Priming versus propagating: distinct immune effects of alpha- versus beta-particle emitting radiopharmaceuticals when combined with immune checkpoint inhibition in mice

Abstract

Radiopharmaceutical therapy (RPT) synergises with immune checkpoint inhibitors (ICI), but comparison of immunomodulation by different radioisotopes is lacking. Here, we evaluate mechanisms of response and timing of ICI administration relative to α- (²²⁵Ac) and β-emitting (⁹⁰Y, ¹⁷⁷Lu) radioisotope therapy, coupled with alkylphosphocholine NM600, when combined with dual (anti-PD-L1 and anti-CTLA4) ICI, using syngeneic poorly immunogenic (B78 and Myc-CaP) and immunogenic (MC38) murine models. Regardless of the isotope, RPT delivering 2 Gy mean tumor dose promotes tumor regression and improves survival in B78 or MC38 tumor-bearing mice when combined with early ICI administration. Greatest anti-tumor responses are seen in MC38 to ⁹⁰Y-NM600 + ICI and in B78 and Myc-CaP to ²²⁵Ac-NM600 + ICI. Flow cytometry and single-cell RNA and T cell receptor sequencing reveal that, combined with ICI, β-emitting radioisotopes expand existing adaptive immunity, whereas α-emitting radiopharmaceuticals initiate immune priming. Thus, appropriate application of α- or β-emitting RPT in combination with ICI achieves distinct antitumor immune responses.

Preclinical studies indicate a synergistic effect of radiotherapy treatment (RT) and Immune checkpoint inhibitors (ICI) on tumor growth and metastasis. However, little is known about the immunomodulatory performance of different radioisotopes on the tumor microenvironment. Here, the authors employ alpha- versus beta-particle emitting radiopharmaceuticals in combination with dual ICI therapy and dissect mechanisms of in vivo immunomodulation and timing of ICI administration relative to RT, by comparing responses in immunogenic and non-immunogenic preclinical mouse models.

Introduction

Combinations of radiation therapy (RT) and immune checkpoint inhibitors (ICI) have shown efficacy preclinically and in certain clinical scenarios^1–5. Preclinical mechanisms by which RT enhances response to immunotherapy include inducing immunogenic cell death, releasing tumor-specific antigens and damage-associated molecular pattern molecules, and causing phenotypic changes on surviving tumor cells (upregulation of MHC-I and FAS, type I interferon secretion)^6–11. However, adding RT to a single tumor site in patients with metastatic disease has not improved the rate or duration of systemic response to standard-of-care ICIs¹². Several mechanisms have been proposed to account for this translational discrepancy, including patient selection (e.g., patients with immune exhausted tumor microenvironments have worse outcomes²), radiation of tumor draining lymph nodes (necessary for propagation of an adaptive immune response¹³), and non-optimal dosing or timing of radiation^14,15.

With a growing understanding of how RT enhances responses to immunotherapy preclinically and clinically in locally advanced disease settings, it is important to consider how these therapies may be combined to improve outcomes in patients with metastatic disease. The role of external beam radiation therapy (EBRT) is limited in treatment of metastatic disease due to potentially prohibitive toxicities associated with irradiating multiple tumor sites or the whole body to cover radiographically occult micrometastatic disease. Additionally, unirradiated distant tumor sites have been shown to abrogate the in situ vaccine effect and resulting adaptive immune response following RT at one tumor site in certain preclinical studies^16,17.

Radiopharmaceutical therapy (RPT), which delivers a radioactive isotope conjugated to a tumor-targeting vector, is administered systemically to deposit radiation absorbed dose at all sites of uptake. RPT agents have the potential to overcome some of the limitations of EBRT in metastatic disease settings. Physical properties of radiopharmaceuticals vary widely, with beta (β)—(e.g., ⁹⁰Y, ¹⁷⁷Lu) and alpha (α)-particle (e.g., ²²³Ra, ²²⁵Ac) emitting radioisotopes being the focus of much clinical interest presently. β particles are characterized by low linear energy transfer (LET) (0.2 keV/µm) and travel millimeters in tissue with sparse ionization events, whereas α particles are high LET particles (50–230 keV/µm) that travel micrometers in tissue with densely clustered ionization events along their track¹⁸. RPT agents have been studied both preclinically and in early clinical studies in combination with immunotherapy¹⁹. To capitalize on the therapeutic potential of combinations of RPT + ICIs, it is critical to understand the effects of RPT properties (e.g., radionuclide, vector), disease characteristics (e.g., cancer type, immune environment), and RPT dosing and timing so that combinations with ICI may be optimized prior to advanced phase clinical investigation.

An unmet question is to determine the radiation absorbed dose of RPT and sequencing of administration of combination RPT and ICI therapies resulting in effective anti-tumor immune responses through comparative studies of ⁹⁰Y-, ¹⁷⁷Lu-, and ²²⁵Ac-NM600. NM600 is an alkylphosphocholine analog with selective uptake and retention in a wide array of tumors²⁰. The capacity of ⁹⁰Y-, ¹⁷⁷Lu-, and ²²⁵Ac-NM600 to generate type I interferon (IFN1) responses and immunomodulate the tumor microenvironment has been reported^15,21–23. ⁹⁰Y-, ¹⁷⁷Lu-, and ²²⁵Ac-NM600 can be utilized to examine RPT + ICI combinations in both poorly immunogenic (B78 melanoma and Myc-CaP prostate cancer) and immunogenic (MC38 colorectal carcinoma) syngeneic murine tumor models. B78, derived from the spontaneous melanocytic tumor B16F10, is characterized as poorly immunogenic, given its low levels of MHC-I expression, low numbers of baseline tumor-infiltrating lymphocytes, and limited response to immune checkpoint inhibition alone^15,24,25. MC38, a carcinogen-induced model, exhibits mismatch repair deficiency²⁶ and is characterized as immunogenic, having the highest mutational load amongst ten syngeneic tumor models and a propensity to respond to anti-CTLA4²⁷ and anti-PD-L1 alone²⁸.

In the current study, using these models and NM600, we seek to evaluate the development, efficacy, and key mechanisms of anti-tumor immunity in response to differing radionuclides and varied timing of ICI delivery. Systematically delineating differing tumor responses to ²²⁵Ac-⁹⁰Y- or ¹⁷⁷Lu-NM600 in combination with dual ICI therapy (anti-PD-L1 and anti-CTLA4) in syngeneic murine tumor models is critical for efficacious translation to patients. Mechanistic insights from the immunogenicity of the tumor model may impact future study design and isotope selection and influence treatment regimens and patient selection as these therapies progress through clinical translation. We observe that ⁹⁰Y-NM600 + ICI promotes the greatest tumor regression in an immunogenic tumor model by propagating existing antitumor immune responses, while ²²⁵Ac-NM600 + ICI leads to superior antitumor responses in poorly immunogenic tumor models by priming CD8⁺ T cells. These results underscore the relevance of understanding tumor mechanisms of resistance to ICI in order to employ effective RPT + ICI combination therapy.

Results

Murine tumor models exhibit tumor growth delay and survival improvement following low absorbed dose ²²⁵Ac-NM600 + ICI

To perform absorbed dose-matched comparative immunologic studies of ⁹⁰Y-, ¹⁷⁷Lu-, and ²²⁵Ac-NM600, image-based radiation dosimetry studies were performed (NM600 chemical structure: Fig. 1A). Tumor uptake and dosimetry of ⁸⁶Y/⁹⁰Y-NM600¹⁵ and ¹⁷⁷Lu/²²⁵Ac-NM600 in mice bearing B78 melanoma tumors were reported previously²³. Maximum intensity projections of PET or SPECT scans showing MC38 selective uptake of ⁸⁶Y-NM600 (imaging surrogate for ⁹⁰Y-NM600) and ¹⁷⁷Lu-NM600, respectively, are shown in Fig. 1B, C. Absorbed doses (mean ± standard deviation) of 2.03 ± 0.05, 0.50 ± 0.01, 1.24 ± 0.08, and 2.11 ± 0.21 Gy/MBq ⁹⁰Y-NM600 and 0.47 ± 0.04, 0.08 ± 0.01, 0.46 ± 0.05, and 1.40 ± 0.24 Gy/MBq ¹⁷⁷Lu-NM600 were determined for blood, bone, spleen, and MC38 tumors, respectively (Fig. 1D, E). These absorbed dose estimates were obtained using a Monte Carlo voxel-based dosimetry method²⁹ and serial PET/CT or SPECT/CT data as well as ex vivo biodistribution data at one timepoint (Supplementary Fig. 1A, B). Estimates of absorbed dose to MC38 and normal tissues for ²²⁵Ac-NM600 were obtained by S-value calculations based on the Medical Internal Radiation Dose (MIRD) formalism using serial ex vivo biodistribution data (Supplementary Fig. 1C)³⁰. Absorbed doses of 0.163, 0.248, and 0.437 Gy/kBq ²²⁵Ac-NM600 were estimated for blood, bone, and MC38 tumor, respectively (Fig. 1F). Tumor absorbed dose estimates for all studies in this work are reported in Table 1

Fig. 1. Murine tumor models exhibit tumor growth delay and survival improvement following low absorbed dose ²²⁵Ac-NM600 + ICI.

A NM600 chemical structure. Maximum intensity projections at the indicated times post injection from B serial PET/CT images (9.25 MBq ⁸⁶Y-NM600) or C serial SPECT/CT images (18.5 MBq ¹⁷⁷Lu-NM600) for MC38 tumor-bearing mice (arrow indicates tumor). D–F Tissue dosimetry performed using Monte Carlo methods. Excised tumors activity distribution imaged with ionizing-radiation quantum imaging detector (iQID) 48 h (G, H; MC38) or 72 h (I, K; B78) post IV injection of the following RPT agents and activities: G 3.7 MBq ⁸⁶Y-NM600 + 18.5 MBq ⁹⁰Y-NM600, H 13.0 MBq ¹⁷⁷Lu-NM600, I 0.74 MBq ⁹⁰Y-NM600, J 1.70 MBq ¹⁷⁷Lu-NM600, K 18.5 kBq ²²⁵Ac-NM600. K iQID scan overlaid with photomicrograph of tumor section. Right axis denotes number of events recorded. L Treatment scheme. MC38 tumor-bearing mice were randomized to IgG2b control on days -5/0/5 (IgG2b), or 0.2 Gy (0.4625 kBq), 2 Gy (4.625 kBq) ²²⁵Ac-NM600, or no RPT on day 1 ± ICI on days −3/0/3 (ICI). M–O Tumor growth and overall survival. P Treatment scheme. B78 two-tumor-bearing mice received no treatment (No Tx), 0.2 Gy (1.85 kBq), 2 Gy (18.5 kBq) ²²⁵Ac-NM600, or no RPT on day 1 ± ICI on days 4/7/10 (ICI). Q–S Tumor growth and overall survival. B, D; C, E; F; G–K; M, N; Q–S Results of one experiment. O Results of two independent experiments. Representative image/timepoint of n = 3 (B) or n = 4 (C) replicates. D N = 3 mice: ⁹⁰Y-NM600; E n = 4 mice: ¹⁷⁷Lu-NM600; n = 9 mice used for dosimetry calculation, F n = 1 data point: ²²⁵Ac-NM600. G–K Representative image/group from n = 2 mice per treatment group. M, N N = 7: all groups. O n = 17: 2 Gy ²²⁵Ac-NM600 ± ICI, ICI, IgG2b; n = 7: 0.2 Gy ²²⁵Ac-NM600 ± ICI. Q–S N = 9: 2 Gy ²²⁵Ac-NM600 + ICI; n = 10: all other treatment groups. One-way ANOVA with Tukey’s post hoc test was used to compare ⁹⁰Y-NM600 and ¹⁷⁷Lu-NM600 absorbed doses between tissues. Linear mixed models were used to compare tumor volumes over time between various treatment groups, adjusted for multiple comparisons using Tukey’s method. Log-rank test was used to compare survival. Error bars are SEM.

Table 1.

Injected RPT activity for tumor absorbed dose prescriptions

Tumor model	RPT agent	Tumor absorbed dose prescription (Gy)	Injected activity
MC38	90Y-NM600	2	0.925 MBq
	177Lu-NM600	2	1.85 MBq
	225Ac-NM600	0.2	0.4625 kBq
	225Ac-NM600	2	4.625 kBq
	225Ac-NM600	8	18.5 kBq
B78	90Y-NM600	2	0.733 MBq
	177Lu-NM600	2	1.702 MBq
	225Ac-NM600	0.2	1.85 kBq
	225Ac-NM600	2	18.5 kBq
Myc-CaP	90Y-NM600	2	1.18 MBq
	225Ac-NM600	2	8 kBq

To determine the effects of different radionuclides on the heterogeneity of the spatial distribution of ionization radiation events in MC38 and B78 tumors, ionizing-radiation quantum imaging detector (iQID) scans of ionizing radiation event distribution were obtained on RPT treated tumors (Fig. 1G–K). Microdosimetric heterogeneity that may be present given the short tissue range of ²²⁵Ac compared to ⁹⁰Y/¹⁷⁷Lu (⁹⁰Y: 11.3 mm, ¹⁷⁷Lu: 1.8 mm, ²²⁵Ac: <0.1 mm) and not resolved by in vivo imaging can be assessed by digital autoradiography with the iQID system. These studies showed greater heterogeneity in ionizing radiation event distribution following ²²⁵Ac-NM600 than ¹⁷⁷Lu-NM600 and ⁸⁶Y-/⁹⁰Y-NM600, resulting in “cold,” unirradiated regions within the B78 tumor adjacent to “hot” regions receiving high linear energy transfer (LET) radiation from ²²⁵Ac-NM600. No large differences in ⁹⁰Y- or ¹⁷⁷Lu-NM600 ionizing radiation event distribution between MC38 and B78 tumor types were observed.

Blood and additional normal tissue toxicity assessments for ⁹⁰Y-NM600 and ¹⁷⁷Lu-NM600 have previously been reported^22,31. The acute toxicity profile for the maximum tolerated activity (MTA) of ²²⁵Ac-NM600 in naïve C57BL/6 mice (18.5 kBq) is shown in Supplementary Fig. 2. ²²⁵Ac-NM600 could be delivered to naïve mice without physical signs of toxicity during the therapy studies but induced transient cytopenias (Supplementary Fig. 2A). Importantly, no mice treated in any of the studies reported in this manuscript were identified by veterinary staff as exhibiting signs or symptoms of illness/distress and no mice lost >20% of their body weight.

Although acute hepatotoxicity was not observed following ²²⁵Ac-NM600 (Supplementary Fig. 2B), we investigated late toxicities of ²²⁵Ac-NM600 (Supplementary Fig. 3A–D), and noted elevated liver transaminases and serum creatinine, suggesting hepatic and renal dysfunction. One parameter of radiopharmaceuticals that can be adjusted to improve toxicities and lesion:non-lesion residence times is molar activity (A_m). Frequently, high A_m is desired to reduce toxicities associated with certain RPT vectors or with target saturation at low A_m³². However, in the case of ¹¹¹I-J591, lowering molar activity of the RPT agent by adding unlabeled antibody increased the lesion residence time and the ratio of lesion:liver residence times³³. For this reason, we investigated the effect of lowering the A_m of ²²⁵Ac-NM600 (Supplementary Fig. 3). The liver uptake of ²²⁵Ac-NM600 was significantly decreased in the low A_m group (0.011 MBq/nmol ²²⁵Ac-NM600) vs. high A_m (0.336 MBq/nmol ²²⁵Ac-NM600), and the MC38 uptake was not significantly different. However, late (6 month post administration) changes in mouse weight and hepatic and renal toxicity were similar in low and high A_m treatment groups. Thus, this approach of lowering the A_m did not markedly improve the biodistribution of ²²⁵Ac-NM600. A possible explanation for the minor impact on biodistribution of altering NM600 A_m is that NM600 has a lipid raft cellular uptake mechanism distinct from the receptor-ligand interaction employed by ¹¹¹I-J591 and many other RPTs, which may not be saturable.

We also examined the cytotoxicity of ²²⁵Ac relative to EBRT in both MC38 and B78, observing an expected high relative biological effectiveness (RBE) of ²²⁵Ac (Supplementary Fig. 4A, B). Additionally, we confirmed that anti-PD-L1 antibody uptake and distribution was similar between MC38 and B78 tumors (Supplementary Fig. 5). We determined that MC38 and B78 tumor mRNA expression of Pdl1 and Ctla4 was not significantly changed following treatment with ⁹⁰Y-, ¹⁷⁷Lu-, or ²²⁵Ac-NM600 + ICI (Supplementary Fig. 6A–D). Given that ²²⁵Ac-NM600 was tolerable with evidence of hematologic recovery and stable weight, we began testing ²²⁵Ac-NM600 in combination with ICI (Fig. 1L–S).

In MC38 (Fig. 1L–O and Supplementary Tables 1–3) and B78 (Fig. 1P–S and Supplementary Tables 4–6), dose finding studies were conducted to determine an absorbed dose of ²²⁵Ac-NM600 that could be administered in each tumor model with a therapeutic benefit. From our prior observation that a 2 Gy mean tumor absorbed dose was optimal for response to ⁹⁰Y-NM600 + anti-CTLA-4¹⁵, we investigated 2 and 0.2 Gy ²²⁵Ac-NM600 in MC38 and B78. B78 has lower uptake of ²²⁵Ac-NM600 (0.10 Gy/kBq)²³ than MC38 (0.437 Gy/MBq), and an estimated tumor absorbed dose greater than 2 Gy could not be achieved without exceeding toxicity limits. In the MC38 model, 2 Gy alone provided only nominal survival benefit, and while addition of ICI to 2 Gy increased survival compared to both 2 Gy alone and the control, it was not significantly different from ICI alone (Fig. 1O). In the B78 model, however, while 2 Gy provide a moderate increase in survival relative to control, addition of ICI to the 2 Gy treatment increased survival compared to control, 2 Gy, and ICI alone. Because 2 Gy ²²⁵Ac-NM600 was tolerable and demonstrated a cooperative therapeutic interaction with ICI in the B78 model, we selected this dose for further study (²²⁵Ac-NM600 + ICI vs. ²²⁵Ac-NM600 alone, overall survival) (Fig. 1M–O, Q–S). The differences in the patterns of response among MC38 and B78 tumors led us to hypothesize that the effects of RPT + ICI may be different based on the immunogenicity of the tumor model.

⁹⁰Y-NM600 + early/intermediate-timed ICI enhances therapeutic response in the immunogenic MC38 colorectal carcinoma model

Three different timings of ICI (dual anti-CTLA4 and anti-PD-L1) were investigated relative to treatment day 1 injection of activities of ⁹⁰Y-NM600, ¹⁷⁷Lu-NM600, or ²²⁵Ac-NM600 that were estimated from dosimetry to deliver 2 Gy mean tumor absorbed dose. ICI delivery was defined as early (days −3/0/3), intermediate (days 4/7/10), or delayed (days 11/14/17). The intermediate-timed ICI has been shown to be safe and effective in combination with all three RPT compounds and EBRT in murine tumor models^15,21,23,34. Considering the open clinical question of optimal sequencing of RT + ICI and effects of radioisotope half-life on tumor immunomodulation²³, we compared the effect of sequencing of RPT + ICI on tumor growth and survival (Fig. 2A–I); 16 treatment groups, repeated two separate times). Comparisons between varied ICI timings are shown in Fig. 2F, H and Supplementary Tables 7–13, and comparisons between varied radioisotopes are shown in Fig. 2G, I and Supplementary Tables 14–20.

Fig. 2. ⁹⁰Y-NM600 + early/intermediate-timed ICI enhances therapeutic response in the immune sensitive MC38 colorectal carcinoma model.

A Treatment regimen options for in vivo therapy studies. Mice either received early (days −3/0/3), intermediate (days 4/7/10), or delayed (days 11/14/17) dual anti-CTLA4 and anti-PD-L1 relative to RPT on day 1. B–I MC38 tumor-bearing mice were randomized to one of 16 treatment groups: IgG2b control on days −5/0/5 (IgG2b), dual anti-CTLA4 and anti-PD-L1 (ICI) on days −3/0/3, days 4/7/10, or days 11/14/17, 2 Gy ⁹⁰Y-NM600 (0.925 MBq), ¹⁷⁷Lu-NM600 (1.85 MBq), or ²²⁵Ac-NM600 (4.625 kBq) on day 1, or combination 2 Gy ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 on day 1 with dual ICI on days −3/0/3, days 4/7/10, or days 11/14/17. B–D Individual MC38 tumor growth curves corresponding to (F, G). F, H Effects of varied ICI timing for each radionuclide on tumor growth and overall survival. G, I Effects of varied radionuclide for each ICI timing of combination therapy on tumor growth and overall survival. ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600, and no RPT + early ICI extended overall survival compared to IgG2b. B–G Results of one experiment. H, I Results of two independent experiments. B–G N = 7: ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 ± ICI −3/0/3, ICI −3/0/3, IgG2b; n = 5: ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 + ICI 4/7/10 or + ICI 11/14/17, ICI 4/7/10, ICI 11/14/17. H, I N = 24: ICI −3/0/3, IgG2b; n = 17: ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 ICI −3/0/3; n = 10: ICI 4/7/10, ICI 11/14/17; n = 5: ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 + ICI 4/7/10 or + ICI 11/14/17. Linear mixed models were used to compare tumor volumes over time between various treatment groups. Statistical testing of the pairwise contrasts was adjusted for multiple comparisons using Tukey’s method. Log-rank test was used to compare survival. Error bars are SEM.

⁹⁰Y- and ¹⁷⁷Lu-NM600 + early ICI reduced MC38 growth rate and prolonged overall survival relative to all other groups. For ²²⁵Ac-NM600, overall survival prolongation was observed following only ²²⁵Ac-NM600 + early ICI (Fig. 2F, H). For early ICI, ⁹⁰Y-NM600 + ICI slowed tumor growth rate and prolonged overall survival compared to ²²⁵Ac-NM600 + ICI (Fig. 2G, I). For intermediate-timed ICI, ⁹⁰Y-NM600 combination therapy showed the greatest therapeutic benefit. For delayed ICI, there were a few differences between any RPT + ICI combination therapy and the control groups (Fig. 2G, I). Rates of complete response (CR) for all mice included in Fig. 2 therapy studies are shown in Supplementary Fig. 7A–D, E demonstrates the rates at which mice rendered disease-free subsequently rejected re-engraftment on the contralateral flank with the same tumor cell line they had been cured from (as a measure of immunologic memory), relative to age-matched naïve control mice. This re-engraftment was performed on day 100 after initial RPT treatment, and rates of rejection were: ⁹⁰Y-NM600 + early ICI: 8/9; ⁹⁰Y-NM600 + intermediate ICI: 4/4; ¹⁷⁷Lu-NM600 + early ICI: 5/5; ¹⁷⁷Lu-NM600 + intermediate ICI: 1/1; early ICI: 8/10; naïve C57BL/6 mice: 0/15.

²²⁵Ac-NM600 + early ICI prolongs overall survival in the poorly immunogenic B78 melanoma model

We next turned to a poorly immunogenic tumor model, B78 melanoma^24,25, to test whether the immunogenicity of a tumor model may impact patterns of response to RPT + ICI (Fig. 3A–I; 14 treatment groups). Comparisons between varied ICI timings are shown in Fig. 3F, H and Supplementary Tables 21–27, and comparisons between varied radioisotopes are shown in Fig. 3G, I and Supplementary Tables 28–34. For each radioisotope delivering 2 Gy mean tumor absorbed dose, early ICI administration led to a statistically significant reduction in tumor growth rate and prolongation of overall survival compared to RPT monotherapy and an untreated control group. ²²⁵Ac-NM600 + ICI groups demonstrated the slowest tumor growth rates for each ICI timing. ²²⁵Ac-NM600 + delayed ICI resulted in a significant extension of overall survival compared to ⁹⁰Y-NM600 + delayed ICI. Supplementary Fig. 8A–D shows the complete response and rechallenge rejection rates for B78 tumor cells injected on the contralateral flank of the complete responders or naïve mice on day 120: ⁹⁰Y-NM600 + early ICI: 2/2; ⁹⁰Y-NM600 + intermediate ICI: 2/2; ¹⁷⁷Lu-NM600 + early ICI: 1/1; ²²⁵Ac-NM600 + intermediate ICI: 1/1; naïve C57BL/6 mice: 0/5.

Fig. 3. Early ICI prolongs overall survival following RPT + ICI in the immune resistant B78 melanoma model.

A Treatment regimen options for in vivo therapy studies. Mice either received early (days −3/0/3), intermediate (days 4/7/10), or delayed (days 11/14/17) dual anti-CTLA4 and anti-PD-L1 relative to RPT on day 1. B–I B78 tumor-bearing mice were randomized to one of 14 treatment groups: untreated control (No Tx), dual anti-CTLA4 and anti-PD-L1 (ICI) on days 4/7/10, or 2 Gy ⁹⁰Y-NM600 (0.7326 MBq), ¹⁷⁷Lu-NM600 (1.702 MBq), or ²²⁵Ac-NM600 (18.5 kBq) on day 1, or combination 2 Gy ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 on day 1 with dual ICI on days −3/0/3 (⁹⁰Y/¹⁷⁷Lu only), days 4/7/10, or days 11/14/17, or ²²⁵Ac-NM600 + ICI on days −8/−5/−2. B–E Individual tumor growth curves corresponding to (F, G). F, H Effects of varied ICI timing for each radionuclide on tumor growth and overall survival. G, I Effects of varied radionuclide for each ICI timing of combination therapy on tumor growth and overall survival. B–I Results of one experiment. B–I N = 9: ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 + ICI 4/7/10 or + ICI 11/14/17; n = 7: ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 + ICI −3/0/3; n = 5: ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600, ICI 4/7/10, No Tx. Linear mixed models were used to compare tumor volumes over time between various treatment groups. Statistical testing of the pairwise contrasts was adjusted for multiple comparisons using Tukey’s method. Log-rank test was used to compare survival. Error bars are SEM.

CD8⁺ T cells are key immune mediators of effector and memory immune response to ⁹⁰Y-, ¹⁷⁷Lu-, and ²²⁵Ac-NM600 + ICI

In prior work with B78 melanoma, we have shown the necessity of T cells for the therapeutic response to ⁹⁰Y-NM600 + anti-CTLA4¹⁵. Separately, in response to immunotherapy, including ICIs, CD8⁺ T cells have been shown to be major mediators of the antitumor and memory response in mice bearing MC38 tumors^35–37. Radiation therapy directed against MC38 tumors has also been shown to enhance the antitumor effect of adoptively transferred CD8⁺ T cells³⁸. In MC38 colorectal carcinoma, the therapeutic benefit of ⁹⁰Y-, ¹⁷⁷Lu-, and ²²⁵Ac-NM600 + ICI as compared to RPT monotherapy was abrogated when CD8⁺ T cells were depleted (Fig. 4A–K and Supplementary Tables 35–41). This suggests the necessity of functional CD8⁺ T cells for mediation of antitumor responses to these combination systemic therapies in mice bearing MC38 tumors.

Fig. 4. CD8⁺ T cells are key immune mediators of effector and memory immune response to ⁹⁰Y-, ¹⁷⁷Lu-, and ²²⁵Ac-NM600 + ICI.

A Confirmation of CD8⁺ depletion by flow cytometry. Gating strategy: Supplementary Fig. 12. B–K MC38 tumor-bearing mice received 2 Gy ⁹⁰Y-, ¹⁷⁷Lu-, or ²²⁵Ac-NM600 + ICI (days −3/0/3) + αCD8 (anti-CD8 on days −5/0/5), RPT + ICI, RPT alone, or αCD8 alone. Effects of CD8⁺ depletion on tumor growth (B–H) and overall survival (I–K) are shown. L Flow cytometry treatment scheme. MC38 tumor-bearing mice were randomized to receive either 2 Gy ⁹⁰Y-, ¹⁷⁷Lu-, or ²²⁵Ac-NM600 + ICI (days −3/0/3), ICI alone, or untreated control (No Tx). M, N Tumors were harvested on days −3, 4, 8, or 11 and dissociated. The tumor immune infiltrate was analyzed by flow cytometry. Gating strategy: Supplementary Fig. 13. O Co-culture experiment scheme. Created https://BioRender.com/22ha1p4. P, Q TNF and IFNγ expression were measured by intracellular flow cytometry staining and analysis of peripheral CD8⁺ effector memory cells (CD8⁺CD44⁺CD62L⁻) harvested from naïve C57BL/6 mice or treated MC38 tumor-bearing mice following 24 h of co-culture with MC38 or no tumor cells. Gating strategy: Supplementary Fig. 14. A–H; M, N; P, Q Results of one experiment. I–K Results of two independent experiments. A N = 15: depleted; n = 7: IgG2b control. B–H N = 7: ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 ± ICI; n = 5: ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 + ICI + αCD8, αCD8. I–K N = 17: ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 ± ICI; n = 10: αCD8; n = 5: ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 + ICI + αCD8. M, N N = 5/treatment group and timepoint except n = 4: ⁹⁰Y-NM600 + ICI Day 4; ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 + ICI Day 11. P, Q N = 5: naïve, 2 Gy ⁹⁰Y-NM600 + ICI; n = 6: ICI, 2 Gy ¹⁷⁷Lu-NM600 + ICI; n = 7: 2 Gy ²²⁵Ac-NM600 + ICI. Unpaired two-tailed t-test was used to compare CD8⁺ frequency between groups. One-way ANOVA with Tukey’s HSD post hoc test was used to compare flow cytometry results between treatment groups for each marker and timepoint and TNF or IFNγ expression between treatment groups within each co-culture condition. Linear mixed models were used to compare tumor volumes over time between various treatment groups. Statistical testing of the pairwise contrasts was adjusted for multiple comparisons using Tukey’s method. Log-rank test was used to compare survival. Error bars are SEM. O Created in BioRender. Jin, W. (2026) https://BioRender.com/22ha1p4.

MC38 tumors were harvested and dissociated for flow cytometry at various timepoints over the course of RPT + early ICI to evaluate longitudinal changes in the tumor immune infiltrate, including pre-treatment (day −3), and one harvest timepoint at the half-life of each radioisotope: day 4 (⁹⁰Y half-life: 2.7 days), day 8 (¹⁷⁷Lu: 6.7 days), and day 11 (²²⁵Ac: 10 days), following 2 Gy RPT administration on day 1 (Fig. 4L–N). Overall, the frequency of CD45⁺ tumor infiltrating immune cells was similar with or without RPT, with a slight reduction in CD45⁺ cell frequency in the ¹⁷⁷Lu- and ²²⁵Ac-NM600 + ICI tumors compared to ICI and ⁹⁰Y-NM600 + ICI on day 11. The ⁹⁰Y-NM600 + ICI treated tumors additionally demonstrated an increased CD8⁺/Treg ratio compared to ICI alone, and increased activation marker expression on CD8⁺ T cells and interferon-γ (IFNγ) production by CD8⁺ and CD4⁺ T cells compared to other groups. These results suggest that ⁹⁰Y-NM600 + ICI treatment leads to an inflamed tumor immune microenvironment in MC38 colorectal carcinoma. Notably, all groups had a reduction in Tregs following ICI on days −3/0/3, as expected following treatment with a Treg depleting anti-CTLA4 antibody (IgG2c, clone 9D9).

We next aimed to determine whether a tumor-specific memory response was elicited following RPT + ICI. To do this, we investigated the functionality of peripheral CD8⁺ effector memory (EM) cells 23 days after 2 Gy mean tumor absorbed dose from ⁹⁰Y-, ¹⁷⁷Lu-, or ²²⁵Ac-NM600 + ICI or ICI alone compared to CD8⁺ EM cells isolated from naïve C57BL/6 mice (Fig. 4O–Q). In this assay, peripheral blood cells were co-cultured with or without MC38 tumor cells for 24 h, and flow cytometry was performed to measure TNF and IFNγ effector cytokine production by CD8⁺ EM cells (CD8⁺CD44⁺CD62L⁻). Overall, ⁹⁰Y-NM600 + ICI resulted in the greatest induction of TNF and IFNγ compared to other treatment groups, and treatment with ¹⁷⁷Lu-NM600 did not induce increased TNF and IFNγ production by CD8⁺ EM cells. There were few or no differences in the level of TNF and IFNγ production by peripheral CD8⁺ EM cells between the groups in the absence of MC38 co-culture, consistent with tumor-specific activation.

Long-term immune memory response following ²²⁵Ac-NM600 dose escalation in immunogenic MC38 tumors

In prior studies, we had observed no improvement in tumor response to ⁹⁰Y-NM600 RPT + ICI combo with dose escalation beyond 2 Gy of ⁹⁰Y-NM600—a low LET, longer range RPT¹⁵. Given that dose escalation of ²²⁵Ac-NM600 to a mean tumor absorbed dose of 8 Gy was tolerable with evidence of hematologic recovery and stable weight in mice bearing MC38 tumors, we began testing ²²⁵Ac-NM600 in combination with ICI in MC38 (Fig. 5A, B). A dose-dependent increase in TNF and IFNγ production by peripheral CD8⁺ EM cells following 0.2, 2, or 8 Gy mean tumor dose from ²²⁵Ac-NM600 + ICI was observed (Fig. 5C).

Fig. 5. Long term immune memory following ²²⁵Ac-NM600 absorbed dose escalation in immunogenic MC38 tumors.

MC38 tumor-bearing mice received IgG2b control on days −5/0/5 (IgG2b), or 0, 0.2 Gy (0.4625 kBq), 2 Gy (4.625 kBq), 8 Gy (18.5 kBq) ²²⁵Ac-NM600 on day 1 ± ICI on days −3/0/3. A Body weights. B Complete blood counts on day 23. WBC white blood cells, LYM lymphocytes, PLT platelets, HGB hemoglobin, RBC red blood cells. C TNF and IFNγ on peripheral CD8⁺ effector memory cells (CD8⁺CD44⁺CD62L⁻) from naïve or MC38 tumor-bearing mice receiving the indicated treatment and following 24 h of co-culture with MC38 or no tumor cells (from Fig. 4O). D–L MC38 tumor-bearing mice received 8 Gy ²²⁵Ac-NM600 + ICI, 8 Gy ²²⁵Ac-NM600, ICI, or were untreated (No Tx). D Experimental scheme. Created with BioRender.com. Tumor growth (E) and overall survival (F) are shown. G Complete response (CR) rates (H) and percent of CR mice and naïve controls rejecting rechallenge with MC38 cells are shown. I–K Flow cytometry of blood from CR (MC38, Ac + ICI) or naïve mice (MC38, naïve). L Flow cytometry of splenocytes from CR ²²⁵Ac-NM600 + ICI mice that did not (Ac + ICI tumor) or did reject rechallenge (Ac + ICI no tumor), CR ICI mice (ICI), naïve C57BL/6 mice following IV injection of MC38 cells (MC38 IV), and completely naïve C57BL/6 mice (Naïve). Gating strategies: Supplementary Figs. 14–16. A, B; C; E–L Results of one experiment. A N = 7: all groups. B N = 5: naïve, 2 Gy ⁹⁰Y-NM600 + ICI; n = 6: ICI, 2 Gy ¹⁷⁷Lu-NM600 + ICI; n = 7: 0.2, 2, 8 Gy ²²⁵Ac-NM600 + ICI. C N = 5: naïve; n = 6: ICI; n = 7: 0.2, 2, 8 Gy ²²⁵Ac-NM600 + ICI. E–G N = 5/treatment group. H N = 5: ²²⁵Ac-NM600 + ICI, naïve; n = 3: ICI. I–K N = 4/treatment group. L N = 4: MC38 IV, Naïve; n = 3: ICI; n = 2: Ac + ICI tumor, Ac + ICI no tumor. One-way ANOVA with Tukey’s HSD post hoc test was used to compare TNF or IFNγ expression, blood counts, CR rates, rechallenge rejection rates, and spleen flow cytometry. Unpaired two-tailed t-test was used to compare blood flow cytometry. Tumor volumes were compared using linear mixed models adjusted for multiple comparisons using Tukey’s method. Log-rank test was used to compare survival. Error bars are SEM. D Created in BioRender. Berg, T. (2026) https://BioRender.com/llbyoho,.

We sought to investigate whether a memory response could be observed following 8 Gy mean tumor dose from ²²⁵Ac-NM600 + ICI. Mice bearing MC38 colorectal carcinoma were treated with ²²⁵Ac-NM600 on day 1 ± ICI on days −3/0 (Fig. 5D–L and Supplementary Tables 42–44). Combination therapy significantly prolonged overall survival compared to RPT monotherapy and control (Fig. 5F). ²²⁵Ac-NM600 + ICI enhanced the CR rate of primary MC38 tumors (5/5 CR) relative to RPT (0/5 CR) and ICI (3/5 CR) monotherapies (Fig. 5G). CRs were rechallenged with injection of MC38 cells in the flank on day 100. However, only 3/5 ²²⁵Ac-NM600 + ICI CR mice rejected tumor rechallenge as compared to 3/3 ICI CR mice (Fig. 5H).

From mice showing rejection of this rechallenge, blood and splenocyte cell populations were evaluated by flow cytometry 24 h following an intravenous injection of MC38 tumor cells on day 136 after RPT. Flow cytometry revealed a significant increase in activation of circulating CD8⁺ memory T cells by IFNγ and TNF expression compared to naïve mice following MC38 re-exposure for all ²²⁵Ac-NM600 + ICI mice regardless of rechallenge response (Fig. 5I). No CD4⁺ memory cell activation differences were observed relative to age-matched naïve mice (Fig. 5J), suggesting the memory response may be mediated by CD8⁺ cells. Increased CD8⁺ CM/EM cell ratios have been reported to be associated with long-term immunologic memory³⁹. Circulating lymphocyte percentages were maintained in ²²⁵Ac-NM600 + ICI treated mice (Fig. 5K). Additionally, we noted increased CD4⁺ and CD8⁺ central memory (CM)/EM ratios in spleens of all ²²⁵Ac-NM600 + ICI CR mice as compared to ICI monotherapy CR mice (Fig. 5L), regardless of whether the CR mice rejected rechallenge (Ac + ICI no tumor) or not (Ac + ICI tumor). Thus, long-term memory immune memory responses were observed following ²²⁵Ac-NM600 dose escalation in the immune sensitive MC38 tumor model.

²²⁵Ac-NM600 + ICI showed therapeutic benefit over ⁹⁰Y-NM600 + ICI and did not decrease immune cell populations in immune-resistant tumors

To test our hypothesis that the immunogenicity of the tumor model may impact the optimal radioisotope for antitumor responses, we evaluated the therapeutic efficacy of low absorbed dose ²²⁵Ac-NM600 vs. ⁹⁰Y-NM600 + dual anti-CTLA4 and anti-PD-L1 in a second poorly immunogenic model. The Myc-CaP prostate cancer model was selected, as it has an immunosuppressive tumor microenvironment and limited responses to RPT or ICI alone⁴⁰. Tumor uptake and dosimetry of ⁸⁶Y/⁹⁰Y-NM600²⁰ and ²²⁵Ac-NM600⁴¹ in mice bearing Myc-CaP prostate cancer tumors were reported previously. As was observed in B78 melanoma, Mice bearing Myc-CaP tumors showed a statistically significant reduction in tumor growth rate and prolonged overall survival when treated with 2 Gy ²²⁵Ac-NM600 + ICI as compared to 2 Gy ⁹⁰Y-NM600 + ICI or monotherapies (Fig. 6A–C and Supplementary Tables 45–47). This finding supported the hypothesis that tumor immunogenicity may impact therapeutic response to ²²⁵Ac alpha-particle emitting vs. ⁹⁰Y or ¹⁷⁷Lu beta-particle emitting RPT when combined with ICI therapy.

Fig. 6. ²²⁵Ac-NM600 + ICI showed therapeutic benefit over ⁹⁰Y-NM600 + ICI and did not decrease immune cell populations in immune-resistant tumors.

A–C Myc-CaP tumor-bearing mice received the indicated treatment. A Experimental scheme. Tumor growth (B) and overall survival (C) are shown. D–H Single cell RNA sequencing of B78 tumors. D Cell type fractions in lymphoid compartment. Cell types having >5% fraction in at least one sample are represented. E Single cell RNA sequencing treatment scheme. B78 tumor-bearing mice received either 2 Gy ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 or EBRT + ICI (anti-CTLA4 and anti-PD-L1) on days −3/0/3 or ICI alone. Tumors were harvested on day 15, dissociated, and immune cells were isolated by Ficoll density gradient centrifugation for single cell RNA sequencing analysis. Created in BioRender. Jin, W. (2026) https://BioRender.com/22ha1p. F UMAP of single cell RNA sequencing data colored by samples. Each panel shows identical cell population with one sample in color and all the others in gray. Regions containing T cells, myeloid cells, B cells, and epithelial cells/fibroblasts are indicated. G Representative B78 tumor photomicrographs for B cell marker CD20 at day 15 following 2 Gy ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 or 2 Gy EBRT + ICI or ICI alone; brown = positive immunolabeling. H Quantification of CD20⁺ cell clusters/tumor. Cell cluster: ≥5 cells. Scale bar = 200 μm. B–H Results of one experiment. B, C n = 10/treatment group. D–F 3 mice/treatment group pooled into one sample/treatment group for single cell RNA sequencing analysis. Numbers of cells analyzed provided in Supplementary Table 48. ⁹⁰Y-NM600 + ICI, n = 6067; ¹⁷⁷Lu-NM600 + ICI, n = 9808; ²²⁵Ac-NM600 + ICI, n = 6503; EBRT + ICI, n = 7045; ICI, n = 11,249. D–F n = 3 tumors/treatment group. G Representative image/treatment group from (H). H n = 4 tumors: ¹⁷⁷Lu-NM600 + ICI; n = 3 tumors: ⁹⁰Y-NM600 + ICI, ²²⁵Ac-NM600 + ICI, EBRT + ICI, ICI. Linear mixed models were used to compare tumor volumes over time between various treatment groups. Statistical testing of the pairwise contrasts was adjusted for multiple comparisons using Tukey’s method. Log-rank test was used to compare survival. One-way ANOVA with Tukey’s HSD post hoc test was used to compare CD20 IHC quantification between treatment groups. Error bars are SEM. E Created in BioRender. Jin, W. (2026) https://BioRender.com/22ha1p4.

In the immune resistant B78 melanoma, we sought to identify tumor immune microenvironment changes following ⁹⁰Y-, ¹⁷⁷Lu-, or ²²⁵Ac-NM600 + ICI combination therapy that would explicate the relative therapeutic benefit of alpha-emitting low absorbed dose ²²⁵Ac-NM600 compared to low absorbed dose ⁹⁰Y-NM600 observed in Fig. 3. We performed paired single cell (sc)RNA-sequencing (seq) and T cell receptor (TCR) seq among tumor-infiltrating immune cells from mice bearing B78 melanoma tumors receiving 2 Gy mean tumor absorbed dose from either ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600, or EBRT + early ICI or ICI alone. Tumors were harvested, and 3 tumors/treatment group were pooled for CD45⁺ cell isolation and scRNA-seq and TCR seq on treatment day 15 (Fig. 6D–F). Supplementary Fig. 9A, B shows the tumor growth curves of the B78 melanoma tumors harvested and pooled for scRNA-seq analysis. Following filtering and quality control, data were analyzed for 40,672 cells with 11,312 mean reads per cell. The number of cells (independent biological replicates) used for scRNA-seq analysis per treatment group is indicated in Supplementary Table 48. Dimensionality reduction and graph-based clustering of all cells were performed to distinguish cells with unique transcriptional profiles. UMAPs showing identical cell populations (overall cell population analyzed) with each treatment group highlighted are shown in Fig. 6F.

Lymphoid cell type fractions are shown in Fig. 6D. Of note, the frequency of follicular B cells, which participate in T-cell-dependent antibody responses, was increased in the ²²⁵Ac-NM600 + ICI and ICI alone treatment groups. The ²²⁵Ac-NM600 follicular B cells had a 2.57-fold increase compared to ICI alone in expression of the MHC-I gene, H2-K1. Frequencies of both CD8⁺ effector T cells and Tregs were highest in the ⁹⁰Y-NM600 + ICI treatment group. CD8⁺/Treg ratios for each treatment group were similar between groups in this analysis: ²²⁵Ac-NM600 + ICI: 0.44, ¹⁷⁷Lu-NM600 + ICI: 0.37, ⁹⁰Y-NM600 + ICI: 0.49, EBRT + ICI: 0.44, ICI: 0.58. With portions of the same tumor samples that were used for single cell RNA sequencing, CD4⁺, CD8⁺, Tregs (FOXP3⁺ cells), and NK1.1⁺ cells were quantified by IHC (Supplementary Fig. 9). Consistent with the single cell RNA sequencing results, there were no statistically significant differences in these lymphoid cell populations between treatment groups (Fig. 6D and Supplementary Fig. 10A–D).

External beam radiation to tumor-draining lymph nodes (TDLN) decreases the effectiveness of RT + anti-CD25 immunotherapy¹³. In metastatic settings, TDLNs may not be readily identifiable or avoidable. Furthermore, it is unknown what effect EBRT may have on other immune structures, including tertiary lymphoid structures (TLS), which have been shown to be critical to antitumor immunity with ICIs⁴². We hypothesize that RPT, by molecularly rather than spatially targeting tumors, could avoid such structures. EBRT + ICI disrupted existing CD20⁺-rich lymphoid clusters in B78 melanoma (Fig. 6G, H). These CD20⁺ structures were present in a significantly greater number in the ²²⁵Ac-NM600 + ICI tumor, which is the radioisotope with the shortest emission pathlength (<100 μm)²³. The short tissue pathlength signifies that irradiation is minimized to non-targeted cells in the tumor, including stroma and islands of immune cells, as well as neighboring structures, including lymph nodes.

²²⁵Ac-NM600 + ICI leads to diversified effector T cell responses with key shared clonotypes

With the same cells used for scRNA-seq analysis, we performed TCR seq analysis to gain insight into the functional status of CD8⁺ T cells following ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600, or EBRT + ICI. After quality control, TCR sequencing analysis provided data for 3,517 cells with productive V-J spanning pairs. The number of cells (independent biological replicates) used for TCR sequencing analysis per treatment group is shown in Supplementary Table 49. UMAPs of all T cells stratified by sample are shown in Fig. 7A. Clonotype expansion ranges for Fig. 7 are represented in Fig. 7B. UMAPs stratified by T cell subtype are shown in Fig. 7C, verifying that most medium and large T cell clones pertain to CD8⁺ effector and memory subtypes. Clonal expansion by T cell subtype and CDR3 sequence diversity measurement (D50) are shown in Fig. 7D, E. ¹⁷⁷Lu- and ⁹⁰Y-NM600 + ICI have more clonal expansion of CD8⁺ effector and memory T cells, with ²²⁵Ac-NM600 and EBRT + ICI showing more diversity and a higher D50 score for CD8⁺ effector T cells. Figure 7F presents the clonal expansion data in a different way, demonstrating the % of cells with clonotypes per expansion. These results indicate that the majority of effector CD8⁺ clones in the ¹⁷⁷Lu- and ⁹⁰Y-NM600 + ICI groups are medium or large clones. These findings are consistent with previously reported CD8⁺ clonal expansion following low absorbed dose ⁹⁰Y-NM600 + ICI¹⁵. Finally, shared clonotypes between various subsets of T cells are shown in Fig. 7G, H, with ²²⁵Ac-NM600 + ICI treated tumors having the greatest number of shared clonotypes between effector and memory CD8⁺ T cells (11), as compared to seven shared clonotypes for the same T cell subsets in each other treatment group.

Fig. 7. ²²⁵Ac-NM600 + ICI leads to diversified effector T cell responses with key shared clonotypes.

Paired TCR sequencing analysis performed on samples from B78 tumor-bearing mice used for single cell RNA sequencing. A–H Data presented are from samples from experiment depicted in Fig. 6E. A UMAP plots of T cells stratified by sample. B TCR clonotype expansion ranges. C UMAP plots of T cells stratified by T cell subtype. D Clonal expansion by treatment group and T cell subtype. E D50 score by treatment group and T cell subtype. F Relative abundance of clonal indices for each treatment group and T cell subtype. G UpSet plot showing shared clonotypes between T cell subtypes for the ²²⁵Ac-NM600 + ICI sample. Red box demarcates CD8⁺ effector and memory subsets; red arrow demarcates # of shared clonotypes between CD8⁺ effector and memory subsets (=11). H UpSet plots showing shared clonotypes between T cell subtypes for the ⁹⁰Y-, ¹⁷⁷Lu-, EBRT + ICI, and ICI alone samples. For each plot, red arrow demarcates # of shared clonotypes between CD8+ effector and memory subsets (=7 for each sample). A–H Results of one experiment. 3 mice/treatment group pooled into one sample/treatment group for TCR sequencing analysis. Numbers of cells analyzed provided in Supplementary Table 49. Number of cells analyzed: ⁹⁰Y-NM600 + ICI, n = 408; ¹⁷⁷Lu-NM600 + ICI, n = 993; ²²⁵Ac-NM600 + ICI, n = 854; EBRT + ICI, n = 285; ICI, n = 977.

While ²²⁵Ac-NM600 + ICI treated tumors had greater T cell diversity overall, they exhibited shared clonotypes between effector and memory CD8⁺ T cells. This is consistent with an in situ vaccine effect and greater priming of an adaptive T cell response with high LET α-particle RPT, as compared to the greater clonal expansion or propagation of existing immune response in the absence of apparent T cell priming with low absorbed dose, low LET β-particle RPT.

²²⁵Ac-NM600 + ICI enhances CD8⁺ cell expression of interferon-α and γ response genes compared to ⁹⁰Y-/¹⁷⁷Lu-NM600/EBRT + ICI

Pathway analysis and differential gene expression analysis were performed to evaluate the relative expression of genes in each combination RT + ICI treatment vs. ICI alone. These results for CD8⁺ effector T cells and M1 macrophages for genes involved in IFNγ and IFNα responses are shown in Fig. 8A–D and Supplementary Fig. 11A–C. In the ²²⁵Ac-NM600 + ICI treatment group, both the hallmark interferon alpha response and hallmark interferon gamma response pathways were noted to be altered compared to the other treatment groups (CD8⁺ cells: Fig. 8A, M1 macrophages: Supplementary Fig. 11A). These pathways were selected for further differential gene expression analysis (Fig. 8B, C). Following ²²⁵Ac-NM600 + ICI, several CD8⁺ effector T cell genes were upregulated to a greater degree than other RT + ICI groups: Rnf213, Txnip, Isg15, Bst2, Rtp4, Epsti1, Zbp1, Stat1, Irf7, and Ly6e. Of these, Rnf213, Isg15, Rpt4, and Stat1 are interferon-induced genes; IRF7 regulates several interferon α genes; and ZBP1 activates both IFN1 responses and NF-κB signaling⁴³. This IFN1 activation reflects the reported IFN1 signaling following radiation due to the accumulation of cytosolic micronuclei⁴⁴. TXNIP has been reported to be associated with increased tumor-infiltrating immune cells and a carcinostatic role overall⁴⁵. Epithelial-stromal interaction 1 (EPST1) is a protein that promotes monocyte adhesion to endothelial cells⁴⁶. BST2 and LY6E are pro-tumor proteins. B2m, which encodes β₂ microglobulin, a component of MHC-I molecules, was upregulated for all groups relative to ICI alone. This increase in expression for all RT + ICI groups reflects the well-documented ability of RT to enhance MHC-I expression⁴⁷.

Fig. 8. ²²⁵Ac-NM600 + ICI enhances CD8⁺ cell expression of interferon-α and γ response genes compared to ⁹⁰Y-/¹⁷⁷Lu-NM600/EBRT + ICI.

A–D Data presented are from samples from experiment depicted in Fig. 6E. A Pathway enrichment results for CD8⁺ effector T cells. The bluer the pathway, the lower the adjusted P, and thus the more significantly the pathway is altered. Differential gene expression heat map for interferon γ (B) and interferon α (C) response genes for CD8⁺ effector T cells. The color of each box represents the log₂(fold change) of gene expression of ²²⁵Ac-, ¹⁷⁷Lu-, ⁹⁰Y-NM600, or EBRT + ICI compared to ICI alone. D Incidence matrix of differential gene expression between treatment groups of the following: comparing cells where the shared clonotypes between effector and memory CD8⁺ T cells are between 7 and 11 (as shown in Fig. 7G, H), and they are amplified. Red arrow marks interferon-inducible gene, Ifi27l2a, which is upregulated in these shared clonotype CD8⁺ cell populations treated with ²²⁵Ac-NM600 + ICI over each treatment group. A–D Results of one experiment. 3 mice/treatment group pooled into one sample/treatment group for single cell RNA sequencing TCR sequencing analysis. Benjamini–Hochberg method was used to correct for multiple comparison. Numbers of cells analyzed provided in Supplementary Table 48.A–C ⁹⁰Y-NM600 + ICI, n = 6067; ¹⁷⁷Lu-NM600 + ICI, n = 9808; ²²⁵Ac-NM600 + ICI, n = 6503; EBRT + ICI, n = 7045; ICI, n = 11,249 and Supplementary Table 50.D ⁹⁰Y-NM600 + ICI, n = 123; ¹⁷⁷Lu-NM600 + ICI, n = 151; ²²⁵Ac-NM600 + ICI, n = 45; EBRT + ICI, n = 36; ICI, n = 100.

For M1 macrophages (Supplementary Fig. 11B, C), there were several genes upregulated following ²²⁵Ac-NM600 + ICI but downregulated for other groups, including Ly6e, Irf7, Isg15, Bst2, and unique from CD8+ effector T cells, Ccl7 and Ifitm3. Ccl7 has been shown to facilitate antitumor responses to anti-PD-1 through recruitment of classical dendritic cells⁴⁸, and Ifitm3 encodes an interferon-inducible transmembrane protein. Conversely, both the MHC-II gene (H2-Aa) and Cd74, involved in the formation and transport of MHC-II⁴⁹, and were downregulated following ²²⁵Ac-NM600 + ICI in M1 macrophages but upregulated in all other RT + ICI groups (Supplementary Fig. 11B, C). Taken together, genes induced by ²²⁵Ac-NM600 + ICI have mixed roles in tumor progression and immune response, with a clear prevalence of MHC-I expression among all RT groups, and IFN1-inducible genes following ²²⁵Ac-NM600 + ICI, consistent with the greater amount of DNA damage induced by the α-emitting radioisotope.

Finally, we performed differential gene expression analysis on CD8⁺ cells with shared clonotypes within effector and memory cells (the population of cells highlighted in Fig. 7G, H). The number of cells (independent biological replicates) with clonotypes shared within effector and memory CD8⁺ T cells, stratified by treatment group, is shown in Supplementary Table 50. Interestingly, one gene that was upregulated in this cell population in the ²²⁵Ac-NM600 + ICI treatment group over each other treatment group was an interferon inducible gene, Ifi27l2a (red arrow, Fig. 8D). This observation further suggested that an active adaptive T cell response was present in the ²²⁵Ac-NM600 + ICI-treated tumor.

Discussion

Low absorbed dose ⁹⁰Y-NM600 + ICI was more effective than tumor absorbed dose-matched ²²⁵Ac-NM600 + ICI in the immunogenic MC38 colorectal carcinoma model. This observation was surprising, given that greater STING activation has been observed with ²²⁵Ac²³. The MC38 response to RPT alone (Fig. 2E) was not affected by difference in RBE, suggesting that the therapeutic response in MC38 may be predominantly mediated by tumor-resident immune cells following low absorbed dose RPT. The higher frequency of tumor infiltrating CD45⁺ cells in the ⁹⁰Y-NM600 + ICI group on day 11 may contribute to the superior tumor responses in that group (Fig. 4M). Additionally, circulating CD8⁺ effector memory cells isolated after 2 Gy mean tumor absorbed dose from ⁹⁰Y-NM600 + ICI treated mice had higher effector cytokine production than those isolated after the same mean tumor absorbed dose from ²²⁵Ac-NM600 + ICI when co-cultured with MC38 tumor cells. This suggests that ⁹⁰Y-NM600 + ICI is effective at enhancing the existing effector immune response in immune sensitive MC38 (Fig. 4P, Q).

The combination of enhanced clonal expansion and preservation of existing tumor immune response following ⁹⁰Y-NM600 + ICI may be particularly beneficial in patients with immune-sensitive metastatic cancers. These tumors, like the MC38 model, have an existing degree of immunogenicity (increased tumor mutational burden, increased immune cell infiltration at baseline, increased MHC-I expression, or a combination of these) and may have developed acquired resistance to ICIs. However, response to ICI may be rescued in such settings following RPT + ICI therapy where low absorbed dose levels permit immune cell sparing (Fig. 6G, H), enhanced immune cell infiltration of tumor (Fig. 4M, N), increased peripheral immune cell activation (Figs. 4P, Q and 5C), and bolstered immunological memory (Fig. 5I–L).

Conversely, ²²⁵Ac-NM600 + ICI was more effective than ⁹⁰Y-NM600 + ICI and ¹⁷⁷Lu-NM600 + ICI in the poorly immunogenic B78 melanoma and Myc-CaP prostate cancer models. There are several potential contributing factors to this result. ²²⁵Ac-NM600 is a high LET radioisotope with a greater RBE in the B78 tumor model as a monotherapy (Fig. 3E). Additionally, irradiation of B78 melanoma in vitro with low absorbed dose ²²⁵Ac was shown to increase expression of type I interferon response-associated genes, which was not observed following ⁹⁰Y or ¹⁷⁷Lu at the same low absorbed dose²³. Additionally, single cell RNA sequencing and TCR sequencing of B78 tumors following ²²⁵Ac-NM600+early ICI compared to ⁹⁰Y- and ¹⁷⁷Lu-NM600+early ICI showed that ²²⁵Ac-NM600 + ICI treated tumors had fewer infiltrating Tregs and greater infiltrating follicular B cells (Fig. 6D). IFNγ and IFNα response genes were upregulated in the ²²⁵Ac-NM600 + ICI group specifically (Fig. 8).

Additionally, the short tissue range of α-emitting ²²⁵Ac-NM600 may result in greater absorbed dose heterogeneity and sparing of B cell-rich TLSs compared to β-emitting radioisotopes or EBRT⁵⁰ (Fig. 6G, H). Radiation absorbed dose heterogeneity, TILs, nearby TDLNs, and TLSs have all been implicated in anti-tumor immune responses following RT + ICI^{13,42,50–52}. Low-dose radiotherapy and anti-PD-1 were shown to increase quantity and maturity of TLSs in murine lung cancer, with a strong antitumor effect associated with CD8⁺ number within TLSs⁵². RPT has the potential to maintain and possibly promote TILs, TDLNs, and TLSs by molecularly targeting tumor cells rather than spatially targeting tumors, as with EBRT.

A unique combination of an effector CD8⁺ T cell repertoire with greater diversity overall but increased clonal expansion of shared clonotypes between effector and memory CD8⁺ T cells was noted for the ²²⁵Ac-NM600 + ICI group compared to the ⁹⁰Y- and ¹⁷⁷Lu-NM600 + ICI groups (Fig. 7). In a report by Fairfax et al., large CD8⁺ T cell clone counts 21 days following treatment, regardless of clonal specificity, were associated with durable responses to combination (anti-CTLA4 and anti-PD-1) ICI in metastatic melanoma patients⁵³. The T cell populations analyzed in this study could signal a progressing adaptive immune response. These findings are consistent with ²²⁵Ac-NM600 + ICI leading to an enhanced systemic antitumor immune response in the immune resistant B78 melanoma model. High LET ²²⁵Ac-NM600 may have a role of priming antitumor immune responses through recruitment of CD8⁺ T cells through type I IFN signaling, whereas low LET may have a larger role in propagating existing immune responses in immune sensitive tumors. All studies here were performed with low absorbed dose RPT. Therefore, it is unknown if low-LET RPT may also prime immune resistant tumors to respond to ICI if administered at higher absorbed doses or in a more heterogeneous distribution. Further investigation is warranted with additional RPT agents and tumor models.

Early or intermediate-timed dual anti-CTLA4 and anti-PD-L1 in combination with RPT significantly prolongs overall survival compared to groups receiving delayed dual ICI in both MC38 and B78 (Figs. 2 and 3). Recently, trials have highlighted the importance of timing of RT + ICI to achieve anti-tumor responses^54,55 (NCT03519971). In our study, administering a Treg depleting isotype of anti-CTLA4 (Treg depletion shown in Fig. 4N) before RPT in the early ICI groups may have had an immune priming effect or effect in depleting Tregs⁵⁶. Depleting this immune suppressive population may have aided activated CD8⁺ T cells (Figs. 4P, Q and 8). Additionally, as is shown in Fig. 6, RPT preserved CD20⁺ tumor immune structures, potentially sparing antigen presentation processes within the tumor and helping to promote clonal expansion of a shared lineage of CD8⁺ effector and memory cells (Fig. 7G). Future studies delineating interactions of other immune cells with RPTs will provide better mechanistic insight to further optimize its combination with ICIs.

This study had several limitations. To perform exploratory investigations of tumor immunomodulation with novel RPT agents, absorbed doses, and ICI treatment schedules, syngeneic murine tumor models are a valuable tool. However, murine tumor models do not fully recapitulate the immune microenvironment and metastatic potential of human tumors and disease progression. Based on prior work with companion canines and human cell lines, we anticipate that the image-based dosimetry methods employed in this study will enable safe translation of RPT dosing studied here to larger tumors, including human tumors^15,20,57; however, the effects of specific isotopes may vary as a function of tumor size, given the fixed physical range of emitted radiation from these. Timing of administration of RPT and ICI agents warrants further investigation as factors including tumor size and host immune fitness change quickly in the short life cycle of a mouse. However, efforts were made to control the influence of tumor size in the therapeutic studies, including permuted block randomization by tumor size and separate ICI control treatment groups for each administration timing (Fig. 2G–I). In this work, we compared equal absorbed doses of radiation delivered to a tumor with the same vector (NM600). However, changing the isotope could alter the biodistribution of the agent, resulting in differing absorbed doses for normal tissues, which potentially confounds the immune response. ²²⁵Ac-NM600, for example, had higher liver uptake and lower tumor:liver uptake ratios than ⁹⁰Y- or ¹⁷⁷Lu-NM600 (²²⁵Ac-NM600 tumor:liver absorbed dose ratio: 0.23; ¹⁷⁷Lu-NM600: 1.37; ⁹⁰Y-NM600: 0.66). Given this variation, it is important to use caution when selecting surrogate imaging radioisotopes for dosimetry, as the radioisotope itself may greatly impact the biodistribution and pharmacokinetics of a given RPT agent. Unfortunately, there may not be a better method for comparison of tumor immune effects than using equivalent tumor absorbed doses and the same RPT chelator and vector.

Additionally, we acknowledge that many features of syngeneic tumor lines aside from immunogenicity, including differential penetration, uptake, and resulting absorbed doses of treatments, impact responses to any given treatment. These potential confounding factors were investigated in the following ways: (1) image (⁹⁰Y, ¹⁷⁷Lu) or biodistribution (²²⁵Ac)-based dosimetric methods were used to estimate injected activities to achieve the same average tumor radiation absorbed doses across tumor models and RPTs (Table 1); (2) iQID scans of NM600 compounds in MC38 and B78 tumors confirmed no major differences in NM600 uptake across tumor types (Fig. 1G–K); (3) immunohistochemistry of anti-PD-L1 antibody penetration in MC38 and B78 tumors showed a similar distribution of anti-PD-L1 antibody uptake across the two tumor types (Supplementary Fig. 5); (4) mRNA expression of checkpoint molecules on B78 and MC38 was confirmed by qPCR to not significantly change following treatment with 2 Gy ⁹⁰Y-, ¹⁷⁷Lu-, or ²²⁵Ac-NM600 (Supplementary Fig. 6). Following these analyses, immunologic differences between MC38 and B78 and Myc-CaP remain a key tumor feature that may relate to the therapeutic divergence observed in response to 2 Gy absorbed tumor dose from ²²⁵Ac alpha-particle emitting vs. ⁹⁰Y or ¹⁷⁷Lu beta-particle emitting RPT when combined with ICI therapy. Certain aspects of tumor immunogenicity, including mismatch repair status, directly impact radiation response and the capacity of cells to repair DNA damage, making the mechanism of therapeutic response to combination RPT + ICI therapy difficult to disentangle.

In poorly immunogenic tumors with intrinsic resistance to ICIs, combination with high LET RPT may prime adaptive CD8⁺ T cell recognition. Conversely, in immunogenic tumors with acquired resistance to ICIs, low absorbed doses of low LET RPT may provide tumor inflammation and tumor cell susceptibility needed to renew clonal expansion of existing tumor-specific CD8⁺ T cells and generate immunologic memory. This study emphasizes the critical importance of understanding the mechanisms of ICI resistance in order to appropriately implement combined modality treatment approaches that may overcome such resistance. In this work, we begin to clarify differential immune effects of distinct radioisotopes when delivered by the same RPT vector. A mechanistic understanding of these effects will be vital to the rational design and optimization of clinical studies combining RPT and ICIs.

Methods

Cell lines and culture

The murine melanoma B78-D14 (B78) cell line, derived from B16 melanoma, was obtained from Dr. Ralph Reisfeld (Scripps Research Institute) in 2002⁵⁸. The murine colorectal carcinoma MC38 cell line, isolated from a female C57BL/6 mouse, was obtained from Millipore Sigma (Cat # SCC172), and the murine prostate cancer Myc-CaP cell line, isolated from a male FVB mouse, was obtained from American Type Culture Collection (ATCC) (Cat # CRL-3255). B78 cells were grown in a humidified incubator at 37 °C with 5% CO₂, in RPMI-1640 (Corning, Cat # 10-040-CV) supplemented with 10% FBS (Avantor, Cat # 97068-085), and 100 U/mL penicillin + 100 µg/mL streptomycin (Corning, Cat # 30-002-CI); MC38 and Myc-CaP cells were grown in DMEM (Corning, Cat # 10-013-CV) supplemented with 10% FBS, 100 U/mL penicillin + 100 µg/mL streptomycin. Cell line authentication was performed per ATCC guidelines using morphology, growth curves, and Mycoplasma testing within 6 months of use.

Murine tumor models

Mice were housed in a specific pathogen-free animal facility and treated under a protocol approved by the Institutional Animal Care and Use Committee at the University of Wisconsin-Madison (protocol: M005670). Control and experimental mice were co-housed. All mice were monitored daily by independent veterinary staff and weighed weekly. All mice were maintained under a tightly controlled temperature (22 °C), humidity (40–50%), light/dark (12/12 h) cycle conditions, with water and food ad libitum. C57BL/6NTac (C57BL/6 genetic background, stock number F151) and FVB/NTac (FVB genetic background, stock number F59) mice were purchased at age 6–8 weeks from Taconic. Both male and female C57BL/6 mice were included in MC38 and B78 therapy studies. Female mice were used for correlate studies. Male FVB/N mice only were included in Myc-CaP prostate cancer therapy studies as the prostate gland is present in male mammals only.

Mouse therapy studies

MC38 and Myc-CaP tumors were engrafted by subcutaneous flank injection of 1 × 10⁶ tumor cells. B78 tumors were engrafted by subcutaneous flank injection of 2 × 10⁶ tumor cells. Tumor size was determined using Vernier calipers, and volume approximated as (width² × length)/2. Mice were randomized immediately before treatment by permuted block randomization. Only mice with palpable flank tumors the day before treatment began were included in the study. Treatment began when tumors were well-established (~50–100 mm³), approximately 12 days after tumor implantation for MC38, 17 days for Myc-CaP, and 4 weeks for B78. The day of EBRT or RPT was defined as “day 1” of treatment. Anti-murine CTLA-4 (IgG2c, clone 9D9, produced by NeoClone) and anti-murine PD-L1 (IgG2b, clone 10F.9G2, BioXCell, Cat # BE0101) were administered by 100 μg intraperitoneal injection on days indicated in figures and figure legends. Tumors were measured twice weekly for at least 60 days after starting treatment unless mice died or were euthanized. Mice were euthanized via CO₂ asphyxiation when tumor size exceeded 20 mm in the longest dimension or recommended by an independent animal health monitor for morbidity or moribund behavior. Due to institutional radiation safety protocols and the mice being treated with radioactive isotopes, all investigators were aware of mouse treatment groups. The number of mice in therapy groups was based on anticipated effect size of outcomes to power experiments, determined in collaboration with biostatisticians, and therefore may vary between treatment groups and tumor models. Mouse therapy experiments were repeated in duplicate (two independent experiments). Final replicates are presented for tumor response and aggregate data for survival; number of animals (independent biological replicates) per group is indicated in figure legends.

Mouse CD8⁺ depletion studies

CD8⁺ T cell depletion was performed by injection of anti-murine CD8α (Rat IgG2b, κ; Clone 2.43; BioXCell, Cat # BE0061) or IgG2b control (Rat IgG2b, κ; Clone LTF-2; BioXCell, Cat # BE0090) by 300 µg intraperitoneal injection on days −5, 0, and 5. Depletion was confirmed on treatment day −3 by flow cytometry.

Mouse immune memory studies

Mice without palpable tumors on day 100 were considered complete responders (CR). CR mice were rechallenged with 1 × 10⁶ MC38 cells or 2 × 10⁶ B78 cells (whichever cell line matched the original tumor), contralateral flank injection on day 100. For each therapy study, n = 5 age-matched naïve female C57BL/6 mice (independent biological replicates) were injected with either 1 × 10⁶ MC38 cells or 2 × 10⁶ B78 cells as a control. On day 136, CR mice and age-matched naïve controls (n = 4 independent biological replicates) received 2 × 10⁵ MC38 cells IV. An additional n = 4 naïve controls (independent biological replicates) received no cells IV. Twenty-four hours post antigen re-exposure, blood and splenocytes were harvested for flow cytometry.

Radionuclides and radiochemistry

⁹⁰Y was purchased as ⁹⁰YCl₃ from Eckert and Ziegler. ¹⁷⁷Lu and ²²⁵Ac were purchased as ¹⁷⁷LuCl₃ and solid ²²⁵Ac(NO₃)₃ from Oak Ridge National Laboratory. The radiolabeling of 2-[[hydroxy[[18-[4-[[2-[4,7,10-tris(carboxymethyl)-1,4,7,10-tetraazacyclododec-1-yl]acetyl]amino]phenyl]octadecyl]oxy]phosphinyl]oxy]-N,N,N-trimethyl ethanaminium (NM600) with ⁹⁰Y, ¹⁷⁷Lu, or ²²⁵Ac proceeded by mixing the radiometal with NM600 (0.336–3.36 MBq/nmol) in 0.1 M NaOAc buffer (pH 5.5) for 30 min at 90 °C for ⁹⁰Y-NM600²². The labeled compounds were purified using a reverse-phase Waters Oasis HLB Light cartridge (Milford, MA), eluted using 100% ethanol, dried with a stream of N₂, and reconstituted in 0.9% NaCl with 0.1% v/v Tween 20. Yield and purity were determined with instant thin-layer chromatography using 50 mM EDTA and iTLC-SG glass microfiber chromatography paper impregnated with silica gel (Agilent Technologies). A cyclone phosphor image reader was used to analyze the chromatograms; the labeled compound remained at the spotting point while free radiometals moved with the solvent front.

PET/CT imaging

Mice bearing MC38 (n = 3 male, n = 3 female independent biological replicates) flank tumors (100–150 mm³) were injected via tail vein with 9.25 MBq of ⁸⁶Y-NM600 and imaged with an Inveon microPET/microCT scanner (Siemens Medical Solutions, Knoxville, TN) at 3, 24, 48, and 72 h post injection of the radiotracer. For each scan, mice were anesthetized with isoflurane and placed in the prone position on the scanner bed. Sequential CT (80 kVp; 1000 mAs; 220 angles) and static PET scans (80 million coincidence events; time window: 3.432 ns; energy window: 350–650 keV) were collected. A three-dimensional ordered subset expectation maximization algorithm was used to reconstruct the PET images. These were then fused with corresponding CT images for attenuation correction and anatomical referencing. Tumor and organs of interest were contoured for region-of-interest analysis of the PET images to determine the magnitude and kinetics of ⁸⁶Y-NM600 uptake, which is reported as percent injected activity per gram of tissue (%IA/g; mean ± SEM). An ex vivo biodistribution study was carried out after the last scan time point.

SPECT/CT imaging

Animals with subcutaneous MC38 (C57BL/6) tumors were administered 18.5 MBq of ¹⁷⁷Lu-NM600 in the lateral tail vein. Individual mice (n = 4 independent biological replicates) were placed prone into a MILabs U-SPECT6 /CTUhr system (Houten, The Netherlands) under 2% isoflurane for longitudinal scans at 3, 24, 96, and 180 h post-injection. CT scans (5 min) were acquired for anatomical reference and attenuation correction and fused with the SPECT scans (45 min). Image reconstruction used a similarity-regulated ordered-subset expectation maximization (SROSEM) algorithm. Images were quantitatively analyzed by drawing volumes of interest (VOI) over the tumor and organs of interest to determine the percent injected activity (IA) per gram (%IA/g) for each tissue. An ex vivo biodistribution study was carried out after the last scan time point.

Ex vivo biodistribution

Mice bearing MC38 subcutaneous tumors were injected with 9.25 MBq ⁸⁶Y-NM600, 18.5 MBq ¹⁷⁷Lu-NM600, or 9.25 kBq ²²⁵Ac-NM600, then euthanized via CO₂ asphyxiation at 72 h (⁸⁶Y-NM600), 180 h (¹⁷⁷Lu-NM600), or 2, 48, and 168 h (n = 3 mice (independent biological replicates)/timepoint) post-injection, and organs of interest were collected for ex vivo biodistribution for dosimetry. For comparison between low A_m and high A_m ²²⁵Ac-NM600 biodistributions, MC38 tumor bearing mice were injected with 9.25 kBq of low A_m (81.1 μg NM600/MBq ²²⁵Ac; n = 3 independent biological replicates) or high A_m (2.70 μg NM600/MBq ²²⁵Ac; n = 2 independent biological replicates) ²²⁵Ac-NM600, and organs of interest were collected 12 days after injection. Organs from animals that received ²²⁵Ac-NM600 were allowed to decay at 4 °C overnight to reach secular equilibrium with ²¹³Bi, which could be quantified. Organs were wet weighed, analyzed with a Perkin Elmer Wizard2 Gamma Counter (Westham, MA), and decay corrected to calculate the %IA/g for each tissue.

In vivo dosimetry estimation

To estimate the dosimetry of ²²⁵Ac-NM600, ex vivo biodistribution results were used. Allometric scaling was first performed to estimate the organ mass with respect to body mass, assuming 20 g for C57BL/6. Then, the residence time in each organ, MBq-sec/MBq(inj), was determined using trapezoidal integration, assuming only physical decay after the last timepoint. Lastly, the residence time for each organ was multiplied by the dose factor, mGy/MBq-sec, for each organ, which was assumed to be a sphere of equal mass that only receives self-dose. The total dose for each organ was computed by summing all contributions from the complete decay chain of ²²⁵Ac, considering negligible redistribution of daughter isotopes. Total absorbed dose estimates were not multiplied by an RBE/weighting factor. SPECT/CT-based ¹⁷⁷Lu-NM600 dosimetry was estimated according to a previously described method^29,31. Total injected activity within each organ was calculated from the extrapolation of %IA/g at a given time point. The standard mouse model was used to convert organ-specific cumulative activity into absorbed dose per injected activity (Gy/MBq). Dose contributions from surrounding organs were also included in calculations.

Ionizing-radiation quantum imaging detector scanning

Mice bearing subcutaneous MC38 tumors were injected with 3.70 MBq ⁸⁶Y-NM600 and 18.5 MBq ^90Y-NM600 or 13.0 MBq ¹⁷⁷Lu-NM600, then euthanized via CO₂ asphyxiation at 48 h (n = 2 mice (independent biological replicates)/treatment group). Tumors were harvested and frozen over dry ice in Tissue Tek Optimal Cutting Temperature (OCT) compound (Cat # 4583). Each tumor was sectioned in 10 µm sections on a Leica CM1950 cryostat and placed on charged glass slides (Fisher Scientific, Cat # NC1693867) for imaging on Ionizing-radiation Quantum Imaging Detector (iQID). The ⁸⁶Y/⁹⁰Y-NM600 tumor was sectioned at a depth of 1520 µm from the edge of the tissue. The ¹⁷⁷Lu-NM600 tumor was sectioned at a depth of 1500 µm from the edge of the tissue. Two serial sections were placed on an individual slide. For ¹⁷⁷Lu and ⁹⁰Y, an image intensifying screen to detect low-and-medium-energy beta particles was used (Carestream BioMax TranScreen LE). For this study, an in-house slide holder was used for the scans in order to fix scan bed position of the glass slides for subsequent image registration purposes between other imaging modalities. One scan was performed per slide; scan durations: 12 h for ⁸⁶Y/⁹⁰Y-NM600 and 24 h for ¹⁷⁷Lu-NM600. Histological images of each tumor section were captured and stitched together in the Infinity Analyze 7 software using an Infinity2-2M monochrome USB camera coupled to an Olympus CKX41 Inverted Brightfield Microscope using a 2× objective. Once histological and iQID autoradiography images were acquired, shadowgraph images used to overlay iQID images and histological images were obtained. Using a phosphorescent disk attached to the lid of the iQID enclosure, shadowgraphs were obtained. Shadowgraphs and histology images were imported into ImageJ. Once imported, the use of the SIFT feature extraction identified shared image features between the shadowgraph and the corresponding histological image. Rigid transform was selected between the two images, with the shadowgraph being used as the source image and the histological image being used as the target image. Once sufficient features were identified, bunWarpJ, a deformable image registration application, was used to register histology images to their corresponding shadowgraphs. Adobe Photoshop was used to produce a comprehensive iQID/histology overlay.

Mice bearing subcutaneous B78 tumors were injected with 0.74 MBq ⁹⁰Y-NM600, 1.70 MBq ¹⁷⁷Lu-NM600, or 18.5 kBq ²²⁵Ac-NM600, then euthanized via CO₂ asphyxiation at 72 h (n = 2 mice (independent biological replicates)/treatment group). Tumors were harvested and frozen over dry ice in Tissue Tek Optimal Cutting Temperature (OCT) compound (Cat # 4583). Each tumor was sectioned in 20 µm sections on a Leica CM1950 cryostat and placed on charged glass slides (Fisher Scientific, Cat # NC1693867) for imaging on iQID. Four serial sections from each tumor were placed on a single slide. Slides were placed against the iQID detector window separated from the input window by a thin sheet of mylar (Ludlum Measurements 01−5859) and the scintillator of choice depending on the radioisotope being scanned. The same scintillator/mylar setup as for the MC38 study was utilized for ⁹⁰Y- and ¹⁷⁷Lu-NM600 tumors. For ²²⁵Ac, a ZnS:Ag detector was used (Eljen Technology EJ-440). One scan was performed per slide; scan durations: 1 h for ⁹⁰Y-NM600, 2 h for ¹⁷⁷Lu-NM600, 12 h for ²²⁵Ac-NM600.

Toxicity assessments

To evaluate ²²⁵Ac-NM600 toxicity, comprehensive metabolic panel (CMP) and complete blood count (CBC) studies were performed in control and therapy mice. Groups of naïve C57BL/6 mice (n = 21/group) were administered 18.5 kBq (MTA) high A_s (2.70 μg NM600/MBq ²²⁵Ac) ²²⁵Ac-NM600 or 20 ng cold NM600 intravenously. Weights were recorded for n = 6 mice (independent biological replicates) on days 0, 7, 14, 22, 29, 36, 43, and 6 months post-injection. Three mice (independent biological replicates) were culled from each cohort on days 0 (pre-injection), 7, 22, 28, 43, and 6 months post-injection, and 500 μL of blood was collected via axillary bleed. CBC analysis was completed using whole blood and an Abaxis VetScan HM5 hematology analyzer (Union City, CA). The remaining blood was centrifuged at 1968 × g for 10 min to separate the serum from red blood cells; the serum was then run on an Abaxis VetScan VS2 analyzer (Union City, CA). Additionally, n = 3 mice (independent biological replicates) were administered 18.5 kBq low As (81.1 μg NM600/MBq ²²⁵Ac) ²²⁵Ac-NM600 intravenously. Weights were recorded for n = 3 mice (independent biological replicates) on days 0, 14, 41, and 6 months post-injection. Six months post injection, CBC and CMP analyses were performed by the same method as the other treatment groups.

In vitro radiation therapy

Delivery of EBRT (195 kV) in vitro was performed using a RS225 Cell Irradiator (Xstrahl). Delivery of in vitro radionuclide therapy was performed as follows. Briefly, a stock solution of ²²⁵Ac (unconjugated radionuclide)-containing cell culture media was prepared based on previously described dosimetry calculations²³ to deliver either 0.25 or 1 Gy absorbed dose in 24 h to a cell monolayer in a six-well plate. For 0.25 Gy, 6.882 kBq ²²⁵Ac was added to each well; for 1 Gy, 27.75 kBq ²²⁵Ac was added to each well.

In vivo radiation therapy

⁹⁰Y-, ¹⁷⁷Lu-, or ²²⁵Ac-NM600 therapy was administered via intravenous tail vein injection on treatment day 1 for MC38, B78, or Myc-CaP tumor-bearing mice. Table 1 shows injected activity for each RPT absorbed dose studied.

Delivery of EBRT (300 kV) in vivo was performed using an X-ray biological cabinet irradiator X-RAD CIX3 (Xstrahl). The dose rate for EBRT delivery in all experiments was approximately 2 Gy/min. Dosimetric calibration and monthly quality assurance checks were performed on these irradiators by the University of Wisconsin Medical Physics staff. Tumor-targeted EBRT was delivered at a dose of 2 Gy on day 1 with the right flank tumor exposed and the rest of the animal shielded using custom lead blocks.

Clonogenic assays

In vitro clonogenic assay of B78 and MC38 cells was performed²⁵. Briefly, cells were plated at a density of 100–4000 cells/well in a six-well plate (MIDSCI, Cat # TP92006), and irradiated with either 0, 2, 4, 6, or 8 Gy EBRT or 0.25 or 1 Gy ²²⁵Ac (n = 6 wells/treatment group; n = 12 wells/0 Gy control group). For EBRT-treated cells, cells were washed with phosphate-buffered saline (PBS) (Corning, Cat # 21-040-CV), and cell culture media was exchanged immediately following radiation. For ²²⁵Ac-treated cells, cells were washed with PBS and radioactive cell culture media was exchanged with non-radioactive cell culture media 24 h following start of radiation. Surviving colonies were stained with crystal violet to aid colony counting. Colonies containing >50 cells were scored to determine plating efficiency and the fractions of the cells surviving after each radiation dose. The log surviving fraction of control and irradiated colonies were calculated and plotted.

In vitro co-culture

MC38 cells or no cells were plated in 24 well plates (20,000 cells per well) containing DMEM media. Fresh media was exchanged after 24 h of culture, and peripheral blood mononuclear cells (PBMCs) harvested from treated or naïve mice were added. Blood (200 µL) was collected by submandibular vein collection. Fifty microliters of whole blood was used for CBC analysis. The remaining ~150 µL whole blood treated with RBC lysis buffer (BioLegend, Cat # 420301) and washed with PBS prior to co-culture. 200,000 PBMCs were added per well, for a 10:1 effector to target ratio. After 48 h of co-culture, PBMCs were harvested for analysis of activation markers using flow cytometry. Number of mice (independent biological replicates)/group is indicated in figure legend.

Flow cytometry

Flow cytometry was performed⁵⁹ using fluorescent beads (UltraComp Beads eBeads, Invitrogen Cat # 01-2222-42) to determine compensation and fluorescence minus one (FMO) methodology to determine gating. Rainbow beads (Spherotech, Cat # NC9025287) were used to set voltages for each flow cytometry timepoint. For in vivo analysis, tumors were harvested and gently dissociated. Spleens and tumor draining lymph nodes were harvested and manually dissociated. Blood was collected by submandibular vein collection. Spleens and blood were treated with RBC lysis buffer and washed with PBS prior to staining. Cells were treated with CD16/32 antibody (BioLegend, Cat # 156603) to prevent non-specific binding. Live cell staining was performed using Ghost Red Dye 780 (Tonbo Biosciences, Cat # 50-105-2988) according to manufacturer’s instructions. After live-dead staining, a single cell suspension was labeled with the surface antibodies at 4 °C for 30 min and washed three times using flow buffer (2% FBS + 2 mM EDTA (Thermo Fisher, Cat # 15575020) in PBS). For intracellular staining, the cells were fixed and stained for internal markers with permeabilization solution according to manufacturer’s instructions (BD Cytofix/Cytoperm, BD Biosciences, Cat # 554714). Flow cytometry was performed using an Attune NxT Flow Cytometer (Thermo Fisher). Data were analyzed using FlowJo Software. A complete list of antibody targets, clones, and fluorophores is provided in Supplementary Table 51. Gating strategies for each flow cytometry experiment are shown in Supplementary Figs. 12–16. The number of mice (independent biological replicates) per group is indicated in figure legends.

Immunohistochemistry

Standard immunohistochemistry (IHC) methods were performed for Fig. 6 and Supplementary Fig. 10^25,60. IHC antibody targets and clones are listed in Supplementary Table 52. All labeling was performed with no primary antibody negative controls. CD20⁺ cell clusters were counted for an entire tumor cross-section, with groupings of ≥5 cells being considered a cell cluster. N = 3 (⁹⁰Y-NM600 + ICI, ²²⁵Ac-NM600 + ICI, EBRT + ICI, ICI) or n = 4 (¹⁷⁷Lu-NM600 + ICI) tumors (independent biological replicates) were quantified for positive labeling by a blinded observer.

For Supplementary Fig. 5 PD-L1 IHC, mice bearing 6–10 mm MC38 or B78 tumors were administered 100 µg anti-murine PD-L1 (rat IgG2b, clone 10F.9G2, BioXCell, Cat # BE0101) or PBS by intraperitoneal injection. Mice were euthanized via CO₂ asphyxiation, and tumors were collected 24 h post-injection of anti-PD-L1 or PBS. Tumors were embedded (unfixed) in OCT and sectioned at a thickness of 8 µm. The anti-rat detection ImmPRESS kit (Vector Laboratories, Cat # MP-7444-15) was used for anti-PD-L1 detection according to manufacturer instructions.

Quantitative messenger RNA analysis

Mouse tumors were dissected and placed in tubes containing 2.8 mm ceramic beads (Fisher, Cat # 15-340-154) with TRIzol reagent (Invitrogen, Cat # 15596018). The tumors were homogenized for 60 s using a Bead Ruptor Elite (OMNI). Total RNA was extracted using the RNeasy Mini Kit (QIAGEN, Cat # 74106) according to manufacturer instructions. Isolated RNA was subjected to complementary DNA synthesis using the QuantiTect Reverse Transcription Kit (QIAGEN, Cat # 205313) according to manufacturer instructions. Quantitative PCR (qPCR) was performed using the CFX96 real-time system (Bio-Rad) with PowerUp SYBR Green master mix (Applied Biosystems, Cat # A25779). Relative messenger RNA (mRNA) expression of target genes was determined according to the 2^−ΔΔCT method using Hprt as a reference gene⁶¹. Forward and reverse primer sequences provided in Supplementary Table 53 were used to detect mRNA expression^15,21. The number of mice (independent biological replicates) per group is indicated in the figure legend.

Single-cell RNA sequencing

Tumors were dissected from three mice/treatment group, pooled by treatment, minced, and digested using the Miltenyi Biotech tumor dissociation kit (mouse, tough tumor dissociation protocol, Cat # 130-096-730) for 40 min at 37 °C. Cells were then strained through a 70 µm filter (Fisher, Cat # 08-771-2) and washed with RPMI-1640, and resuspended in 6 mL of RPMI-1640. Tumor infiltrating lymphocytes were isolated by Ficoll-Paque density gradient centrifugation⁶². Briefly, 4 mL of Ficoll-Paque (Cytiva, Cat # 17544602) was added to a 15 mL tube, and the entire cell suspension was carefully layered on top. The samples were centrifuged at 600 × g for 20 min with no brake or accelerator. Cells were collected from the layer interface, rinsed with RPMI-1640 at 375  × g, strained through a 40 µm filter, and counted on a Countess 3 automated cell counter (Invitrogen). All samples had viability > 90% and were resuspended at a concentration of 1.2 × 10⁶cells/mL in RPMI-1640 for single-cell RNA sequencing.

scRNA-seq was then performed⁶³. Briefly, single-cell suspensions from isolated CD45⁺ cells were loaded on a Chromium iX single cell instrument (10x Genomics) to generate single-cell beads in emulsion and scRNA-seq libraries were prepared. For Gel bead in EMulsion (GEM) generation and library preparation, the following materials from 10x Genomics were used: Chromium Next GEM Single Cell 5′ Kit v2 (PN-1000263); VDJ Amplification, Mouse TCR Kit (PN-1000254); 5′ Feature Barcode Kit (PN-1000256); Library Construction Kit (PN-1000190); Chromium Next GEM Chip K Single Cell Kit (PN-1000286); Dual Index Kit TT Set A (PN-1000215); Dual Index Kit TN Set A (PN-1000254). The Chromium Next GEM Single Cell 5′ v2 User Guide was followed. Single-cell barcoded cDNA libraries were quantified using a Qubit Fluorometer with Qubit dsDNA HS reagent (Invitrogen, Cat # Q32851), assayed using a Tapestation D1000 screentape (Agilent) and sequenced on a NovaSeq 6000 S2 flow cell (Illumina) according to 10X Genomics recommendations (26 cycles read 1, 10 cycles i7 index read, 10 cycles i5 index read, and 90 cycles read 2). Cells were sequenced to greater than 50,000 reads per cell for the 5′ gene expression library and greater than 5000 reads per cell for the V(D)J library as recommended by manufacturer.

Cell Ranger (version 7.2.0) was used for demultiplexing, barcode processing, and gene counting. Reads were aligned to the mouse genome (version mm10). Doublets were removed by scDblFinder (version 1.14.0). We removed low-quality cells that had total read counts fewer than 1000, had log(“total read counts” + 1), log(“number of genes with read count” + 1), or fraction of read counts in the top 20% of genes beyond five times of median absolute deviation (MAD), or had fraction of read counts from mitochondria beyond three times of MAD. Read counts were normalized across all samples by the logNormCounts function in scuttle (version 1.10.1). Dimensionality reduction was performed by Scanpy (version 1.9.1). Cell types were predicted by SingleR (version 2.2.0) using the Immunological Genome Project reference data from celldex (version 1.10.1). Gene differential expression was analyzed by MAST (version 1.26.0)⁶⁴. Gene set enrichment analysis was performed by fgsea (version 1.26.0) using the mouse hallmark gene sets from MSigDB (version 2022.1). Following filtering and quality control, the number of cells (independent biological replicates) used for scRNA-seq analysis per treatment group is indicated in Supplementary Table 48.

For the TCR data, for each cell the cellranger multi pipeline assembles the V(D)J transcripts into contigs, aligns the contigs to the TCR reference sequence, and determines whether the contigs correspond to a CDR3 sequence by annotating both ends of the CDR3 sequences with V and J genes.

We analyzed each sample independently and considered high confidence, productive and full length CDR3 sequences. Following quality control, the number of cells (independent biological replicates) used for TCR analysis per treatment group is indicated in Supplementary Table 49. We defined a clonotype as the unique combination of TRA and TRB CDR3 sequences annotated by different V and J genes. Then, we stratified the T cells according to their clonal expansion (Fig. 7B). To summarize the diversity and clonal expansion at the T cell type level, we counted the number of T cells annotated with a clonotype in each cell type and then calculated the Shannon diversity, clonal expansion, and the D50 index^65,66. We examined the T cell combinations sharing the same clonotypes per sample by using the ComplexUpset R package (v1.3.3)^67,68. The number of cells (independent biological replicates) with clonotypes shared within effector and memory CD8⁺ T cells is shown in Supplementary Table 50.

Statistical analysis

Prism 10 (GraphPad Software) and R 4.1.2 (R Foundation) were used for all statistical analyses. Student’s two-sided t-test was used for two-group comparisons. One-way or two-way ANOVA with Tukey’s honestly significant difference (HSD) test was used to assess the statistical significance of mean differences in CBC, CMP, flow cytometry, immunohistochemistry, complete response rate, rechallenge rejection rate, and mouse body weight data. For tumor growth analysis, all available data was used. To compare treatment groups, linear mixed models after log base 10 transformation of tumor volume were fitted on cell line, time in days, and their interaction. Tumor volumes of 0 were present in all experiments except those represented by Figs. 1Q and 3G. In such cases, a small constant (1 × 10⁻⁴) was added prior to log transformation. Mouse ID was also included as a random intercept, accounting for correlation between measurements taken from the same mouse (Supplementary Tables 1, 4, 7, 9, 11, 14, 16, 18, 21, 23, 25, 28, 30, 32, 35, 37, 39, 42 and 45). Pairwise contrasts were adjusted using Tukey’s method (Supplementary Tables 2, 5, 8, 10, 12, 15, 17, 19, 22, 24, 26, 29, 31, 33, 36, 38, 40, 43 and 46). Kaplan–Meier method was used to estimate the survival distribution for the overall survival. Then, pairwise comparison of the overall survival was made using a log-rank test with Benjamini–Hochberg adjustment of p-values between levels of factors (Supplementary Tables 3, 6, 13, 20, 27, 34, 41, 44 and 47). All data are reported as mean ± standard error of the mean (SEM) unless otherwise noted. For all graphs, *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001. Number of data points and independent experiments are stated in all figure legends.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

C.P.K., A.J.O., P.A.C., and M.P.: animal work and analysis. C.P.K., W.J.J., and M.T.: flow cytometry staining and analysis. C.A.F., H.C.R., M.B.I., A.N.P., C.F.M., and R.H.: RPT radiochemistry. C.P.K., A.J.O., C.C., and C.F.M.: RPT biodistribution and toxicity studies. A.A.: iQID scans and analysis. C.P.K., W.J.J., P.A.C., A.K.E., A.G.S.: scRNA sequencing sample preparation. J.J.G., O.K., B.P.B.: PET/SPECT analysis and ⁹⁰Y-, ¹⁷⁷Lu-, ²²⁵Ac-NM600 dosimetry. C.P.K., M.H.: biostatistical analysis. P.L., R.W.S., J.M.V., and I.M.O.: bioinformatics analysis. C.P.K., W.J.J., and Z.S.M.: experimental design. J.P.W. and Z.S.M.: funding, planning, and supervision of the project and data interpretation. C.P.K. and Z.S.M. wrote the manuscript. All authors read and approved the final manuscript.

Peer review

Peer review information

Nature Communications thanks Udo Gaipl and the other anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

The single-cell RNA sequencing and T cell receptor sequencing datasets generated and analyzed in the current study have been deposited in the GEO repository under accession GSE275609. All data are included in the Supplementary Information or available from the authors, as are unique reagents used in this Article. The raw numbers for charts and graphs are available in the Source Data file whenever possible. Source data are provided with this paper.

Competing interests

J.J.G. is cofounder and Chief Innovation Officer of Voximetry, Inc. H.C.R. provides consulting services to Archeus Technologies, which holds the license rights to NM600 related technologies. O.K. provides consulting services to Voximetry, Inc. R.H. is a member of the scientific advisory board for Archeus Technologies. B.P.B. is cofounder and Chief Science Officer of Voximetry, Inc. J.P.W. is a founder and Chief Science Advisor for Archeus Technologies. Z.S.M. has served as a member of the scientific advisory board for Seneca Therapeutics, Archeus Technologies, NorthStar Medical Radioisotopes, and Cali Biomedical, as a consultant for Johnson and Johnson and Telix Pharmaceuticals, and has sponsored research agreements with Point Biopharmaceuticals and Telix Pharmaceuticals. Z.S.M. has received material support for research (drug reagents) from Bayer Pharmaceuticals, B.M.S., XRD therapeutics, Seneca Therapeutics, AstraZeneca, HiberCell, Apeiron, Nektar Therapeutics, and Invenra. R.H., J.P.W., and Z.S.M. are inventors on patents held by the University of Wisconsin Alumni Research Foundation related to select radiopharmaceutical therapies and the interaction of radiopharmaceutical therapies with immunotherapies. All other authors declare they have no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68834-1.