 (PMID 37729184), Jinushi 2025 (PMID 40180420), Roelens 2022 (PMID 35041763), Gordon 2025 (+rhIL-15; PMID 40546724).
Fc-enhanced anti-CTLA-4 — Ager 2026 BMS-986218+ADT (PMID 41759531), Chand 2024 botensilimab+balstilimab (PMID 39083809).
Low-dose / metronomic cyclophosphamide — Ghiringhelli 2007 (PMID 16960692), Audia 2007 (PMID 17956583).
Denileukin diftitox (DD / ONTAK / E7777) — Dannull 2005 (PMID 16308572), Attia 2005 (PMID 16224276), Atchison 2010 (PMID 20664355), Luke 2016 (PMID 27330808), Thibodeaux 2021 (PMID 33771857), Geskin 2018 (PMID 29204699), Liao 2024 (PMID 39362046), Gwin 2025 (PMID 40006664).
Class-I HDAC inhibitors — Brinkmann 2018 panobinostat/HIV (PMID 29468194), Pili 2017 entinostat+IL-2 (PMID 28939740), Terranova-Barberio 2020 vorinostat+tamoxifen+pembro (PMID 32681091), Roussos Torres 2021 entinostat+nivo (PMID 34135021), Govindaraj 2014 aza+panobinostat (PMID 24297862).
Standard anti-CTLA-4 — Huang 2011 tremelimumab (PMID 21558401), Hamid 2011 ipilimumab (PMID 22123319), Sharma 2019 ipi/treme (PMID 30054281), Comin-Anduix 2008 treme (PMID 18452610), Ribas 2009 ipi/treme (PMID 19118070), Nancey 2012 ipi (PMID 22069060), Yi 2017 chemo+ipi (PMID 28951518), Penter 2023 decitabine+ipi (PMID 36706355).

Combination-regimen rows were rolled into their mechanism-of-interest bucket (entinostat+nivo → HDACi; decitabine+ipi → standard anti-CTLA-4; mogamulizumab+IL-15 → anti-CCR4). See Classes examined but not ranked below for counterexamples.

Top interventions¶

Show / hide

1. Anti-CCR4 (mogamulizumab) — most-replicated, mechanism-coherent Treg depletion across PBMC, tumor, and skin¶

Evidence base. 4 trials (paired pre/post, window-of-opportunity, and translational substudies; total n with Treg measurement = 97 across the group). Effect magnitudes: ~90% PBMC CCR4+ effector-Treg depletion in ATLL/CTCL (Fujikawa 2023, PMID 37729184); median 86.7% intratumoral eTreg reduction in 16/16 solid-tumor neoadjuvant cases (Jinushi 2025, PMID 40180420); significant PBMC and mixed-direction skin decrease at 4 weeks in Sézary syndrome (Roelens 2022, PMID 35041763); n=6 +rhIL-15 combo with Treg PD recorded but not tested (Gordon 2025, PMID 40546724). Skeptic confidence: 1 High, 2 Moderate, 1 Low.

Likelihood of desired effect. Direction, magnitude, and compartment concordance are all consistent — and the target biology (CCR4 is selectively enriched on effector Tregs) matches the readout. Jinushi 2025 anchors tumor-compartment depletion where most other classes fail. Mogamulizumab depletes the CCR4+ eTreg subset specifically; total intratumoral FOXP3+ counts are mixed (8/16 decreased in Jinushi), so claims should be phrased at the subset level.

Toxicity profile. Prescribing experience comes predominantly from the CTCL/ATLL label. Infusion reactions, skin eruptions (drug rash, severe cutaneous reactions including SJS/TEN have been reported), immune-related AEs, and severe/fatal complications after subsequent allogeneic HSCT are labelled risks (FDA USPI POTELIGEO; see DailyMed). In the solid-tumor combination setting (Jinushi 2025) the safety profile was reported as manageable with no new signals, but the population is small and follow-up is short.

Counter-productive mechanisms. CP severity aggregate: High (paper-internal + replicated). The class concern is depletion of beneficial effectors: CCR4 is co-expressed on central-memory CD8 T cells and Th1 effectors, so ADCC against CCR4+ Tregs collaterally depletes anti-tumor effectors (flagged 4/4 papers; paper-internal in Fujikawa 2023 and Jinushi 2025, external-evidence-only in Roelens 2022 and Gordon 2025; anchored to Tanaka 2021 Nat Commun and Kurose 2015 preclinical). Fujikawa 2023 authors explicitly argue this "dual depletion may cancel anti-tumor immune responses" and tie it to the observed minimal clinical benefit, with high-grade lymphopenia in 25%. Mitigations seen in the set: CTCL indications where CCR4+ malignant cells are themselves a target (Roelens 2022) and IL-15 rescue of memory-CD8/NK post-depletion (Gordon 2025, mechanism plausible but not directly measured).

Practical considerations. Mogamulizumab is FDA-approved for CTCL (2018); off-label use in solid tumors is window-of-opportunity / investigator-led so far. Combinable with anti-PD-1 (Jinushi 2025) and with rhIL-15 (Gordon 2025). Flow gating for CCR4 is straightforward; eTreg quantification (CD45RA−FoxP3^hi, Miyara Fr. II) is recommended given that total-FoxP3 readouts dilute the effect.

Why this rank. Most-replicated positive signal in the shieldbreak, with direct intratumoral evidence, a mechanism-coherent subset readout, and two Moderate-or-better skeptic confidences. No other class combines these three properties.

Per-trial detail.

Therapeutic agent	Efficacy	Toxicity	Reference
Mogamulizumab (KW-0761) monotherapy, CCR4-negative solid tumors	PBMC eTreg median 2.1% → 0.20% of CD4+ at 4 weeks (~90% reduction, dose-independent across 0.1–1.0 mg/kg, n=37 evaluable). 1 PR + 9 SD of 49 (modest monotherapy activity despite profound PD)	N=49; treatment-related AEs in 93.9%; grade 3–4 in 21 (42.9%). Dominant categories skin disorders (n=34) and lymphopenia (n=34, with 15 grade 3–4 lymphopenia events ≈30%). No grade 3–4 skin disorders; no drug-related deaths	Fujikawa 2023
Mogamulizumab + nivolumab (neoadjuvant, renal/lung/esophageal/oral)	Tumor CCR4+ FoxP3+ eTreg median −86.7% (range −94.8% to −52.7%), 16/16 patients depleted; total FoxP3+ median −11.1% (decreased in 8/16). 1 pCR + 3 PRs / 16; trend toward improved PFS/OS with lymphocyte infiltration	N=16; grade 3–4 TRAEs in 6/16 (38%): lymphopenia 25%, maculopapular rash 13% most frequent. 1 grade 5 interstitial pneumonia (cause of death adjudicated as disease progression). Cohort 2 (higher dose) skin AEs led to early closure without proceeding to cohort 3	Jinushi 2025
Mogamulizumab (Sézary syndrome, long-term PD)	Drastic decrease in activated PBMC Tregs within first 4 weeks (Unknown - non-OA for numeric values); long-term skin immune restoration qualitative; 17/26 with early complete blood response correlated with higher baseline CCR4	N=26; Unknown - non-OA for AE quantitative data. Mogamulizumab on-label cutaneous and infusion-reaction risks per FDA POTELIGEO USPI apply	Roelens 2022
Mogamulizumab + rhIL-15 (R/R T-cell malignancies)	Treg PD not measured in this trial — PD focus is NK expansion and ADCC. Treg row retained as mechanism-targeted regimen (uncritiqued for Treg-specific contribution); 1 PR (ATLL)	N=6; most common AEs rash, infection, fever (67% each); grade 4 AKI in 33%; grade 3+ anemia in 25% of cycles; 2 DLTs at dose level 2 (grade 4 acidosis/capillary leak/AKI; grade 4 myositis); MTD = dose level 1	Gordon 2025

2. Fc-enhanced anti-CTLA-4 — mechanism-rescue for the anti-CTLA-4 failure mode, with thin but directionally clean human evidence¶

Evidence base. 2 trials (n with Treg measurement = 36 total; randomized phase I and paired pre/post phase I). Ager 2026 BMS-986218 + ADT reduced intratumoral Tregs relative to ADT alone (p=0.031) in high-risk localized prostate cancer (PMID 41759531). Chand 2024 botensilimab + balstilimab reported significant intratumoral FOXP3+ reduction by IHC in MSS mCRC (PMID 39083809). Skeptic confidence: both Moderate; Ager RoB Some concerns, Chand RoB Moderate.

Likelihood of desired effect. Pharmacologically distinct from standard anti-CTLA-4: Fc engineering for increased FcγR affinity is the variable that reconciles the preclinical ADCC-on-Treg model with the null human result for ipilimumab/tremelimumab (see Rank 6). Ager 2026 is the most rigorous intratumoral-Treg PD readout in the shieldbreak — it pairs a randomized control (ADT alone) with a mechanism-specific comparator, and the contrast is significant. Chand 2024 is single-arm IHC, industry-sponsored; the direction is consistent but the evidentiary weight is lower. Compartment dissociation is on display — Ager also observed expansion of tdLN Tregs while tumor Tregs contracted.

Toxicity profile. Botensilimab has reported immune-related AEs including colitis, hepatitis, pneumonitis and infusion reactions at rates roughly comparable to or slightly exceeding ipilimumab at equivalent doses in the Chand 2024 combination (PMID 39083809, per the published safety section). Class-level FcγR-engagement predicts potential for heightened irAE burden relative to non-Fc-enhanced anti-CTLA-4, but head-to-head data are not available in this dataset. BMS-986218 safety in Ager 2026 was reported as manageable at the neoadjuvant dose tested.

Counter-productive mechanisms. CP severity aggregate: Moderate (2/2 papers). Two flagged patterns: (i) compartment dissociation / tdLN Treg expansion — Ager 2026 directly documents tumor-Treg contraction alongside tdLN Treg expansion (p<0.0001), a paper-internal finding in the priming compartment; (ii) ADCC on activated effectors — Fc-engineering increases FcγR engagement on any CTLA-4-high cell, which transiently includes activated CD8 effectors (external: Simpson 2013, Arce Vargas 2018). Ager 2026 partially bounds (ii) by showing CTLA-4 protein was largely confined to the Treg compartment in that tumor. Chand 2024 does not report compartment-specific data and adds anti-PD-1 combination autoimmune-toxicity compounding (paradoxical-autoimmunity tag).

Practical considerations. Both agents are investigational (botensilimab AGEN1181 and BMS-986218 are in active clinical development as of 2026). Industry-controlled supply; academic access primarily through sponsored trials. Combinable with anti-PD-1 and ADT as demonstrated.

Why this rank. The strongest mechanism-rescue story in the dataset and the most rigorous intratumoral randomized comparison (Ager 2026), but total n across the class is small and single-vendor; one more replication outside the sponsors would shift this to Rank 1 territory.

Per-trial detail.

Therapeutic agent	Efficacy	Toxicity	Reference
BMS-986218 (Fc-enhanced anti-CTLA-4 / afucosylated ipilimumab) + ADT, neoadjuvant high-risk localized prostate	Tumor TI-Treg frequency reduced in ADT+BMS-986218 vs ADT alone (p=0.031, n=14 vs n=10); ADT-only arm INCREASED TI-Tregs vs untreated (p=0.002); tdLN Tregs simultaneously expanded (p<0.0001). Densities figure-only, not text-quantified. Depth of Treg reduction associated with improved clinical outcomes; CD16a+ macrophage correlation p=0.033 supports ADCC	N=24 randomized; any-grade TRAEs 80% (ADT-only) vs 71% (ADT+NF), mostly attributed to ADT and injection-site reactions. Grade ≥3 TRAEs rare — 1 patient had grade 3 asymptomatic lipase elevation that resolved without intervention; 3 grade 1–2 GI events on anti-CTLA4-NF arm. No interim safety boundary crossings; no unexpected surgical complications	Ager 2026
Botensilimab (AGEN1181, Fc-enhanced anti-CTLA-4) + balstilimab (anti-PD-1), MSS mCRC	Significant intratumoral FOXP3+ reduction by IHC and RNA-seq signature in n=12 paired biopsies (fold-change and p not text-stated; figure-only, Fig 5H). PBMC/serum Tregs unchanged (tumor-selective). Activity in ICI-resistant tumor type	N≈65 phase 1 enrolled; most common grade ≥3 TRAE was diarrhea/colitis (combined preferred terms diarrhea/colitis/enteritis; per-AE frequencies in Supplementary Table S9 not in extracted text). Notably absent: hypophysitis (consistent with V11L/L30L Fc point mutations abrogating C1q binding)	Chand 2024

3. Low-dose / metronomic cyclophosphamide — schedule-dependent; the positive report is small but statistically strong¶

Evidence base. 2 trials (n with Treg measurement = 58). Ghiringhelli 2007 reported a 61% Treg frequency reduction and 78% absolute count reduction (p<0.0001 for both) in n=9 end-stage solid-cancer patients on metronomic oral cyclophosphamide, with restored NK/T-cell function (PMID 16960692). Audia 2007 reported no Treg reduction in n=49 with a single IV cyclophosphamide dose prior to intratumoral BCG (PMID 17956583). Skeptic confidence: Moderate (Ghiringhelli), Low (Audia).

Likelihood of desired effect. Dose-schedule-dependent. The metronomic oral schedule (Ghiringhelli) depleted; the single IV dose (Audia) did not. This is a real biological distinction, not a discrepancy to explain away — both critiques converge on schedule as the reconciling variable. The positive report is small (n=9) but the effect size and significance are unambiguous.

Toxicity profile. Metronomic cyclophosphamide has a well-characterized safety profile: myelosuppression, hemorrhagic cystitis (rare at metronomic doses), infections, and gonadotoxicity are the load-bearing labelled risks (cyclophosphamide USPI; FDA DailyMed). At the 50 mg/day oral metronomic dose used by Ghiringhelli, cumulative exposure is lower than standard chemotherapy but long-term bladder and fertility monitoring is standard.

Counter-productive mechanisms. CP severity aggregate: Low for the metronomic regimen (Ghiringhelli 2007), Moderate for single-IV dosing (Audia 2007). The mechanism-level concern is non-selective lymphopenia at cytotoxic doses; metronomic oral dosing exploits the lower ATP reserves of Tregs for preferential depletion (Lutsiak 2005 Blood; Ghiringhelli 2004) and the paper-internal evidence is that effector function is preserved/restored. This is arguably the cleanest dose-selective Treg-depletion mechanism in the shieldbreak — but it is not effector-free, and the single-IV Audia 2007 null is mechanism-consistent with non-selective depletion eliminating the BCG-responding effectors.

Practical considerations. Generic, inexpensive, orally available, globally accessible. Combinable with checkpoint blockade and vaccines; used widely as a Treg-depleting pretreatment in adoptive-cell-therapy and vaccine trials outside this dataset. Schedule must be metronomic — single-IV dosing is a likely null.

Why this rank. High replication floor for schedule biology and a strong statistical result in the positive report, but the total evidence base is 2 trials and the positive n is 9. Ranked above DD despite the smaller n because the CD25-gating confound makes much of the DD literature hard to interpret.

Per-trial detail.

Therapeutic agent	Efficacy	Toxicity	Reference
Metronomic oral cyclophosphamide 50 mg/day (advanced end-stage solid cancer)	PBMC CD4+CD25high frequency 7.9 ± 1.5% → 3.1 ± 1.8% (−61%, p<0.0001); absolute count 28.7 ± 9.4 → 6.4 ± 5.4 cells/mm³ (−78%, p<0.0001) at 1 month. T-cell proliferation and NK cytotoxicity restored to healthy-volunteer levels	N=9; AE frequencies not reported in this PD-focused paper. Cyclophosphamide class risks (myelosuppression, hemorrhagic cystitis, infections, gonadotoxicity) per generic USPI apply; metronomic 50 mg/day cumulative exposure substantially below standard chemotherapy dosing	Ghiringhelli 2007
Single IV cyclophosphamide + intratumoral BCG (metastatic mixed solid cancer)	PBMC Treg baseline 9.2% (vs 7.1% healthy); no significant modulation of Treg numbers or function post-cyclophosphamide. Authors explicitly state cyclophosphamide may not represent optimal Treg-elimination therapy	N=49; Unknown - non-OA for detailed AE frequencies. Standard single-dose cyclophosphamide toxicity expected	Audia 2007

4. Denileukin diftitox (DD / ONTAK / E7777) — large body of work, but the CD25-gating confound and the Attia-vs-Dannull split make this ambiguous¶

Evidence base. 8 trials (n with Treg measurement = 153 across the group; designs span paired pre/post, randomized phase 2, single-arm phase 1/2, and case series). Positive signals: Dannull 2005 (~51% PBMC reduction in RCC; PMID 16308572), Atchison 2010 (56.3% with DD+HD IL-2; PMID 20664355), Thibodeaux 2021 (n=2 DD+IFNα "worked" before drug shortage; PMID 33771857), Geskin 2018 (29% in CTCL, p=0.03 — responder- dependent; PMID 29204699), Liao 2024 (73% PBMC p=0.0275, 67% ascites p=0.27 n.s.; PMID 39362046), Gwin 2025 (p=0.10 overall n.s., partial; PMID 40006664). Negative signals: Attia 2005 (null by FoxP3 mRNA with retained ≥50% suppressive function, n=13 melanoma; PMID 16224276), Luke 2016 (phase 2 n=60 with peptide vaccine, no depletion; PMID 27330808). Skeptic confidence: 1 Moderate (Attia), 2 Moderate (Liao, Geskin), 5 Low; RoB distribution includes 2 Serious.

Likelihood of desired effect. Genuinely uncertain. The Attia 2005 vs Dannull 2005 split is the foundational conflict in this class, and the skeptic identified a pervasive structural confound: CD25-gating during CD25-targeting therapy means many of the "positive" reports are measuring surviving CD25^lo cells rather than genuine Treg depletion. The two reports that sidestep the confound with FoxP3-mRNA readouts go in opposite directions (Liao 2024 positive in PBMC; Attia 2005 negative in melanoma). The response-stratified finding in Geskin 2018 (responders deplete, non-responders expand) is intriguing but post-hoc. Directional consistency within CD25-gated studies is not strong evidence when the gating is the confound.

Toxicity profile. DD's labelled risks include capillary-leak syndrome, visual/vascular events, infusion reactions, and hepatotoxicity (ONTAK/E7777 USPI, FDA DailyMed). The product has had multiple manufacturing/supply interruptions, including the shortage that halted Thibodeaux 2021 enrollment — a non-trivial practical toxicity for research planning.

Counter-productive mechanisms. CP severity aggregate: Moderate (7/8 Moderate, 1/8 Low). Dominant flag: CD25+ activated-effector collateral — DD depletes CD25-high cells indiscriminately, including transiently activated CD8 and CD4 effectors (external: FDA ONTAK label, Baur 2013). Paper-internal concerns sharpen this in two specific contexts: (a) vaccine-priming window (Dannull 2005, Luke 2016) — if DD exposure overlaps priming, it may ablate the very effectors the vaccine generates; (b) IFNα-CD25 interaction (Thibodeaux 2021) — IFNα upregulates CD25 on activated T cells, plausibly making the DD+IFNα combination worse for effector collateral than DD alone. Liao 2024's intraperitoneal route bounds systemic exposure and is the Low-severity case. In CTCL (Geskin 2018) the CD25+ malignant-cell confound operates in both directions and the Treg-depletion mechanism is not cleanly isolable.

Practical considerations. E7777 (now also referred to as I/ONTAK) received FDA approval for CTCL (2023); supply has historically been inconsistent. Academic access outside CTCL requires sponsored or investigator-led trials. Any future Treg-PD endpoint should use FoxP3 mRNA or TSDR methylation to sidestep the CD25 confound — surface-only CD25-based gating should be treated as uninformative here.

Why this rank. Largest evidence base in the shieldbreak after standard anti-CTLA-4, but the skeptic-weighted magnitude shrinks substantially once the CD25 confound is applied and the Attia-vs-Dannull split is taken seriously. Ranked here rather than lower because Liao 2024 and Geskin 2018 provide at least some confound-resistant positive signal.

Per-trial detail.

Therapeutic agent	Efficacy	Toxicity	Reference
Denileukin diftitox (DAB389IL-2, single 18 µg/kg dose) prior to RCC RNA-DC vaccine	PBMC CD4+CD25high Tregs reduced 26–76% relative (median −51%) at day 4 in 7/7 treated; FoxP3 mRNA −30 to −80%; in vitro suppressive activity abrogated; nadir d4, ~75% recovery by 2 months. Enhanced vaccine-induced T-cell responses	N=7 treated + 4 vaccine-only controls; AE quantification not in extractable text — abstract describes Treg depletion "without inducing toxicity on other cellular subsets". DD class risks (capillary leak, infusion reactions, hepatotoxicity, visual events) per ONTAK USPI apply	Dannull 2005
Denileukin diftitox 9 or 18 µg/kg/day × 5 (metastatic melanoma, two cohorts)	PBMC FoxP3 mRNA cycle-1 change: 9 µg/kg cohort −1.27±2.57 (p=0.656, n=4); 18 µg/kg cohort −2.01±0.618 (p=0.031, n=5); pooled p=0.167. After ≥4 cycles −3.30±3.21 (p=0.380). In vitro Treg suppression retained at ≥50% in 5/5 patients tested. 0/13 objective responses	N=13; detailed AE frequencies not extractable from PMC text. Generally tolerated at studied doses (single grade ≥3 hypersensitivity reaction was DLT-defining in protocol)	Attia 2005
Denileukin diftitox + high-dose IL-2 (metastatic RCC)	PBMC Tregs median −56.3% pre-DD to post-DD (p=0.013, pooled cohorts B+C n=15). 33% RR (not distinguishable from HD IL-2 monotherapy historical benchmark)	N=18; Unknown - non-OA for detailed AE frequencies. HD IL-2 toxicity (capillary leak, hypotension, organ dysfunction) is the dominant burden in this combination	Atchison 2010
Single-dose denileukin diftitox + gp100 peptide vaccine (advanced melanoma, RCT)	PBMC Tregs not significantly altered (no quantification, Fig 2 unquantified); 1/1 paired tumor biopsy showed INCREASED intratumoral FoxP3 post-DD. No improvement in vaccine-induced T-cell responses vs vaccine alone. 1 PR + 8 SD across 17 treated (4 DD: 5 vaccine-only)	N=17 treated; no drug-related grade 3–4 adverse events reported. DLTs defined as ≥grade 3 or grade 2+ autoimmunity / visual impairment per protocol	Luke 2016
Denileukin diftitox + IFNα (advanced ovarian)	DD monotherapy phase II "failed"; DD + IFNα2a phase II 2/2 patients responded before DD shortage halted enrollment (Unknown - non-OA for numerics). Qualitative claim of Treg depletion + IFNα-augmented anti-tumor immunity	N=2 (DD+IFNα completed); Unknown - non-OA for AE quantitative data. DD class risks apply	Thibodeaux 2021
Denileukin diftitox 18 µg/kg/day × 5 (CTCL: MF and Sézary)	PBMC CD4+FoxP3+ median relative change −29% (94% CI −83% to −20%) post one DD cycle, p=0.03; clinical responders (9/12 long-term) achieved 20–45% absolute Treg reductions; non-responders 2/3 EXPANDED Tregs	N=77; abstract-only AE detail in extracted text. DD-typical infusion reactions, capillary-leak risk per ONTAK USPI; in CTCL cohort the CD25+ malignant-cell substrate confounds attribution	Geskin 2018
Intraperitoneal denileukin diftitox (ONTAK), recurrent ovarian	PBMC FoxP3 mRNA pooled −73% (mean 0.1726±0.0442 → 0.0374±0.0101, p=0.0275; 15 µg/kg subset p=0.0374); ascites mean −67% (0.1855±0.0945 → 0.0597±0.0304) but p=0.2737 n.s. (n=3). 5/9 met ≥25% Treg reduction efficacy criterion; 3 had CA-125 decreases, no PRs	N=10 across 3 dose levels; majority of AEs transient grades 1–2; 1 DLT in 6-patient 15 µg/kg expansion. 1 grade 4 cytokine storm at 25 µg/kg requiring prolonged hospitalization closed that arm. MTD = 15 µg/kg	Liao 2024
Denileukin diftitox (E7777 / I/ONTAK) 18 µg/kg/day × 5 + pDC-targeted vaccine (stage IV breast)	Overall PBMC Treg change p=0.10 (n.s.); 6/15 (40%) achieved ≥25% reduction (responder subset 56.0% ± 10.96%); anti-DT IgG response in 100% by week 6 likely limited efficacy. 0 CR/PR; 4 SD (27%)	N=15; 11 (73%) had at least one grade 3–4 AE; 2 (13%) discontinued for toxicity, 9 (60%) for progressive disease; per-AE category frequencies not extracted	Gwin 2025

5. Class-I HDAC inhibitors (entinostat / vorinostat / panobinostat) — context-dependent direction; oncology signals plausibly favorable, HIV/cART signal opposite¶

Evidence base. 5 trials (n with Treg measurement = 76 across the group). Oncology context — directionally favorable: Pili 2017 entinostat+HD IL-2 RCC (lower Treg associated with response, p=0.03; PMID 28939740), Terranova-Barberio 2020 vorinostat+tamoxifen+pembro in ER+ breast (tumor Tregs 11.8%→2.9% overall, p=0.0067; PMID 32681091), Roussos Torres 2021 entinostat+nivo (CD8:FoxP3 ratio 4.11→9.03, p=0.002; PMID 34135021), Govindaraj 2014 aza+panobinostat in AML (TNFR2+ Treg subset reduction in PBMC and BM with associated benefit; PMID 24297862). HIV/cART context — opposite: Brinkmann 2018 panobinostat HIV cART cohort (+40% Tregs at day 4, p=0.003; PMID 29468194). Skeptic confidence: all 5 Moderate; 2 Some concerns RoB, 3 Moderate.

Likelihood of desired effect. Plausibly favorable in oncology, clearly unfavorable in HIV/cART reactivation — a context-dependent class. Terranova-Barberio 2020 is the strongest single intratumoral result in this group (large magnitude, p<0.01, in tumor). Roussos Torres 2021 is properly a ratio-shift rather than Treg-only reduction (the +119.7% pct_change field reflects the CD8:FoxP3 ratio shift, not a Treg expansion — a data-quality note for downstream readers). Govindaraj 2014 targets a minority subset (TNFR2+ Tregs), not total Tregs, and panobinostat is the in-vitro driver.

Toxicity profile. Class-I HDACi labelled toxicities include thrombocytopenia, neutropenia, fatigue, GI (nausea/diarrhea), QT-interval prolongation (panobinostat more than entinostat/vorinostat), and a black-box diarrhea/cardiac warning for panobinostat in myeloma (FARYDAK USPI, FDA DailyMed). Vorinostat (ZOLINZA USPI) and entinostat (investigational) carry broadly similar hematologic toxicity at oncology doses. IL-2 combination (Pili 2017) and pembrolizumab combination (Terranova-Barberio 2020) add the respective partner toxicity profiles.

Counter-productive mechanisms. CP severity aggregate: Moderate (spread: 2 Low, 2 Moderate, 1 High context-outlier). Flagged patterns: (i) effector-function suppression — pan-HDAC inhibitors (vorinostat, panobinostat) can reduce effector CD8 function (external: Kroesen 2014); bounded for class-I-selective entinostat (Pili 2017, Roussos Torres 2021, both Low). (ii) Opposite-direction mechanism in the wrong context — Brinkmann 2018 panobinostat in HIV/cART increases Tregs 40%, a High-severity outlier that is context-discordant (HIV-latency reactivation, not oncology) and footnoted accordingly. (iii) DNMTi confounder in Govindaraj 2014 — the azacitidine partner demethylates the FOXP3 TSDR and can induce Tregs, operating against the HDACi direction. The selectivity/breadth axis (class-I vs pan-HDAC) is the reconciling variable the data point at.

Practical considerations. Vorinostat and panobinostat are FDA-approved (CTCL and multiple myeloma respectively); entinostat is investigational as of 2026. Orally available, broadly combinable. The HIV-reservoir use case motivated Brinkmann 2018 and is not a Treg-depletion indication — inclusion of that data point here is as a counterexample that pins down context-dependence.

Why this rank. Directionally favorable signal in oncology with multiple Moderate-confidence results and one strong intratumoral magnitude, but context-dependence and the single-positive-per-agent pattern prevent a higher ranking. Mechanism is plausible but not precisely specified (FoxP3 stability, TSDR methylation, TNFR2+ subset biology — each HDACi study invokes a different rationale).

Per-trial detail.

Therapeutic agent	Efficacy	Toxicity	Reference
Panobinostat (pan-HDAC) in HIV+cART (reservoir reactivation, context-discordant)	PBMC Treg frequency +40% at day 4 (p=0.003), sustained at day 28 (p=0.004); CTLA-4 MFI on Tregs +25% (p=0.007); CD39 MFI +12% (p=0.02). Tregs ACTIVATED rather than depleted; returned to baseline by week 24	N≈14–15 evaluable; HIV/cART context — paper notes oncology-dose HDACi toxicities have impeded development as immunomodulators in non-cancer; quantitative AE table not extracted	Brinkmann 2018
Entinostat + high-dose IL-2 (metastatic ccRCC)	PBMC Tregs lower in responders (n=5) than progressors (n=7) at C1D1 (p=0.0273); lower Treg overall associated with response (p=0.03). Tumor (n=3 paired): entinostat priming prevented HD IL-2-induced Treg expansion (not statistically tested). ORR 37%; mPFS 13.8 mo; mOS 65.3 mo	N=47; most common grade 3–4 TRAEs hypophosphatemia 16%, decreased lymphocytes 15%, hypocalcemia 7% — all transient. 1 RA flare; 1 death during treatment deemed unrelated (cardiac tamponade from undiagnosed lung primary). HD IL-2 hypotension 3.2%, capillary-leak signal expected	Pili 2017
Vorinostat + tamoxifen + pembrolizumab (ER+ metastatic breast, ≥5 prior regimens)	Tumor CD4+FoxP3+CTLA-4+ Tregs 11.8% → 2.9% overall (−75.4%, p=0.0067); responders 20.3% → 4.2% (p=0.031); non-responders 10.4% → 2.5% (p=0.034). PBMC Tregs unchanged. Treg depletion in BOTH responders and non-responders. ORR 4%; CBR 19%. Trial stopped early for futility	N=34; grade 3–4 toxicities included transaminitis 9% (incl immune hepatitis with discontinuation), fatigue 6%, hyponatremia 6%, thrombocytopenia 6%, anorexia 3%; 1 disabling stroke (relatedness unclear) requiring discontinuation. Grade 2 irAEs included pneumonitis, hypothyroidism, colitis, fatigue	Terranova-Barberio 2020
Entinostat + nivolumab ± ipilimumab (advanced HER2-negative breast / solid tumors)	Tumor CD8/FoxP3 ratio median 4.11 → 9.03 (T0 → T2; +119.7% RATIO, p=0.002, Wilcoxon, n=14 paired); ratio shift required ICI addition (T1 post-entinostat-only median 3.56). Not a Treg-absolute reduction. ORR 16% (incl. CR in TNBC); 4 responders trended toward greater shift (n.s.)	N=33 across 4 dose levels. Treatment-related AEs: fatigue 65%, nausea 41%, anemia 38%, diarrhea 26%, anorexia 26%. Grade 3–4: fatigue 21%, anemia 27%, neutropenia 12%. RP2D entinostat 3 mg weekly + nivolumab 3 mg/kg q2w + ipilimumab 1 mg/kg q6w (max 4)	Roussos Torres 2021
Azacitidine + panobinostat (AML, BM + PBMC)	TNFR2+ Treg subset (minority of total Tregs) decreased in PBMC and BM after 28-day cycles; associated with increased BM effector IFN-γ and IL-2 and clinical benefit in subset. Unknown - non-OA for numerics. Total FoxP3+ Tregs not the readout	N=14; Unknown - non-OA for AE detail. Class panobinostat AEs (cytopenias, GI, QT prolongation per FARYDAK USPI) and azacitidine toxicities expected	Govindaraj 2014

6. Standard anti-CTLA-4 (ipilimumab / tremelimumab) — foundational negative result for the Treg-depletion endpoint; motivates Rank 2¶

Evidence base. 8 trials (n with Treg measurement = 233 across the group; includes foundational papers Huang 2011, Sharma 2019, and Comin-Anduix 2008). Ipilimumab and tremelimumab do not deplete intratumoral Tregs in humans: Huang 2011 tumor FOXP3+ density 35→167 cells/mm² (p=0.0029, an increase; PMID 21558401), Sharma 2019 across melanoma/prostate/bladder paired biopsies — FoxP3+ increase in all cohorts (PMID 30054281), Comin-Anduix 2008 PBMC n.s. (PMID 18452610), Ribas 2009 n.s. tumor (PMID 19118070), Yi 2017 PBMC increase with chemo+ipi, p=0.012 (PMID 28951518), Penter 2023 marrow Treg expansion on decitabine+ipi in AML/MDS (PMID 36706355). Hamid 2011 measured baseline-only biomarker (PMID 22123319); Nancey 2012 is an n=1 enterocolitis case report overclaiming depletion (PMID 22069060). Skeptic confidence: 1 High (Sharma), 3 Moderate, 3 Low, 1 Very low.

Likelihood of desired effect. Low — converging evidence across the most methodologically rigorous reports in the shieldbreak (Sharma 2019 High confidence; Huang 2011 Moderate) indicates that non-Fc-enhanced anti-CTLA-4 does not deplete intratumoral Tregs in humans and in several contexts actively expands them. This is the foundational negative result that motivates the Fc-enhancement program (Rank 2).

Toxicity profile. Ipilimumab's irAE profile (colitis, hepatitis, hypophysitis, pneumonitis, dermatitis, endocrinopathies) is well-characterized (YERVOY USPI, FDA DailyMed). Tremelimumab's profile is similar (IMJUDO USPI). Standard-dose irAEs are clinically significant and rate-limiting.

Counter-productive mechanisms. CP severity aggregate: Low (6/8 Low, 1 Moderate case-report, 1 High is the decitabine-co-drug confound in Penter 2023). The aggregate understates the real story because the proximal Treg-depletion mechanism fails in humans — so CP severity is technically Low (you can't have collateral from a mechanism that isn't engaging). The flagged class concerns are: (i) alt-checkpoint upregulation — CTLA-4 blockade induces PD-1 and LAG-3 compensatory expression (external: Huang 2017 PNAS; Woo 2012); (ii) paradoxical autoimmunity — irAEs (colitis, hypophysitis) consume immune capacity in non-tumor tissue (Nancey 2012 case report). Penter 2023's High severity is a co-drug effect: decitabine expands marrow Tregs via TSDR demethylation, a finding the authors document directly as an ipilimumab resistance mechanism — a warning for any DNMTi + anti-CTLA-4 combination rather than a standalone ipi/treme concern.

Practical considerations. Both agents are FDA-approved in multiple indications (ipi 2011, tremelimumab 2022). Widely available, combinable with anti-PD-1 and platinum chemotherapy. For Treg-depletion PD endpoints specifically, standard anti-CTLA-4 should not be the tool of choice — use Fc-enhanced variants (Rank 2) or move to a different class.

Why this rank. Foundational, but negative. Included in the ranked list because the negative result is load-bearing for the rest of the synthesis — it is the reason Rank 2 exists as a distinct class and the reason this shieldbreak treats Fc engineering as the reconciling variable.

Per-trial detail.

Therapeutic agent	Efficacy	Toxicity	Reference
Tremelimumab 15 mg/kg q3mo (metastatic melanoma, paired biopsies)	Tumor FOXP3+ density 35.20 ± 30.06 → 167.35 ± 162.37 cells/mm² post-dose (+375%, p=0.0029 paired, n=19). FOXP3 INCREASED, not depleted; CD8+ TIL increase independent of clinical response. Caveat: paper applied unpaired Mann-Whitney label to paired biopsy data	N=32 enrolled, 19 paired-biopsy. Tremelimumab + ipilimumab class irAEs (grade ≥3) ~20% in pivotal phase 2; per-AE breakdown not in extracted text for this translational paper. DLT-defined events: grade 4 TRAE, grade ≥3 hypersensitivity, grade ≥2 colitis, autoimmune events in critical organs	Huang 2011
Ipilimumab (advanced melanoma, predictive-biomarker study)	Baseline (NOT longitudinal) FOXP3 quantification only: detected in 75.0% of clinical-benefit pretreatment biopsies vs 36.0% of non-benefit (p=0.014). Not a depletion endpoint — incidental Treg readout	N=82 evaluable across two ipilimumab doses (10 mg/kg vs 3 mg/kg). Drug-related any-grade 82.5% vs 76.2%; grade 3–4 15.0% vs 31.0%. irAE any-grade 55.0% vs 66.7%; grade 3–4 7.5% vs 19.0%. GI most common irAE category. 5 (12.5%) vs 11 (26.2%) discontinued for AE	Hamid 2011
Ipilimumab or tremelimumab (foundational paired tumor analysis: melanoma, prostate, bladder)	Tumor FOXP3+ INCREASED in all cohorts: ipilimumab melanoma (n=16 post vs 19 untreated, p<0.05), tremelimumab paired melanoma (n=18, p<0.05), no reduction in bladder (n=9) or prostate (n=16). CyTOF orthogonal validation in n=5 paired melanoma. Densities figure-only	N=45 across cohorts; this is a translational tumor-IHC paper without primary AE reporting. Standard ipilimumab/tremelimumab irAE profiles per YERVOY and IMJUDO USPIs apply	Sharma 2019
Tremelimumab (melanoma, PBMC PD with detailed flow)	PBMC FoxP3 mRNA showed no statistically significant change pre vs post-dose (n=8 evaluable of 13); Treg functional assay not performed (insufficient cells from 40 mL draws)	N=29 reported on this trial cohort; documented G2/G3 toxicities included diarrhea G2, hepatitis G3, acne rosacea G2, colitis G3, panhypopituitarism (2 cases, hypophysitis), uveitis, leukocytoclastic vasculitis. Per-patient case-list rather than aggregated rates	Comin-Anduix 2008
Ipilimumab or tremelimumab (retrospective biopsy analysis, melanoma)	n=15 biopsies from 7 patients; no consistent FoxP3 reduction; slight FoxP3+ increase in 2/3 paired responders; IDO associated with non-response. Unknown - non-OA for numerics	N=7; Unknown - non-OA for AE detail. Standard anti-CTLA-4 irAEs apply	Ribas 2009
Ipilimumab (case series of patients with severe enterocolitis)	Lamina-propria FoxP3+ Treg "profound long-lasting depletion" reported qualitatively in n=4; Unknown - non-OA for numeric density values. Title overclaims for case-report-level evidence	N=4 case series — all by definition had severe (grade ≥3) ipilimumab-induced enterocolitis, which is the inclusion criterion. Not informative for class-level toxicity rates	Nancey 2012
Neoadjuvant chemotherapy + ipilimumab (early-stage NSCLC)	PBMC: median Treg frequency INCREASED slightly by +1.05% (V1 chemo-alone → V3 post-ipi), p=0.012 — significant but tiny magnitude and opposite direction from depletion. Tumor: no pre-treatment biopsy, change not assessable	N=24; treatment-related grade 1–2 AEs in 54%, grade 3–4 in 46%, no treatment-related deaths. Most AEs attributable to carboplatin/paclitaxel. Ipilimumab-attributed irAEs: grade 2 pneumonitis 4%, grade 3 adrenal insufficiency 17%, diarrhea/colitis grade 1–2 25% + grade 3 13%	Yi 2017
Decitabine + ipilimumab (R/R AML/MDS, BM scRNA-seq + mIF)	Bone-marrow CD3+FOXP3+ density INCREASED post-ipilimumab (qualitative; baseline not numerically reported, no p-value cited); authors interpret BM Treg expansion as ipilimumab-resistance mechanism and explicitly suggest combining with Treg-depleting strategies	N=18; AE quantitative data not in extracted PMC text. Decitabine + ipilimumab combination toxicity (cytopenias, irAEs) per parent ETCTN/CTEP 10026 trial	Penter 2023

Classes examined but not ranked (counterexamples and thin-evidence classes)¶

Show / hide

Seven additional classes were surveyed and deliberately omitted from the ranked list:

Non-α IL-2 variants (nemvaleukin, bempegaldesleukin) — 6 trials (Bentebibel 2019 PMID 30988166; Diab 2020 PMID 32439653; Vaishampayan 2024 PMID 39567211; Gogas 2024 PMID 39025948; Calvo 2025 PMID 40759440; Piha-Paul 2025 PMID 40990800). These expand Tregs absolutely, with a "favorable-ratio" framing that did not hold up: Gogas 2024 PIVOT-02 showed ~8–10× absolute Treg expansion with bempegaldesleukin (+800%) and the PIVOT-IO phase 3 failed. Mechanism does not fit the shieldbreak's target effect.
Anti-CD25 daclizumab — 2 trials (Mahnke 2007 PMID 17315189; Morse 2008 PMID 18519811). Reported reductions are confounded by CD25-based gating during CD25-targeting therapy (same structural issue as DD). Not enough confound-resistant data to rank.
Iberdomide / CELMoDs — 2 studies (Lipsky 2022 PMID 35477518; Amatangelo 2024 PMID 38776914). Iberdomide/mezigdomide degrade Ikaros/Aiolos rather than Helios/IKZF2, and therefore expand Tregs (e.g., +104.9% at 0.45 mg in Lipsky 2022, p<0.001). Wrong direction.
DNMT inhibitors / HMAs as monotherapy — Han 2021 low-dose decitabine in ITP (PMID 33876188) and Penter 2023 decitabine+ipi (PMID 36706355, rolled into anti-CTLA-4 above) both show DNMTi increases Treg quantity/function. Counterexample direction in isolation; Govindaraj 2014 aza+panobinostat is the exception only because the HDACi partner drives the direction.
Anti-GITR — 2 trials (Piha-Paul 2021 GWN323 PMID 34389618; Davar 2022 TRX518 PMID 35499569). Both Low confidence, programs deprioritized by sponsors; insufficient evidence to rank.
Anti-TIGIT — 1 trial with Treg readout (Guan 2024 tiragolumab PMID 38418879). Single translational substudy, PBMC-only, Low confidence; not enough to rank.
PI3Kδ inhibitors — 1 trial (Gadi 2022 idelalisib in CLL, n=9 PBMC; PMID 35170759). Very-small-n, Low confidence.

Ranked prioritization¶

Show / hide

Rank	Intervention	Likelihood of effect	Toxicity burden	Counter-productive MoA	Overall
1	Anti-CCR4 (mogamulizumab)	High in CCR4+ eTreg subset (Fujikawa, Jinushi, Roelens concordant)	Moderate (cutaneous, infusion, post-allo-HSCT signal)	High (CCR4+ CD8 effector-memory collateral; 2/4 paper-internal, 4/4 replicated)	Strongest, most-replicated depletion with direct tumor evidence — but pair with effector rescue if the goal is anti-tumor benefit
2	Fc-enhanced anti-CTLA-4	Moderate (Ager randomized control; Chand consistent)	Moderate-to-high (class-level irAE + FcγR-engagement signal)	Moderate (tdLN Treg expansion, ADCC on activated CTLA-4-high effectors; 2/2)	Mechanism-rescue; thin but directionally clean; CP profile narrows but does not close the Rank-1 gap
3	Low-dose metronomic cyclophosphamide	Moderate (schedule-dependent; p<0.0001 in Ghiringhelli, null in Audia)	Low-to-moderate (generic, well-characterized)	Low (metronomic regimen is the cleanest dose-selective mechanism in the set; single-IV is Moderate)	Cheap, accessible, schedule-sensitive, low collateral — strong combination backbone
4	Denileukin diftitox	Low-to-moderate after CD25-gating discount	Moderate (capillary-leak, supply instability)	Moderate (CD25+ activated-effector collateral; vaccine-priming window and IFNα-CD25 interaction elevate specific combinations)	Large literature but structurally confounded; FoxP3-mRNA readouts needed; avoid priming-overlap scheduling
5	Class-I HDAC inhibitors	Moderate in oncology, unfavorable in HIV/cART (context-dependent)	Moderate (cytopenias; QT for panobinostat)	Moderate (class-I-selective entinostat bounded Low; pan-HDAC and DNMTi-partner cases Moderate-High)	Plausible; context- and selectivity-sensitive; entinostat is the cleanest CP profile in class
6	Standard anti-CTLA-4	Very low for the Treg-depletion endpoint specifically	Moderate-to-high (standard irAE profile)	Low (proximal mechanism fails so collateral is moot; Penter 2023 decitabine-co-drug is the High outlier and a combination-specific warning)	Foundational negative result; motivates Rank 2; not the tool for this endpoint

The Counter-productive MoA column summarizes the skeptic-assessed severity of mechanism-level risks that the intervention may undermine the shieldbreak's target effect even when its proximal endpoint is met. It is distinct from Toxicity burden (which is about patient-level AEs). A severe counter-productive MoA pulls the Overall rating down even when Likelihood of effect is high. Severity aggregates per-group as the modal paper-level severity, bumped up one step when a paper-internal High is replicated across ≥2 papers or documents a wrong-direction mechanism in the intended context. Wrong-direction context-outliers (e.g., Brinkmann panobinostat in HIV; Penter decitabine co-drug) are footnoted rather than allowed to move the aggregate.

Caveats¶

Show / hide

Total-n is small for several ranked groups. Fc-enhanced anti-CTLA-4 has n=36 across 2 trials; cyclophosphamide positive report is n=9; mogamulizumab solid-tumor tumor data is n=16 (Jinushi 2025). Rankings reflect magnitude × confidence, not magnitude alone.
CD25-gating confound is pervasive across the DD and anti-CD25 literature. Any future Treg-PD endpoint in a CD25-targeting program should use FoxP3 mRNA, TSDR methylation, or Helios/IKZF2 readouts.
Compartment dissociation is pervasive. Tumor-Treg depletion often coexists with tdLN-Treg expansion (Ager 2026) or with skin expansion (Roelens 2022 mixed). PBMC-only data should not be extrapolated to tumor.
Industry sponsorship is flagged by the skeptic for Chand 2024 (botensilimab) and several IL-2-variant trials; this is a caveat, not a disqualifier.
One extraction-fidelity discrepancy carried over from the screener pass: Huang 2011 applied a Mann-Whitney test label to paired pre/post biopsy data. The skeptic flagged it; the directional conclusion is unaffected.
Data-quality note for Roussos Torres 2021: the pct_change of +119.7% reflects a CD8:FoxP3 ratio shift, not a Treg-specific change — read the underlying biology, not the field value.
What would change the ranking.
An independent (non-sponsor) replication of Fc-enhanced anti-CTLA-4 intratumoral depletion would plausibly move Rank 2 above Rank 1.
A second confound-resistant (FoxP3 mRNA / TSDR) DD study with concordant positive direction would move Rank 4 up.
A randomized HDACi ± ICI study with a pre-specified intratumoral Treg endpoint would sharpen Rank 5 considerably.
A head-to-head of metronomic oral vs low-dose IV cyclophosphamide with matched Treg PD would resolve the schedule question at Rank 3.
Shared CP pattern — CD25+ effector collateral spans Rank 4 (denileukin diftitox) and the non-ranked anti-CD25 daclizumab class. Any CD25-targeting Treg-depletion program should assume the readout is structurally confounded and that activated effectors are being co-depleted in the priming window. Use FoxP3 mRNA / TSDR and schedule around (not over) effector priming.
Shared CP pattern — beneficial-effector collateral via surface-marker promiscuity is the dominant concern for the two highest-likelihood classes: anti-CCR4 (CCR4+ CD8 central-memory) and Fc-enhanced anti-CTLA-4 (CTLA-4-upregulating activated effectors). If the target effect is re-scoped from "proximal Treg depletion" toward "restored anti-tumor immunity," this shared pattern is load-bearing against both top ranks and favors Rank 3 (metronomic cyclophosphamide) as a combination backbone.
CP aggregation rule used here: modal per-group severity, bumped up one step when a paper-internal High is replicated across ≥2 papers or documents a clearly wrong-direction mechanism in the intended context. Wrong-direction outliers in a discordant context (Brinkmann panobinostat in HIV; Penter decitabine-co-drug) are footnoted rather than allowed to move the aggregate.
Rankings reflect Target-effect-as-written ("reduce Treg abundance/frequency/dominance/suppressive function"). If the Target effect is re-scoped toward anti-tumor benefit, CP severity should weigh more heavily and Rank 1 narrows substantially against Rank 2 and Rank 3.

Sources¶

Show / hide

Trial data and critiques: data/shieldbreaks/treg-depletion/trials.jsonl and data/shieldbreaks/treg-depletion/critiques.jsonl; per-trial citations below are also linked from the main table.
Ager CR et al. 2026 (BMS-986218+ADT, prostate) — PMID 41759531
Atchison E et al. 2010 (DD + HD IL-2, RCC) — PMID 20664355
Attia P et al. 2005 (DD, melanoma, null) — PMID 16224276
Audia S et al. 2007 (single-IV cyclophosphamide + BCG, null) — PMID 17956583
Amatangelo M et al. 2024 (iberdomide + dex, myeloma) — PMID 38776914
Bentebibel SE et al. 2019 (bempegaldesleukin, PD) — PMID 30988166
Brinkmann CR et al. 2018 (panobinostat, HIV/cART) — PMID 29468194
Calvo E et al. 2025 (nemvaleukin monotherapy) — PMID 40759440
Chand D et al. 2024 (botensilimab+balstilimab, MSS mCRC) — PMID 39083809
Comin-Anduix B et al. 2008 (tremelimumab) — PMID 18452610
Dannull J et al. 2005 (DD + DC vaccine, RCC) — PMID 16308572
Davar D et al. 2022 (TRX518 anti-GITR) — PMID 35499569
Diab A et al. 2020 (bempegaldesleukin + nivolumab) — PMID 32439653
Fujikawa K et al. 2023 (mogamulizumab, PBMC) — PMID 37729184
Gadi D et al. 2022 (idelalisib, CLL) — PMID 35170759
Geskin LJ et al. 2018 (DD, CTCL) — PMID 29204699
Ghiringhelli F et al. 2007 (metronomic cyclophosphamide) — PMID 16960692
Gogas H et al. 2024 (bempegaldesleukin + nivolumab, PIVOT-02) — PMID 39025948
Gordon MJ et al. 2025 (mogamulizumab + rhIL-15) — PMID 40546724
Govindaraj C et al. 2014 (azacitidine + panobinostat, AML) — PMID 24297862
Guan X et al. 2024 (tiragolumab+atezolizumab, anti-TIGIT) — PMID 38418879
Gwin WR et al. 2025 (DD / E7777 combo, breast) — PMID 40006664
Hamid O et al. 2011 (ipilimumab baseline biomarker) — PMID 22123319
Han P et al. 2021 (low-dose decitabine, ITP) — PMID 33876188
Huang RR et al. 2011 (tremelimumab, melanoma) — PMID 21558401
Jinushi K et al. 2025 (mogamulizumab+nivo, neoadjuvant) — PMID 40180420
Liao JB et al. 2024 (intraperitoneal DD, ovarian) — PMID 39362046
Lipsky PE et al. 2022 (iberdomide, SLE) — PMID 35477518
Luke JJ et al. 2016 (DD + gp100 vaccine, melanoma) — PMID 27330808
Mahnke K et al. 2007 (daclizumab) — PMID 17315189
Morse MA et al. 2008 (daclizumab + CEA vaccine) — PMID 18519811
Nancey S et al. 2012 (ipi + enterocolitis, case report) — PMID 22069060
Penter L et al. 2023 (decitabine+ipilimumab, AML/MDS) — PMID 36706355
Piha-Paul SA et al. 2021 (GWN323 anti-GITR ± spartalizumab) — PMID 34389618
Piha-Paul SA et al. 2025 (nemvaleukin, less-frequent IV) — PMID 40990800
Pili R et al. 2017 (entinostat + HD IL-2, RCC) — PMID 28939740
Ribas A et al. 2009 (ipi/treme retrospective, melanoma) — PMID 19118070
Roelens M et al. 2022 (mogamulizumab, Sézary) — PMID 35041763
Roussos Torres ET et al. 2021 (entinostat+nivo) — PMID 34135021
Sharma A et al. 2019 (ipi/treme tumor FOXP3, foundational) — PMID 30054281
Terranova-Barberio M et al. 2020 (vorinostat+tamoxifen+pembro) — PMID 32681091
Thibodeaux SR et al. 2021 (DD + IFNα, ovarian) — PMID 33771857
Vaishampayan UN et al. 2024 (nemvaleukin monotherapy) — PMID 39567211
Yi JS et al. 2017 (chemo+ipilimumab, NSCLC) — PMID 28951518
FDA prescribing information — POTELIGEO (mogamulizumab), YERVOY (ipilimumab), IMJUDO (tremelimumab), ONTAK / E7777 / I-ONTAK (denileukin diftitox), ZOLINZA (vorinostat), FARYDAK (panobinostat), cyclophosphamide generic USPI — all via DailyMed.

This summary is an evidence-synthesis aid for research planning. It does not constitute clinical advice and must not be used to guide patient care.

Scope inventory¶

Intervention classes: combo (14), denileukin-diftitox (9), anti-CTLA-4 (6), anti-CCR4 (4), IL-2-variant (4), anti-GITR (3), low-dose-cyclophosphamide (2), anti-CD25 (2), Fc-enhanced-anti-CTLA-4 (2), HDACi (1), PI3Kdelta-i (1), anti-TIGIT (1), other (1), DNMTi (1)
Tissue compartments: PBMC (33), tumor (12), bone-marrow (3), ascites (1), skin (1), other (1)
Design types: paired pre/post, treated-vs-untreated, single-arm, window-of-opportunity, case series (all flagged per row)
Row grain: one row per (study × tissue × timepoint-cluster × dose-cohort)
Primary-research-only; reviews / meta-analyses live in the side-list below.

See prompts/shieldbreaks/treg-depletion/search.md for the full search specification and prompts/shieldbreaks/treg-depletion/extract.md for the extraction schema.

Pharmacodynamic results¶

All PBMC tumor TDLN bone marrow ascites skin other

Show all columns (dose, treg defn, baseline/post values, durability, notes)

Intervention	Disease	N	Report	Tissue	Assay	Treg change	Result	Confidence	Bias & confounding	Source	PMID	Design	Dose/schedule	Treg defn	Readout	Timepoint	Timing detail	Baseline	Post	Magnitude	Significance	Durability	Intent	Data source	Notes
Denileukin diftitox (ONTAK)	Metastatic melanoma	13	Attia P ··· Rosenberg SA J Immunother, 2005	PBMC	flow	≈ n.s.	null result null	Moderate	Low small-nunderpowered	PMID PMCID DOI critique	16224276	paired-pre-post	Variable per prior dosing schedules	CD4+CD25+FOXP3+	frequency-of-CD4	early-on-treatment	post-dose serial sampling	Not reported in cells/µL; FoxP3 mRNA expression (copies/10^3 β-actin) only	FoxP3 mRNA change (cycle 1): 9 µg/kg cohort −1.27±2.57 (p=0.656, n=4); 18 µg/kg cohort −2.01±0.618 (p=0.031, n=5); pooled −1.68±1.11 (p=0.167, n=9); after ≥4 cycles (n=4) −3.30±3.21 (p=0.380)	Small, dose-dependent FoxP3 mRNA reduction; no elimination of CD4+CD25+ cells; suppressive function preserved (50.6-98.2% suppression retained in 5 patients post-cycle-2)	p=0.031 (18 µg/kg only); pooled n.s.	—	mechanism-targeted	pmc_full_text	Attia 2005 — key null result: DAB/IL-2 failed to eliminate functional Tregs in melanoma. Full-text backfill 2026-04-23: statistically significant FoxP3 mRNA decrease only in 18 µg/kg cohort (p=0.031), pooled analysis n.s. (p=0.167); in vitro suppression retained in 5/5 patients (50.6-98.2%). Conflicts with Dannull 2005 (16308572): Dannull saw d4 functional abrogation in RCC while Attia saw preserved suppression post-cycle-2 in melanoma; gating (FOXP3+ vs CD4+CD25+) and timepoint differences likely reconcile.
Denileukin diftitox (DAB389IL-2) prior to RCC tumor-RNA DC vaccineAutologous RCC tumor-RNA dendritic cell vaccine	Metastatic renal cell carcinoma	11	Dannull J ··· Vieweg J J Clin Invest, 2005	PBMC	flow+fn	↓ −51%	significant reduction depletion	Low	Moderate cd25-gating-confoundsmall-nno-multiplicity-correction	PMID PMCID DOI critique	16308572	paired-pre-post	Single infusion prior to vaccination	CD4+CD25+ (functional Treg assay)	frequency-of-CD4	early-on-treatment	pre- and post-denileukin diftitox; then post-vaccine	2.5-4.6% CD4+CD25high/FoxP3+ of CD4+ (2.1±1.2-fold above healthy volunteers n=10)	26-76% relative reduction of CD4+CD25high Tregs at day 4 post-DAB389IL-2 (all 7 subjects analyzed)	26-76% reduction in CD4+CD25high; FoxP3 mRNA reduced ~30-80% within purified CD4+ cells at day 4	reported significant (all 7 subjects)	Transient: ~75% of Tregs restored within 2 months (d4 nadir, d60 recovery)	mechanism-targeted	pmc_full_text	Dannull 2005 JCI — first clear demonstration that DAB389IL-2 depletes functional Tregs in cancer patients and enhances vaccine response. Full-text backfill 2026-04-23: explicit 26-76% range for frequency reduction, 30-80% for FoxP3 mRNA, nadir d4, 75% recovery by 2 months, suppressive function abrogated.
Low-dose oral cyclophosphamide (metronomic)	Advanced end-stage solid cancers	9	Ghiringhelli F ··· Chauffert B Cancer Immunol Immunother, 2007	PBMC	flow	↓ −61%	significant reduction depletion	Moderate	Moderate very-small-nsingle-arm-no-control	PMID PMCID DOI critique	16960692	paired-pre-post	50 mg PO twice daily, 1 week on / 1 week off	CD4+CD25highFOXP3+	frequency-of-CD4	early-on-treatment	after 1 month (day 28-ish)	7.9 ± 1.5% (CD4+CD25high of CD4+); 28.7 ± 9.4 cells/mm3 (absolute)	3.1 ± 1.8% (CD4+CD25high of CD4+); 6.4 ± 5.4 cells/mm3 (absolute) at 1 month	~61% reduction in frequency (7.9% to 3.1%) and ~78% reduction in absolute count (28.7 to 6.4 cells/mm3)	p<0.0001 (both frequency and absolute count)	Maintained during metronomic dosing; 2/2 patients with extended follow-up had Treg return to pretreatment levels ~2 months after stopping CTX	mechanism-targeted	pmc_full_text	Landmark Ghiringhelli 2007 metronomic-cyclo PD paper. Full-text backfill 2026-04-23: explicit mean±SD frequency and absolute count values, p<0.0001 for both, functional recovery data, 2-month off-treatment recovery. Side-list Ghiringhelli 2004 preclinical excluded from table.
Daclizumab (anti-CD25)	Metastatic melanoma	8	Mahnke K ··· Enk AH Int J Cancer, 2007	PBMC	flow	↓ sig.	significant reduction depletion	Low	Moderate cd25-gating-confoundabstract-onlyvery-small-nsurface-marker-gating	PMID DOI critique	17315189	paired-pre-post	Single bolus, then successive	CD4+CD25+ (FOXP3-validated in subset)	frequency-of-CD4	early-on-treatment	immediately post-first bolus through day 13	—	—	significantly reduced; successive doses further reduced	reported significant	Recovery observed over ~13 days after single dose; sustained with repeated dosing	mechanism-targeted	pubmed_abstract	Mahnke 2007 — daclizumab Treg depletion kinetics. Full-text backfill 2026-04-23: not openly accessible (Int J Cancer, inEPMC=N, subscription required). Abstract-only values retained: significant PBMC Treg reduction after single bolus, sustained with repeat dosing, ~13-day recovery window.
Single IV cyclophosphamide + intratumoral BCGIntratumoral BCG (non-specific immunotherapy)	Metastatic solid cancer (mixed)	49	Audia S ··· Bonnotte B Clin Exp Immunol, 2007	PBMC	flow	≈ n.s.	null result null	Low	Moderate abstract-onlysurface-marker-gating	PMID critique	17956583	treated-vs-untreated	Single IV dose (300 mg/m2 per paper)	CD4+CD25+ (FoxP3 referenced)	frequency-of-CD4	early-on-treatment	post-infusion serial	9.2%	not significantly changed	baseline 9.2% (patients) vs 7.1% (healthy) — post-cyclo: no significant modulation	n.s.	—	mechanism-targeted	pubmed_abstract	Audia 2007 — single-dose IV cyclo + BCG failed to modulate Treg numbers or function despite elevated baseline in cancer patients. Important null result anchoring the 'dose schedule matters' argument. Full-text backfill 2026-04-23: article not openly accessible (not in PMC, Europe PMC inEPMC=N, subscription required). Abstract-derived baseline 9.2% vs 7.1% healthy and 'no significant modulation' remains the only numeric available.
Tremelimumab	Melanoma	13	Comin-Anduix B ··· Ribas A J Transl Med, 2008	PBMC	flow	≈ n.s.	null result null	Moderate	Moderate small-nno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI critique	18452610	paired-pre-post	Per trial	CD4+CD25highFOXP3+	frequency-of-CD4	early-on-treatment	Serial post-infusion	FoxP3 mRNA pre-dose (qPCR); per-patient values not tabulated	No statistically significant change between pre- and post-dosing FoxP3 mRNA (n=8 evaluable of 13)	No significant change in FoxP3 mRNA pre→post; insufficient cells for functional suppression assay	n.s. (p-value not stated)	—	exploratory-but-prespecified	pmc_full_text	Comin-Anduix 2008 — PBMC PD of tremelimumab. Full-text backfill 2026-04-23: n=8 with baseline+follow-up FoxP3 mRNA data; 'failed to detect a statistically significant change (either positive or negative)'; Treg functional assay not performed due to insufficient cells from 40 mL draws.
Daclizumab (anti-CD25) prior to CEA-based cancer vaccineCEA-targeted cancer vaccine	Advanced CEA-expressing cancers (mixed solid)	15	Morse MA ··· Clay TM Blood, 2008	PBMC	flow	↓ sig.	significant reduction depletion	Moderate	Moderate cd25-gating-confoundsmall-nschedule-dependent	PMID PMCID DOI critique	18519811	paired-pre-post	Multiple doses	CD4+CD25highFoxP3+	frequency-of-CD4	early-on-treatment	pre- and post- multi-dose daclizumab; then post-vaccine	Baseline CD4+CD25highFoxP3+ frequency (absolute µL count) not tabulated; per-patient trajectories shown in figures	Cohort 2 (multi-dose, n=4): all 4 patients had lower Treg % at study end; Cohort 1 (single-dose, n=5): only 2/5 reduced	Dose-schedule dependent: multi-dose DD → 4/4 reduction; single-dose DD → 2/5 reduction; absolute Treg counts per µL began declining at week 5 and continued after final vaccination (Cohort 2 only)	Time-to-first immune response p=0.003 (Cohort 2 vs Cohort 0); individual Treg p-values not reported	Multi-dose sustained decline through post-vaccine follow-up; single-dose rebounded	mechanism-targeted	pmc_full_text	Morse 2008 Blood — pilot vaccine study. Full-text backfill 2026-04-23: schedule-dependent depletion confirmed (4/4 multi-dose vs 2/5 single-dose reduced Treg%); absolute µL counts only in Cohort 2. No tabulated baseline/post numeric means. Conflict pair with Luke 2016 (27330808): both show single-dose DD fails; Morse shows multi-dose DD succeeds — not a true conflict but a schedule-dependence finding.
Ipilimumab or tremelimumab	Melanoma	11	Ribas A ··· Cochran AJ Clin Cancer Res, 2009	tumor	flow	≈ n.s.	null result null	Low	Serious retrospective-analysisfoxp3-ihc-single-markerabstract-onlyvery-small-n+1	PMID DOI critique	19118070	paired-pre-post	Per trial	FoxP3+	frequency-of-lymphocytes	early-on-treatment	Post-dose biopsy	—	—	No consistent FoxP3 reduction post-CTLA-4 blockade	n.s.	—	exploratory-but-prespecified	pubmed_abstract	Ribas 2009 — intratumoral FoxP3/IDO in CTLA-4 blockade. Full-text backfill 2026-04-23: Clin Cancer Res article not in Europe PMC OA (inEPMC=N, subscription required). Abstract retains qualitative 'no consistent FoxP3 reduction' finding supporting early null-for-intratumoral-depletion; numeric values not extractable.
Denileukin diftitox (DD) + high-dose IL-2High-dose IL-2	Metastatic renal cell carcinoma	18	Atchison E ··· Kuzel TM J Immunother, 2010	PBMC	flow	↓ −56.3%	significant reduction depletion	Low	Serious cd25-gating-confoundabstract-onlysurface-marker-gatingretrospective-analysis+1	PMID DOI critique	20664355	paired-pre-post	Groups A/B/C dose-escalation	CD4+CD25+	frequency-of-CD4	early-on-treatment	pre-DD to post-DD, serial	—	—	median 56.3% reduction from pre-DD to post-DD	p=0.013	—	mechanism-targeted	pubmed_abstract	Atchison 2010 pilot. Clean median 56.3% reduction from pre-DD to post-DD, p=0.013 (abstract-reported). Full-text backfill 2026-04-23: article not openly accessible (J Immunother, inEPMC=N, subscription required); abstract numeric retained as sole source.
Tremelimumab (anti-CTLA-4)	Metastatic melanoma	18	Huang RR ··· Ribas A Clin Cancer Res, 2011	tumor	flow	↑ +375%	increase null	Moderate	Moderate foxp3-ihc-single-markersmall-nno-multiplicity-correctionauthor-coi-sponsor	PMID PMCID DOI critique	21558401	paired-pre-post	Per trial schedule	FOXP3+	frequency-of-lymphocytes	early-on-treatment	Post-dose biopsy	35.20 ± 30.06 FOXP3+ cells/mm^2 (n=19 paired)	167.35 ± 162.37 FOXP3+ cells/mm^2 post-dosing	Mean +132.15 cells/mm^2 (~4.75× INCREASE) in FOXP3+ density; CR patients 993.76±734.18 vs PD 775.63±458.70 (n.s. p=0.3340)	p=0.0029 (paired Mann-Whitney; note: paper uses paired test despite Mann-Whitney label)	—	mechanism-targeted	pmc_full_text	Huang 2011 — tremelimumab INCREASED rather than depleted intratumoral FOXP3+ cells. Full-text backfill 2026-04-23: baseline 35.2 → post 167.4 cells/mm^2 (p=0.0029, paired, n=19). Classic early null-for-intratumoral-Treg-depletion result supporting Sharma 2019's broader conclusion.
Ipilimumab (in patients with severe enterocolitis)	Melanoma (with ipi-induced enterocolitis)	4	Nancey S ··· Flourié B Inflamm Bowel Dis, 2012	other	flow	↓ qual.	significant reduction depletion	Very low	not amenable — case report case-report-n-of-fewabstract-onlyspin-abstract-vs-results	PMID DOI critique	22069060	case-series	Per trial	FoxP3+	frequency-of-lymphocytes	late	At time of colitis flare	—	—	Profound long-lasting FoxP3+ Treg depletion in colonic mucosa	reported	Long-lasting	incidental-but-measured	pubmed_abstract	Nancey 2012 — ipi-enterocolitis colonic mucosa Treg depletion. Full-text backfill 2026-04-23: Inflamm Bowel Dis article not in Europe PMC OA (inEPMC=N, subscription required). Abstract retains 'profound long-lasting FoxP3+ Treg depletion' in lamina propria; numeric density values not extractable from abstract.
Ipilimumab	Advanced melanoma (stage III/IV)	82	Hamid O ··· Berman D J Transl Med, 2011	tumor	flow	—	null result null	Low	Moderate exploratory-endpointno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI critique	22123319	paired-pre-post	Per trial	FoxP3+	frequency-of-lymphocytes	early-on-treatment	Baseline vs week 3 (24-72h post-dose 2)	Benefit group: FoxP3 detected in 75.0% of evaluable pretreatment biopsies; Non-benefit group: 36.0%	Longitudinal change not quantified — paper reports baseline FoxP3 only as predictive biomarker	Not a depletion study; baseline FoxP3 positivity associated with clinical benefit (OR ~10.38, p=0.014); TIL increase post-dose separate finding (p=0.005)	p=0.014 (baseline FoxP3 vs clinical benefit); not a post-treatment-change p-value	—	incidental-but-measured	pmc_full_text	Hamid 2011 — predictive biomarker study, NOT a depletion endpoint. Full-text backfill 2026-04-23 confirmed: FoxP3 quantified only at baseline (pretreatment biopsy), not longitudinally; 24-72h post-dose-2 samples collected but FoxP3 change not quantified. Reclassified intent_to_deplete from 'exploratory-but-prespecified' to 'incidental-but-measured' — the study did not aim to measure Treg depletion. Retained in table because it still reports a Treg-related readout in melanoma ipi patients.
Azacitidine + panobinostatazacitidine (DNMTi) + panobinostat (HDACi)	Acute myeloid leukemia	14	Govindaraj C ··· Plebanski M Clin Cancer Res, 2014	PBMC	flow	↓ sig.	significant reduction depletion	Moderate	Moderate abstract-onlysmall-n	PMID DOI critique	24297862	paired-pre-post	panobinostat + azacitidine 28-day cycles (dose not specified in abstract)	CD4+CD25hi FOXP3+ TNFR2+ subset	frequency-of-CD4	mid	after 28-day cycle(s) of aza+pano	elevated vs healthy donors (TNFR2+ Treg subset)	decreased in vivo after aza+pano	TNFR2+ Treg reduction (no numeric % given in abstract)	reported significant; associated with increased IFN-γ, IL-2 and clinical benefit	measured within 28-day cycles	pre-specified-endpoint	pubmed_abstract	Govindaraj 2014 CCR. Abstract-only — no PMC record (non-OA in 2014 CCR). The reduction is specifically in TNFR2+ Treg subset, not total FoxP3+ Tregs. Panobinostat identified as the primary Treg-reducing agent in combination. Important conceptual anchor for HDACi-driven Treg-subset targeting in AML.
Azacitidine + panobinostatazacitidine (DNMTi) + panobinostat (HDACi)	Acute myeloid leukemia	14	Govindaraj C ··· Plebanski M Clin Cancer Res, 2014	bone-marrow	flow	↓ sig.	significant reduction depletion	Moderate	Moderate abstract-onlysmall-n	PMID DOI critique	24297862	paired-pre-post	panobinostat + azacitidine 28-day cycles (dose not specified in abstract)	CD4+CD25hi FOXP3+ TNFR2+ subset	frequency-of-CD4	mid	after 28-day cycle(s) of aza+pano	elevated vs healthy donors, high BM-migration potential	decreased in vivo after aza+pano	TNFR2+ Treg reduction in bone marrow (no numeric % in abstract)	reported significant; associated with increased IFN-γ, IL-2 in bone marrow effector T cells	—	pre-specified-endpoint	pubmed_abstract	Govindaraj 2014 CCR — bone-marrow companion row to the PBMC row. Same caveats: abstract-only, no numerics, TNFR2+ subset not total FoxP3+.
Single-dose denileukin diftitox + gp100:209-217(210M) peptide vaccinegp100:209-217(210M) peptide vaccine + IFA + IL-2	Advanced melanoma	60	Luke JJ ··· Gajewski TF J Immunother Cancer, 2016	PBMC	flow	≈ n.s.	null result null	Low	Moderate cd25-gating-confoundindustry-sponsored	PMID PMCID DOI critique	27330808	randomized-controlled	Single dose pre-vaccine	CD4+CD25highFoxP3+	frequency-of-CD4	early-on-treatment	pre-vaccine post-DD	Not numerically reported; individual patient data shown in Fig 2 without quantification	Not reproducibly reduced; frequencies similar to control patients	No reproducible reduction in peripheral Treg numbers; 1/1 evaluable paired tumor biopsy (patient 16) showed INCREASED FoxP3 transcripts post-DD	n.s. (no depletion)	—	mechanism-targeted	pmc_full_text	Luke 2016 RCT — single-dose DD failed to deplete Tregs or enhance vaccine responses. Full-text backfill 2026-04-23: no numeric baseline/post Treg values in text (Fig 2 unquantified); 1 paired tumor biopsy showed INCREASED intratumoral FoxP3 post-DD. Conflict pair with Morse 2008 (18519811): Luke=single-dose→fail; Morse=multi-dose→succeed. Reconciled by schedule dependence, not truly conflicting.
Entinostat + high-dose IL-2entinostat (HDACi, class I) + high-dose IL-2	Metastatic clear-cell renal cell carcinoma	12	Pili R ··· Carducci MA Clin Cancer Res, 2017	PBMC	flow	↓ sig.	nonsignificant trend depletion	Moderate	Some concerns surface-marker-gatingno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI NCT critique	28939740	paired-pre-post	entinostat oral priming C1D-14 through C1D1; HD IL-2 standard	CD4+CD25hi T cells	frequency-of-CD4	early-on-treatment	C1D1 (post-priming entinostat, pre-IL-2)	pretreatment CD4+CD25hi frequency (not numeric in manuscript)	statistically significant decline in responders (n=5) vs progressors (n=7)	Treg frequency lower at C1D1 in responders (p=0.0273)	p=0.0273 responders lower than progressors at C1D1; p=0.03 lower Treg overall associated with response	measured across priming phase only (2 weeks entinostat)	mechanism-targeted	pmc_full_text	Pili 2017 CCR. PMC full text (PMC5712266). Class I HDACi entinostat was explicitly selected to downregulate FOXP3 and Treg suppressive function (preclinical rationale). Clinical data show a modest response-associated Treg decline, but absolute numerics are not reported for the whole cohort — only 12 of 47 had complete Treg samples. Design: priming phase (entinostat alone for 2 weeks) then combined with HD IL-2.
Entinostat + high-dose IL-2entinostat (HDACi, class I) + high-dose IL-2	Metastatic clear-cell renal cell carcinoma	3	Pili R ··· Carducci MA Clin Cancer Res, 2017	tumor	flow	mixed	nonsignificant trend depletion	Moderate	Some concerns surface-marker-gatingno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI NCT critique	28939740	paired-pre-post	entinostat oral priming C1D-14 through C1D1; HD IL-2 standard	CD4+FOXP3+ (IHC)	other	early-on-treatment	post-entinostat 2-week priming; before HD IL-2	pretreatment tumor Treg infiltration	stable or decreased Treg infiltration in 3 paired biopsies	stable or decreased (no quantitative increase despite HD IL-2 co-administration)	not tested	—	mechanism-targeted	pmc_full_text	Pili 2017 CCR — tumor biopsy companion row, N=3 only. Value of row is that HD IL-2 alone is known to expand Tregs; entinostat priming prevented that expansion in the intratumoral compartment. Not statistically powered.
Neoadjuvant chemotherapy + ipilimumabNeoadjuvant chemotherapy + ipilimumab	Early-stage NSCLC (IB-IIIA)	24	Yi JS ··· Weinhold KJ Clin Cancer Res, 2017	PBMC	flow	↑ sig.	increase null	Low	Moderate small-npbmc-only-generalizedno-multiplicity-correction	PMID PMCID DOI critique	28951518	window-of-opportunity	Per trial	CD4+FOXP3+	frequency-of-CD4	early-on-treatment	Baseline vs post-neoadjuvant	Per-patient baseline; aggregate not tabulated	Median frequency of CD4+CD25+FOXP3+ Tregs increased slightly by 1.05% (V1 chemo-alone to V3 post-ipi)	Median +1.05% absolute increase in Treg frequency from V1 to V3 (statistically significant but biologically modest); CTLA-4+ Tregs unchanged	p=0.012 (Wilcoxon signed-rank)	—	exploratory-but-prespecified	pmc_full_text	Yi 2017 — ipi+chemo neoadj NSCLC. Full-text backfill 2026-04-23: median +1.05% Treg frequency increase post-ipi (p=0.012, statistically significant but tiny magnitude and opposite direction from depletion); CTLA-4+ Treg subset unchanged.
Neoadjuvant chemotherapy + ipilimumabNeoadjuvant chemotherapy + ipilimumab	Early-stage NSCLC (IB-IIIA)	7	Yi JS ··· Weinhold KJ Clin Cancer Res, 2017	tumor	flow	—	null result null	Low	Moderate small-npbmc-only-generalizedno-multiplicity-correction	PMID PMCID DOI critique	28951518	window-of-opportunity	Per trial	FOXP3+	frequency-of-lymphocytes	early-on-treatment	Surgical resection post-neoadj	Not measured — no pretreatment tumor samples available	Higher and more variable Treg levels than PBMCs in 7 post-resection tumors; no pre/post comparison	Cannot assess change — authors state 'in the absence of pre-treatment tumor samples, we are unable to adequately assess the effect of therapy on Treg frequencies within the tumor microenvironment'	not tested (paired analysis impossible)	—	exploratory-but-prespecified	pmc_full_text	Yi 2017 tumor compartment. Full-text backfill 2026-04-23: authors explicitly disclaim ability to assess Treg change in tumor (no pre-treatment biopsy). This row is 'not-assessed' correctly — retained as a documented limitation and because post-treatment tumor Treg data is still in-scope per schema.
Denileukin diftitox	Mycosis fungoides / Sézary syndrome (CTCL)	15	Geskin LJ ··· Falo LD Cancer Immunol Immunother, 2018	PBMC	flow	↓ −29%	significant reduction depletion	Moderate	Moderate cd25-gating-confoundsmall-nno-multiplicity-correction	PMID PMCID DOI critique	29204699	single-arm-descriptive	Standard	CD4+CD25highFoxP3+	frequency-of-CD4	early-on-treatment	during treatment cycles	Not tabulated; CD4+FoxP3+ baseline count in blood per patient (n=15)	Median relative intra-patient change −29% in CD4+FoxP3+ (94% CI −83% to −20%) after one DD cycle	Median 29% reduction after one cycle (94% CI −83% to −20%); responders had significant 20-45% absolute number decrease after cycle 1; consecutive cycles produced sustained depletion	p=0.03	Transient single-cycle effect with rebound between cycles; sustained with consecutive cycles	mechanism-targeted	pmc_full_text	Geskin 2018 — CTCL DD study. Full-text backfill 2026-04-23: median 29% reduction (94% CI −83%–−20%), p=0.03; responders achieved 20-45% decrease after cycle 1. Skin Treg decrease qualitatively shown in representative patient (unquantified).
Panobinostat	HIV-1 infection on suppressive cART (reservoir reactivation study)	15	Brinkmann CR ··· Tolstrup M mSphere, 2018	PBMC	flow	↑ +40%	increase null	Moderate	Low small-nno-multiplicity-correction	PMID PMCID DOI NCT critique	29468194	paired-pre-post	20 mg PO 3× weekly x 8 weeks	CD4+CD25+FOXP3+	frequency-of-CD4	early-on-treatment	day 4 (peak) and day 28 (on-treatment)	baseline Treg proportion (value not numerically reported in abstract)	+40% at day 4; sustained elevation at day 28	+40% Treg proportion at day 4 (p=0.003); similar elevation at day 28 (p=0.004)	p=0.003 (day 4); p=0.004 (day 28)	Treg elevation returned to baseline by 24-week follow-up post-treatment	pre-specified-endpoint	pmc_full_text	Brinkmann 2018 mSphere. PMC full text (PMC5812898). HIV reservoir-reactivation study — the pre-specified hypothesis was not Treg-depletion per se, but Treg PD was pre-specified. Panobinostat INCREASED Treg frequency and activation markers, opposite to the HDACi-driven Treg depletion paradigm in oncology. Important cross-indication counter-example: HDACi effects are context-dependent. Transient — returned to baseline 24 weeks post-treatment.
Ipilimumab or tremelimumab	Melanoma, prostate, bladder	45	Sharma A ··· Sharma P Clin Cancer Res, 2019	tumor	flow	↑ sig.	increase null	High	Moderate foxp3-ihc-single-markersmall-nbaseline-imbalance	PMID PMCID DOI critique	30054281	paired-pre-post	Per trial	FOXP3+	frequency-of-lymphocytes	early-on-treatment	Paired pre/post	Numeric density values shown only in Fig 2A plots without text quantification	Numeric density values shown only in Fig 2A plots without text quantification; direction is INCREASE across melanoma (ipi n=16 post vs 19 untreated; treme n=18 paired), no reduction in bladder (n=9) or prostate (n=16)	Both ipi (Mann-Whitney unpaired) and treme (Wilcoxon paired) significantly INCREASED FOXP3+ density in melanoma, p<0.05; 2 pre-treme patients with zero baseline FOXP3+ gained FOXP3+ infiltrate post-treme; bladder and prostate: no reduction	p<0.05 for melanoma cohorts (both ipi and treme); exact p-values not stated in text (figure-only)	—	mechanism-targeted	pmc_full_text	Sharma 2019 — foundational human null result. Full-text backfill 2026-04-23: both anti-CTLA-4 agents increased rather than depleted intratumoral FOXP3+ cells across melanoma, bladder, and prostate; exact density numbers in Fig 2A plots not textually quantified in paper. Consistent with Huang 2011 (21558401) baseline 35→167 cells/mm^2; both support the absence of Fc-mediated Treg depletion by non-Fc-enhanced anti-CTLA-4 in humans. Conflict pair with Ager 2026 (41759531): Ager shows Fc-enhanced anti-CTLA-4 DOES deplete intratumoral Tregs — the Fc engineering is the reconciling difference.
Bempegaldesleukin (NKTR-214)	Advanced/metastatic solid tumors	28	Bentebibel SE ··· Diab A Cancer Discov, 2019	PBMC	flow	↑ qual.	ratio-shift only fraction shift	Low	Moderate abstract-onlyspin-abstract-vs-resultsindustry-sponsored	PMID DOI NCT critique	30988166	single-arm-descriptive	Dose-escalation	CD4+CD25+FoxP3+	ratio-to-CD8	early-on-treatment	Post-dose serial	—	—	CD8/Treg ratio increased; absolute Treg expansion but skewed toward effectors	reported	—	pre-specified-endpoint	pubmed_abstract	Bentebibel 2019 — bempeg first-in-human. Full-text backfill 2026-04-23: not in Europe PMC OA (inEPMC=N, subscription required). Abstract retains 'promoted immune cell increase with limited increase of Tregs' — consistent with the ratio-shift path. No numeric baseline/post Treg, CD8:Treg ratio, or fold change extractable from abstract.
Bempegaldesleukin (NKTR-214) + nivolumabNKTR-214 + nivolumab	Advanced solid tumors (melanoma, RCC, NSCLC, UC)	38	Diab A ··· Cho DC Cancer Discov, 2020	PBMC	flow	↑ qual.	ratio-shift only fraction shift	Low	Serious abstract-onlyspin-abstract-vs-resultsindustry-sponsored	PMID DOI NCT critique	32439653	single-arm-descriptive	Dose-escalation	CD4+CD25+FoxP3+	ratio-to-CD8	early-on-treatment	Post-dose serial	—	—	CD8/Treg ratio shift; Treg absolute expansion	reported	—	pre-specified-endpoint	pubmed_abstract	Diab 2020 PIVOT-02 — bempeg+nivo. Full-text backfill 2026-04-23: not in Europe PMC OA (inEPMC=N, subscription required). Abstract: 'increased infiltration, activation, and cytotoxicity of CD8+ T cells, without regulatory T-cell enhancement' — supports ratio-shift framing but no numeric values extractable.
Vorinostat + tamoxifen + pembrolizumabvorinostat (HDACi) + tamoxifen (endocrine) + pembrolizumab	ER+ advanced breast cancer, hormone-refractory	21	Terranova-Barberio M ··· Munster PN Nat Commun, 2020	tumor	flow	↓ −75.4%	significant reduction depletion	Moderate	Some concerns small-nresponder-subset-reportingindustry-sponsored	PMID PMCID DOI NCT critique	32681091	paired-pre-post	vorinostat + tamoxifen run-in, then add pembrolizumab	CD4+FoxP3+CTLA-4+ (activated Treg)	frequency-of-CD4	early-on-treatment	post-run-in (pre-pembrolizumab)	11.8% CD4+FoxP3+CTLA-4+ overall; 20.3% responders; 10.4% non-responders	2.9% overall; 4.2% responders; 2.5% non-responders	overall 11.8% → 2.9% (−75%); responders 20.3% → 4.2% (−79%); non-responders 10.4% → 2.5% (−76%)	p=0.0067 overall; p=0.031 responders; p=0.034 non-responders	measured at end of run-in (~4 weeks)	mechanism-targeted	pmc_full_text	Terranova-Barberio 2020 Nat Commun. PMC full text (PMC7367885). Vorinostat HDACi-driven intratumoral Treg depletion replicated across responders and non-responders — notable because Treg depletion alone was not sufficient to predict benefit in this cohort. PBMC Tregs were not reported as depleted (tumor-specific effect).
Denileukin diftitox + IFNαIFNα	Advanced ovarian cancer	12	Thibodeaux SR ··· Curiel TJ Clin Cancer Res, 2021	PBMC	flow	↓ sig.	significant reduction depletion	Low	Serious cd25-gating-confoundabstract-onlycase-report-n-of-fewspin-abstract-vs-results	PMID DOI critique	33771857	single-arm-descriptive	Per protocol	CD4+CD25highFoxP3+	frequency-of-CD4	early-on-treatment	during combination cycles	—	—	augmented depletion with IFNα	reported significant	—	mechanism-targeted	pubmed_abstract	Thibodeaux 2021 — DD+IFNα in ovarian cancer. Full-text backfill 2026-04-23: Clin Cancer Res article not in Europe PMC OA (inEPMC=N, subscription required). Abstract retains qualitative finding that DD depleted Tregs and IFNα augmented antitumor immunity; no numeric baseline/post Treg values extractable.
Low-dose decitabine	Immune thrombocytopenia (ITP)	32	Han P ··· Hou M Blood, 2021	PBMC	flow	↑ sig.	increase null	Moderate	Moderate no-multiplicity-correction	PMID PMCID DOI critique	33876188	paired-pre-post	3.5 mg/m² IV x 3 days per 4-week cycle; 3 cycles total	CD4+CD25+FOXP3+	frequency-of-CD4	late	day 85 (post-3-cycle completion, ~12 weeks)	pretreatment Treg frequency (not numerically reported)	significantly increased Treg quantity and suppressive function	Treg quantity and function 'substantially improved' (p<0.05); no numeric % reported	p<0.05	measured at day 85 post-initiation; durable through treatment window	pre-specified-endpoint	pmc_full_text	Han 2021 Blood (PMC8394906). Companion suppressive-function assay is paired with the count data — Tregs increased in BOTH quantity AND function (ex vivo CFSE-based suppression assay). This is opposite to Treg-depletion phenotype — an important null/counter-example for DNMTi effect on Tregs. Low-dose decitabine in autoimmunity tilts the balance TOWARD Treg dominance.
Entinostat + nivolumab ± ipilimumabentinostat (HDACi) + nivolumab ± ipilimumab	Advanced solid tumors (multiple histologies)	14	Roussos Torres ET ··· Connolly RM Clin Cancer Res, 2021	tumor	flow	↑ +119.7%	nonsignificant trend ratio only	Moderate	Moderate compartment-dissociationsmall-nno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI NCT critique	34135021	paired-pre-post	entinostat 5 mg PO weekly (2-week run-in, then with ICI)	FOXP3+ cells (tumor IHC); CD8/FoxP3 ratio	ratio-to-CD8	mid	Post-8-week combination (T2) vs baseline (T0)	median CD8/FoxP3 ratio 4.11 (baseline)	median CD8/FoxP3 ratio 9.03 at T2 (post-combination)	CD8:FoxP3 ratio increased 4.11 → 9.03 (2.2× shift in favor of CD8)	p=0.002 (Wilcoxon signed-rank, T0 vs T2)	measured at 8 weeks combination therapy	pre-specified-endpoint	pmc_full_text	Roussos Torres 2021 CCR (ETCTN-9844). PMC full text (PMC8563383). The ratio-shift is driven by both CD8 expansion and FoxP3 decrease; absolute Treg change not separately quantified. Entinostat monotherapy was insufficient — ICI was required for the shift. Note pct_change is for the CD8/FoxP3 ratio, not Tregs directly.
GWN323 (anti-GITR) ± spartalizumab (anti-PD-1)Spartalizumab	Advanced solid tumors and lymphomas	53	Piha-Paul SA ··· Bedard PL J Immunother Cancer, 2021	PBMC	flow	↓ qual.	nonsignificant trend depletion	Low	Moderate dose-mixed-analysisexploratory-endpointsingle-arm-no-controlindustry-sponsored	PMID PMCID DOI NCT critique	34389618	single-arm-descriptive	Dose-escalation	CD4+FoxP3+	frequency-of-CD4	early-on-treatment	On-treatment	Most patients had <1% FoxP3 at screening; 4 patients had >1% FoxP3 at baseline	3 of 4 high-baseline patients had on-treatment FoxP3 decrease; non-dose-dependent	Inconsistent decreases limited to the 4/53 patients with detectable baseline Tregs; 3/4 of those decreased; non-dose-dependent across 10-1500 mg range	not tested; insufficient N	—	pre-specified-endpoint	pmc_full_text	Piha-Paul 2021 JITC GWN323 phase I. Full-text backfill 2026-04-23: authors explicitly note 'because most patients have no Foxp3 Treg cells in the tumor microenvironment at baseline, it may be impossible to test the hypothesis that GWN323 decreases Treg cells'; only 4 patients had assessable baseline Tregs; 3/4 decreased; non-dose-dependent. Classic anti-GITR human mixed result; underpowered for depletion claim.
Mogamulizumab	Sézary syndrome (CTCL)	26	Roelens M ··· Moins-Teisserenc H Br J Dermatol, 2022	PBMC	flow	↓ sig.	significant reduction depletion	Moderate	Moderate abstract-onlyindustry-sponsored	PMID DOI critique	35041763	paired-pre-post	Standard	CD4+CD25+FOXP3+ activated Tregs	frequency-of-CD4	early-on-treatment	Within first 4 weeks	—	—	drastic decrease in activated Tregs within first 4 weeks	reported significant	Long-term immune restoration noted	mechanism-targeted	pubmed_abstract	Roelens 2022 — long-term moga PD in Sézary syndrome. Full-text backfill 2026-04-23: Br J Dermatol article not in Europe PMC OA (inEPMC=N, subscription required). Abstract retains qualitative 'drastic decrease in activated Tregs within first 4 weeks'; 17/26 patients with early complete blood response correlated with higher baseline CCR4 expression. No numeric baseline/post values extractable.
Mogamulizumab	Sézary syndrome (CTCL)	26	Roelens M ··· Moins-Teisserenc H Br J Dermatol, 2022	skin	flow	mixed	significant reduction depletion	Moderate	Moderate abstract-onlyindustry-sponsored	PMID DOI critique	35041763	paired-pre-post	Standard	CD4+CD25+FOXP3+	frequency-of-CD4	late	Long-term follow-up	—	—	Long-term immune restoration; tumor heterogeneity reshaped	reported	Long-term restoration observed	mechanism-targeted	pubmed_abstract	Roelens 2022 skin compartment. Full-text backfill 2026-04-23: article not OA; abstract only describes 'long-term immune restoration' qualitatively for the skin compartment. No numeric skin Treg data extractable.
Idelalisib	Relapsed/refractory chronic lymphocytic leukemia	9	Gadi D ··· Brown JR Br J Haematol, 2022	PBMC	flow	↓ sig.	significant reduction depletion	Low	Moderate very-small-nno-multiplicity-correction	PMID PMCID DOI NCT critique	35170759	paired-pre-post	150 mg PO BID (per-label)	CD4+CD25+FOXP3+	frequency-of-CD4	early-on-treatment	visits 6-12 (roughly month 1-3 on treatment)	pre-treatment Treg frequency (not numerically reported)	uniformly decreased across all 9 patients	uniform decrease in Tregs with PI3Kδ inhibition (no numeric % in paper)	uniform across cohort; CD8:Treg ratio increase only statistically significant in toxicity subgroup (p<0.05)	—	mechanism-targeted	pmc_full_text	Gadi 2022 Br J Haematol. PMC full text (PMC9263710). Second trial is NCT01539291 (GS-US-312-0117). Small translational cohort (N=9) from the idelalisib phase III registration. Uniform Treg decrease reported qualitatively; no absolute numerics in full text. Toxicity-driving mechanism appears to be Th17 activation on top of Treg loss, not Treg loss alone.
Iberdomide (CC-220, cereblon modulator / Ikaros-Aiolos degrader)	Active systemic lupus erythematosus	81	Lipsky PE ··· Schafer PH Ann Rheum Dis, 2022	PBMC	flow	↑ +104.9%	increase null	Moderate	Some concerns no-multiplicity-correctionexploratory-endpointindustry-sponsoredauthor-coi-sponsor	PMID PMCID DOI NCT critique	35477518	randomized-controlled	0.45 mg PO daily	TSDR demethylation (epigenetic Treg definition)	frequency-of-CD4	late	week 24 vs placebo	baseline Treg frequency (not reported numerically in abstract/paper)	+104.9% vs placebo at 0.45 mg	+104.9% increase in Tregs at 0.45 mg vs placebo at week 24	p<0.001 vs placebo	measured at week 24 on continued dosing	pre-specified-endpoint	pmc_full_text	Lipsky 2022 Ann Rheum Dis (PMC9279852). CELMoD iberdomide degrades Ikaros (IKZF1) and Aiolos (IKZF3) — NOT Helios (IKZF2). The consequence in SLE is IL-2-driven Treg EXPANSION, not destabilization. Critical conceptual boundary: Ikaros/Aiolos-degrading CELMoDs do not destabilize Tregs; an IKZF2-selective degrader would be required. This row anchors that distinction and corrects an initial assumption in the scope expansion.
TRX518 (anti-GITR) single agent or combo (gemcitabine / pembrolizumab / nivolumab)Gemcitabine OR pembrolizumab OR nivolumab	Advanced solid tumors	109	Davar D ··· Luke JJ Clin Cancer Res, 2022	PBMC	flow	mixed	nonsignificant trend depletion	Low	Moderate dose-mixed-analysissingle-arm-no-controlno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI NCT critique	35499569	single-arm-descriptive	Dose-escalation	CD4+FoxP3+	frequency-of-CD4	early-on-treatment	Serial; combo-regimen-dependent kinetics	Baseline GITR+ Treg frequency: mean 41.82%, range 9.11-100% (Part C gem+TRX518 cohort); mean 36.63%, range 12.25-54.95% (Parts D+E anti-PD-1 combo cohort)	Monotherapy (Parts A+B, n=43): SD patients (n=11) had 'more substantial decreases in peripheral Tregs and eTregs' than PD patients; GITR+ Tregs consistently reduced in all patients, greater reduction in PD; gem combo: gem reduced Tregs days 1/8, TRX518 expanded on day 2 (cycle-level oscillation); anti-PD-1 combo: less consistent, CR patient had Treg INCREASE	Modest, inconsistent peripheral Treg reductions; kinetics varied by combo regimen; numeric fold-changes not text-quantified (figure-only)	p<0.05 indicated by asterisks in Fig 3A (unpaired t-test); exact values not stated	Combo-regimen-dependent: gem oscillation d1/8 (down) vs d2 (up with TRX518); anti-PD-1 less consistent	pre-specified-endpoint	pmc_full_text	Davar 2022 TRX518 Phase Ib. Full-text backfill 2026-04-23: modest inconsistent peripheral Treg reductions with significant combo-regimen dependence; baseline GITR+ frequency means provided (41.82% gem cohort, 36.63% anti-PD-1 cohort); ORR 3.2-12.5% across parts. Quantitative values mostly in figures without text numbers.
TRX518 (anti-GITR) single agent or comboGemcitabine OR pembrolizumab OR nivolumab	Advanced solid tumors	20	Davar D ··· Luke JJ Clin Cancer Res, 2022	tumor	flow	mixed	nonsignificant trend depletion	Low	Moderate dose-mixed-analysissingle-arm-no-controlno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI NCT critique	35499569	paired-pre-post	Dose-escalation	FoxP3+	frequency-of-lymphocytes	early-on-treatment	Pre and on-treatment biopsy	Per-tumor baseline; Fig 4A shows trajectories without tabulated density	Monotherapy (Parts A+B, n=13 evaluable): 7/13 had decreased intratumoral Tregs at C1D21; anti-PD-1 combo (Parts D+E, n=9 paired): 3 responders had 'profound intratumoral Treg reductions', 6 non-responders had increased or stable tumor Tregs; gem combo (Part C, n=21): no consistent change, clinical-benefit patients had INCREASED GITR+ Treg infiltration	Monotherapy: 7/13 (54%) decreased tumor Tregs; anti-PD-1 combo: 3/3 responders deep-depleters vs 6/6 non-responders not; gem combo: paradoxical increase in clinical benefiters	Paired t-test performed on Fig 4B; exact p not stated in text	C1D21 biopsy timepoint	pre-specified-endpoint	pmc_full_text	Davar 2022 TRX518 tumor. Full-text backfill 2026-04-23: 7/13 monotherapy patients depleted; 3/3 anti-PD-1 responders vs 0/6 non-responders (striking correlation); gem combo paradoxically enriched GITR+ Tregs in responders.
Decitabine + ipilimumabdecitabine (DNMTi) + ipilimumab (anti-CTLA-4)	Relapsed/refractory AML/MDS (post-allo-HCT and transplant-naïve)	36	Penter L ··· Wu CJ Blood, 2023	bone-marrow	flow	↑ sig.	increase null	Moderate	Moderate small-nno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI NCT critique	36706355	paired-pre-post	decitabine 20 mg/m² IV d1-5 lead-in; ipilimumab 3 or 10 mg/kg q3w	CD3+FOXP3+ (scRNA + multiplex IF)	frequency-of-CD3	early-on-treatment	post-decitabine lead-in + post-ipilimumab infusion(s)	baseline marrow Treg density (not numerically reported)	increased CD3+FOXP3+ density post-ipilimumab	ipilimumab increased marrow-infiltrating Treg frequency (validated by scRNA-seq + mIF)	reported as qualitatively significant (no p-value cited in text)	—	mechanism-targeted	pmc_full_text	Penter 2023 Blood (PMC10122106) / Garcia 2023 Blood companion paper (PMID 36332187). This finding CONFLICTS with the canonical view that anti-CTLA-4 depletes tumor Tregs via ADCC. In the AML/MDS bone-marrow niche, ipilimumab (non-Fc-enhanced) appears to expand rather than deplete Tregs. Supports the Ager 2026 / Sharma 2019 Fc-engineering-reconciling-variable framework surfaced in the prior backfill run. Reinforces the rationale for Fc-enhanced anti-CTLA-4 (botensilimab, etc.) in Treg-rich niches.
Mogamulizumab (KW-0761) — anti-CCR4 Treg-targeted	CCR4-negative advanced solid tumors	49	Fujikawa K ··· Wada H PLoS One, 2023	PBMC	flow	↓ −90%	significant reduction depletion	High	Moderate industry-sponsoredauthor-coi-sponsor	PMID PMCID DOI critique	37729184	paired-pre-post	Per 1a/1b dose-escalation	CCR4+FOXP3+ eTreg	frequency-of-CD4	early-on-treatment	2 weeks post-first dose; sustained during treatment	Median 2.1% eTreg of CD4+ T cells (range 0.45-6.1%)	Median 0.20% eTreg of CD4+ (range 0.04-0.92%) at 4 weeks post-first dose	Median ~90% reduction (from 2.1% to 0.20%); eTreg depletion in all but 2 patients (where samples unavailable), 37 evaluable of 49	reported in all patients (p-value not stated)	Sustained eTreg depletion during treatment, particularly durable in patients with clinical response	mechanism-targeted	pmc_full_text	Fujikawa 2023 PLoS One — integrated phase 1a/1b mogamulizumab. Full-text backfill 2026-04-23: median 2.1% → 0.20% at 4 weeks (~90% reduction); dose-independent (0.1-1.0 mg/kg).
Tiragolumab + atezolizumabatezolizumab (anti-PD-L1)	PD-L1-positive NSCLC	16	Guan X ··· Patil NS Nature, 2024	PBMC	flow	↓ qual.	nonsignificant trend depletion	Low	Moderate small-nexploratory-endpointindustry-sponsoredauthor-coi-sponsor	PMID PMCID DOI NCT critique	38418879	paired-pre-post	tiragolumab 600 mg q3w + atezolizumab 1200 mg q3w	FOXP3+ fraction of CD4+ T cells	frequency-of-CD4	early-on-treatment	on-treatment (cycle not specified in text)	circulating Treg fraction of CD4+ (value not numerically reported)	decreased fraction of total CD4+ T cells	Treg fraction of total CD4 decreased on treatment (magnitude not quantified in manuscript)	reported as a group-level change; p-value not given	—	exploratory-but-prespecified	pmc_full_text	Guan 2024 Nature (CITYSCAPE NCT03563716 + phase 1b GO30103 translational cohort). PMC full text (PMC11139643). Peripheral Treg frequency decreased on-treatment but tumor Treg depletion not directly demonstrated in clinical samples — mouse CT26 data supported FcγR-dependent gene-signature shift only. Cleanest anti-TIGIT clinical Treg PD data available as of 2026-04-23; lacking numeric values.
Iberdomide + dexamethasonedexamethasone	Relapsed/refractory multiple myeloma	197	Amatangelo M ··· Thakurta A Cell Rep Med, 2024	bone-marrow	flow	—	null result null	Moderate	Moderate exploratory-endpointno-multiplicity-correctionindustry-sponsoredauthor-coi-sponsor	PMID PMCID DOI NCT critique	38776914	single-arm-descriptive	iberdomide 0.3-1.6 mg PO + dex (CC-220-MM-001)	CD4+ regulatory phenotype (flow)	frequency-of-CD4	baseline	screening bone marrow aspirate	6.9% of CD4+ T cells Treg phenotype (all-comer median)	non-responders significantly higher baseline Treg vs responders (p<0.001)	baseline-only biomarker (no longitudinal depletion reported)	non-responder vs responder Treg: p<0.001 (baseline biomarker)	—	exploratory-but-prespecified	pmc_full_text	Amatangelo 2024 Cell Rep Med (PMC11228401). CC-220-MM-001 translational analysis. This row captures the BASELINE Treg biomarker finding, not a longitudinal depletion. Iberdomide in MM does NOT demonstrably deplete Tregs; baseline Treg elevation is a resistance marker. Confirms Lipsky 2022 SLE finding that iberdomide's mechanism is not Treg-depletion.
Bempegaldesleukin + nivolumabNKTR-214 + nivolumab	Melanoma / RCC / UC	40	Gogas H ··· Lebbé C NPJ Precis Oncol, 2024	PBMC	flow	↑ +800%	ratio-shift only fraction shift	Moderate	Moderate no-multiplicity-correctionindustry-sponsored	PMID PMCID DOI NCT critique	39025948	paired-pre-post	Per trial	CD4+FOXP3+	ratio-to-CD8	early-on-treatment	Serial on-treatment	C1D1 baseline Treg frequency; per-patient data shown in Fig 4	~8-10x increase in absolute CD4+CD25+FoxP3+ Treg counts at C1D8 and C5D8 (bempeg+nivo arm); CD8+ T cells ~2x increase at same timepoints	Bempeg+nivo arm: Treg ~8-10x ABSOLUTE expansion but CD8/Treg ratio DECREASED at C1D8/C5D1/C5D8 vs baseline (Treg expansion outpaced CD8 2x expansion); nivo-monotherapy arm: ratios stable	p<0.00001 / p<0.0001 (multiple comparisons via asterisks in Fig 4; exact p not stated)	Serial sampling C1D1/C1D8/C5D1/C5D8	pre-specified-endpoint	pmc_full_text	Gogas 2024 — PIVOT-02 biomarker re-analysis. Full-text backfill 2026-04-23: 8-10x Treg absolute expansion with bempeg+nivo (ratio shift NEGATIVE — Treg outpaced CD8); this actually argues AGAINST bempeg achieving a favorable Treg-relative effect. Row retained per user's explicit ratio-shift-path inclusion spec but directionality is unfavorable.
Botensilimab (AGEN1181) + balstilimab (anti-PD-1)Balstilimab	Advanced solid tumors (poorly ICI-responsive)	12	Chand D ··· Stein RB Cancer Discov, 2024	tumor	flow	↓ sig.	significant reduction depletion	Moderate	Moderate small-nsingle-arm-no-controlindustry-sponsoredauthor-coi-sponsor	PMID PMCID DOI NCT critique	39083809	paired-pre-post	Per dose-escalation	FOXP3+	frequency-of-lymphocytes	early-on-treatment	On-treatment biopsy	Not quantified in main text; RNA-seq-based signal (Fig 5H)	Significant intratumoral Treg reduction by IHC and RNA-seq signature (Fig 5H); fold-change and p-value not stated in text	Human: significant intratumoral FOXP3+ reduction (IHC); numeric values not text-quantified. Mouse CT26: 'significant intratumoral FOXP3+ Treg reduction up to 10 days posttreatment' and increased CD8/Treg ratio (Fig 1D). Human PBMC mechanism: not attributed to depletion but to reduced TGFβ1 secretion (Suppl Fig S9B).	reported significant (intratumoral IHC); exact p not stated	Mouse: sustained up to 10 days post-treatment; human: single on-treatment biopsy	mechanism-targeted	pmc_full_text	Chand 2024 Cancer Discov — botensilimab translational. Full-text backfill 2026-04-23 (via Europe PMC PMCID PMC11609826): intratumoral FOXP3+ decrease in human IHC+RNA-seq but numerics not text-quantified; PBMC/serum Tregs unchanged (tumor-selective); preclinical mouse data more quantitative. Updated with PMCID.
Intraperitoneal denileukin diftitox (ONTAK)	Recurrent refractory ovarian cancer	10	Liao JB ··· Salazar LG Gynecol Oncol, 2024	PBMC	flow	↓ −73%	significant reduction depletion	Moderate	Moderate very-small-ndose-mixed-analysisno-multiplicity-correctionauthor-coi-sponsor	PMID PMCID DOI critique	39362046	single-arm-descriptive	Dose-escalation; MTD 15 µg/kg	CD4+CD25highFoxP3+	frequency-of-CD4	early-on-treatment	serial during dose escalation	Mean FoxP3 0.1726 ± 0.0442 (n=6) at baseline (pooled dose levels)	Mean FoxP3 0.0374 ± 0.0101 (n=5) at week 8	Mean 73% reduction pooled; by dose: 5 µg/kg cohort 58.9% reduction (p=0.1500); 15 µg/kg cohort 83.2% reduction (p=0.0374); 5/9 patients achieved ≥25% reduction (secondary efficacy threshold)	p=0.0275 (pooled); p=0.0374 (15 µg/kg)	8-week on-treatment measurement; recovery not reported	mechanism-targeted	pmc_full_text	Liao 2024 — IP route in ovarian. Full-text backfill 2026-04-23: pooled p=0.0275 with 73% reduction; dose-level breakdown shows only 15 µg/kg significant (p=0.0374).
Intraperitoneal denileukin diftitox (ONTAK)	Recurrent refractory ovarian cancer	10	Liao JB ··· Salazar LG Gynecol Oncol, 2024	ascites	flow	↓ −67%	nonsignificant trend depletion	Moderate	Moderate very-small-ndose-mixed-analysisno-multiplicity-correctionauthor-coi-sponsor	PMID PMCID DOI critique	39362046	single-arm-descriptive	Dose-escalation; MTD 15 µg/kg	CD4+CD25highFoxP3+	frequency-of-CD4	early-on-treatment	serial	Mean FoxP3 0.1855 ± 0.0945 (n=3) at baseline	Mean FoxP3 0.0597 ± 0.0304 (n=3) at week 8	Mean 67% reduction in ascites FoxP3; 3 patients achieved ≥25% reduction (immunologic efficacy threshold)	p=0.2737 (n.s.; small sample)	8-week on-treatment measurement	mechanism-targeted	pmc_full_text	Liao 2024 ascites compartment. Full-text backfill 2026-04-23: 67% mean reduction, n.s. (p=0.2737) — reclassified as 'partial' because direction is consistent with PBMC but underpowered (n=3 paired).
Nemvaleukin alfa (ALKS 4230) monotherapy	Advanced solid tumors (ARTISTRY-1)	46	Vaishampayan UN ··· Velcheti V J Immunother Cancer, 2024	PBMC	flow	↑ +100%	increase ratio only	Moderate	Moderate single-arm-no-controlno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI NCT critique	39567211	single-arm-descriptive	6 µg/kg/day IV days 1-5 q21d	CD4+CD25+FoxP3+	frequency-of-CD4	early-on-treatment	Serial cycles	C1D1 baseline; per-patient data only	Max fold change from baseline (FCBmax) for Tregs ~2x in both melanoma and RCC cohorts; CD8+ 2.53x, NK 6.52x for comparison (Part B monotherapy)	Tregs expanded ~2x (modest); CD8+ 2.53x; NK 6.52x — ratios favor effector expansion. Treg expansion limited to C1D5 only	reported qualitatively; no per-comparison p-values in main text	Transient Treg expansion at C1D5 only; effector expansion at D8 (cycle 1) and D22 (cycle 2)	pre-specified-endpoint	pmc_full_text	Vaishampayan 2024 ARTISTRY-1 monotherapy. Full-text backfill 2026-04-23: Treg FCBmax ~2x vs NK 6.52x / CD8 2.53x; nemva's IL-2R-βγ bias limited Treg expansion as designed. Kept via ratio-shift path.
Nemvaleukin alfa + pembrolizumabPembrolizumab	Advanced solid tumors (ARTISTRY-1 Part C)	166	Vaishampayan UN ··· Velcheti V J Immunother Cancer, 2024	PBMC	flow	↑ qual.	increase ratio only	Moderate	Moderate single-arm-no-controlno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI NCT critique	39567211	single-arm-descriptive	6 µg/kg/day IV days 1-5 q21d	CD4+CD25+FoxP3+	frequency-of-CD4	early-on-treatment	Serial cycles	C1D1 baseline; per-patient data only	Similar minimal Treg expansion pattern as monotherapy (~2x FCBmax)	Nemva+pembro: minimal Treg expansion; effector expansion preserved	reported qualitatively	—	pre-specified-endpoint	pmc_full_text	ARTISTRY-1 Part C combo. Full-text backfill 2026-04-23: combo cohort showed the same minimal Treg expansion pattern as monotherapy; CD8/NK ratios remained favorable. No separate numeric table for combo vs mono.
Denileukin diftitox (E7777 / I/ONTAK)	Treatment-refractory stage IV breast cancer	15	Gwin WR ··· Disis ML Vaccines (Basel), 2025	PBMC	flow	↓ qual.	nonsignificant trend depletion	Low	Moderate cd25-gating-confoundsmall-nindustry-sponsored	PMID PMCID DOI critique	40006664	single-arm-descriptive	Per protocol, up to six 21-day cycles	CD4+CD25+FOXP3+	frequency-of-CD4	early-on-treatment	post-cycle	Per-patient baseline; aggregate not tabulated	No significant difference between baseline and maximal observed Treg levels (overall p=0.10)	Overall depletion n.s. (p=0.10); 6/15 patients (40%) achieved ≥25% Treg reduction with mean reduction 56.0% ± 10.96% (SD) in this responder subset	p=0.10 (overall, n.s.)	Most reductions occurred after cycle 2; limited by rapid anti-DT IgG response (100% of 9 evaluable patients at week 6, p<0.005)	mechanism-targeted	pmc_full_text	Gwin 2025 — partial depletion in breast cancer. Full-text backfill 2026-04-23: overall p=0.10 (n.s.); responder subset (6/15, 40%) had 56.0% ± 10.96% mean reduction; 100% anti-DT IgG response by week 6 likely limited efficacy.
Mogamulizumab + nivolumab neoadjuvantNivolumab (3 doses) + mogamulizumab (4 doses)	Operable solid tumors (renal, lung, esophageal, oral SCC)	16	Jinushi K ··· Wada H J Immunother Cancer, 2025	tumor	flow	↓ −86.7%	significant reduction depletion	Moderate	Low small-nsingle-arm-no-controlindustry-sponsored	PMID PMCID DOI critique	40180420	window-of-opportunity	Per protocol neoadjuvant	CCR4+FOXP3+ Tregs	frequency-of-CD4	early-on-treatment	Preoperative post-neoadjuvant window	Per-tumor baseline CCR4+FoxP3+ density, not aggregated	Median −86.7% change in CCR4+FoxP3+ (range −94.8% to −52.7%) in all 16 patients; total FoxP3+ median change −11.1% (range −73.2% to +87.8%), decreased in 50% of patients	CCR4+ eTreg: median 86.7% reduction, depleted in 16/16; total FoxP3+ mixed response, 8/16 decreased	reported; p-values not published	Measured at surgery (preoperative neoadjuvant window)	mechanism-targeted	pmc_full_text	Jinushi 2025 JITC — neoadjuvant moga+nivo. Full-text backfill 2026-04-23: median 86.7% reduction in CCR4+FoxP3+ eTregs (100% depleted); total FoxP3+ more variable (median −11.1%, 50% decreased).
Mogamulizumab + recombinant human IL-15 (rhIL-15)Mogamulizumab + rhIL-15	R/R T-cell malignancies (ATLL, MF/SS)	6	Gordon MJ ··· Roschewski M Blood Neoplasia, 2025	PBMC	flow	—	ratio-shift only depletion	Low	Moderate very-small-nsurface-marker-gatingindustry-sponsored	PMID PMCID DOI critique	40546724	single-arm-descriptive	Dose-escalation	CD4+FOXP3+ (presumed)	frequency-of-CD4	early-on-treatment	First cycle	—	—	NK expansion and tumor cell reduction reported; Treg-specific numeric not in abstract	not tested	—	mechanism-targeted	pmc_full_text	Gordon 2025 — mogamulizumab + rhIL-15 phase 1 in R/R T-cell malignancies. Full-text backfill 2026-04-23: PMC article reviewed (PMC12182836). Paper focuses on NK cell expansion and ADCC; no quantitative Treg baseline/post values, p-values, or fractional changes reported. Mechanism of mogamulizumab Treg depletion is assumed by reference to prior moga literature but NOT measured in this trial. Row retained because moga is mechanism-targeted and this is a mogamulizumab-containing regimen, but Treg PD is NOT a pre-specified endpoint here — borderline case.
Nemvaleukin alfa monotherapy	Advanced melanoma and RCC (post-ICI)	74	Calvo E ··· McDermott DF J Immunother Cancer, 2025	PBMC	flow	↑ +100%	increase ratio only	Moderate	Moderate single-arm-no-controlno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI NCT critique	40759440	single-arm-descriptive	6 µg/kg/day IV days 1-5 q21d	CD4+CD25+FoxP3+	frequency-of-CD4	early-on-treatment	Serial on-treatment	C1D1 baseline; per-patient only	Treg fold change ~2x (FCBmax at C1D5); CD8+/NK expansion larger at C1D8/C2D22	Tregs ~2x FCBmax; effector cells (CD8+/NK) FCB larger at D8/D22; post-ICI melanoma and RCC both show this pattern	reported qualitatively	—	pre-specified-endpoint	pmc_full_text	Calvo 2025 ARTISTRY-1 Part B post-ICI cohort. Full-text backfill 2026-04-23: consistent ~2x Treg FCBmax limited to C1D5, contrasting with sustained effector expansion at D8/D22.
Nemvaleukin alfa (less-frequent IV dosing)	Advanced solid tumors	30	Piha-Paul SA ··· Lakhani NJ Oncologist, 2025	PBMC	flow	↑ qual.	increase ratio only	Low	Moderate single-arm-no-controlno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI critique	40990800	single-arm-descriptive	Q-variable IV schedules	CD4+CD25+FoxP3+	frequency-of-CD4	early-on-treatment	Serial across schedules	C1D1 baseline; per-patient only	Low-level transient non-dose-dependent Treg expansion; favorable effector:Treg ratios maintained	Minimal Treg expansion across schedules; figures only	reported qualitatively	—	pre-specified-endpoint	pmc_full_text	Piha-Paul 2025 ARTISTRY-3 less-frequent dosing. Full-text backfill 2026-04-23: 'low-level, transient, non-dose-dependent expansion of Tregs' with favorable NK:Treg and CD8:Treg ratios; Treg quantitative data in Fig 6 only.
BMS-986218 (Fc-enhanced anti-CTLA-4) + androgen deprivation therapy (ADT) neoadjuvantAndrogen deprivation therapy (ADT)	High-risk localized prostate cancer	24	Ager CR ··· Dallos MC Cell Rep Med, 2026	tumor	flow	↓ sig.	significant reduction depletion	Moderate	Some concerns small-nopen-labelno-multiplicity-correctionindustry-sponsored	PMID PMCID DOI NCT critique	41759531	randomized-controlled	Per trial	FOXP3+ (scRNA-seq + CyTOF)	frequency-of-CD4	early-on-treatment	Baseline vs pre-surgery	Per-arm baseline TI-Treg frequency; density values not tabulated in text (figure panels only)	ADT+BMS-986218 arm significantly reduced TI-Treg frequency vs ADT alone (p=0.031); ADT alone arm had INCREASED TI-Tregs vs untreated (p=0.002)	Qualitative TI-Treg reduction in combo arm; residual TI-Tregs remained in all tumors (incomplete depletion); numeric density values in figure panels without text quantification	p=0.031 (ADT+BMS-986218 vs ADT alone); p=0.002 (ADT alone vs untreated, ADT expanded TI-Tregs)	Measured at radical prostatectomy (end of neoadjuvant window)	mechanism-targeted	pmc_full_text	Ager 2026 CRM — randomized phase 1 BMS-986218 (Fc-enhanced anti-CTLA-4) + ADT vs ADT alone. Full-text backfill 2026-04-23: TI-Treg reduction p=0.031 in combo arm; ADT alone expanded TI-Tregs (p=0.002); residual Tregs remained (partial depletion); CD16a+ macrophage correlation p=0.033. Numeric densities in figure panels not text-quantified. Conflict pair with Sharma 2019 (30054281): Sharma shows non-Fc-enhanced anti-CTLA-4 fails to deplete; Ager shows Fc-enhanced anti-CTLA-4 succeeds — reconciled by Fc-engineering difference.

Side-list: systematic reviews and meta-analyses¶

Show / hide

These are excluded from the primary table but retained as follow-up leads.

First author	Year	Journal	Title	Type	Links
Raeber ME	2023	EBioMedicine	A systematic review of interleukin-2-based immunotherapies in clinical trials for cancer and autoimmune diseases	systematic-review	PMID DOI
Mortezaee K	2024	Med Oncol	Selective targeting or reprogramming of intra-tumoral Tregs	narrative-review	PMID DOI
Kumar P	2019	Immunotherapy	Recent advances with Treg depleting fusion protein toxins for cancer immunotherapy	narrative-review	PMID DOI
Churov A	2021	Hum Immunol	Targeting adenosine and regulatory T cells in cancer immunotherapy	narrative-review	PMID DOI
Rudqvist NP	2023	Oncoimmunology	Next-generation CTLA-4 targeting molecules and combination therapy for cancer immunotherapy	narrative-review	PMID DOI
Tang F	2018	Cell Biosci	Anti-CTLA-4 antibodies in cancer immunotherapy: selective depletion of intratumoral regulatory T cells or checkpoint blockade?	narrative-review	PMID DOI