Programable Albumin-Hitchhiking Nanobodies Enhance the Delivery of STING Agonists to Potentiate Cancer Immunotherapy

Stimulator of interferon genes (STING) is a promising target for potentiating antitumor immunity, but multiple pharmacological barriers limit the clinical utility, efficacy, and/or safety of STING agonists. Here we describe a modular platform for systemic administration of STING agonists based on nanobodies engineered for in situ hitchhiking of agonist cargo on serum albumin. Using site-selective bioconjugation chemistries to produce molecularly defined products, we found that covalent conjugation of a STING agonist to anti-albumin nanobodies improved pharmacokinetics and increased cargo accumulation in tumor tissue, stimulating innate immune programs that increased the infiltration of activated natural killer cells and T cells, which potently inhibited tumor growth in multiple mouse tumor models. We also demonstrated the programmability of the platform through the recombinant integration of a second nanobody domain that targeted programmed cell death ligand-1 (PD-L1), which further increased cargo delivery to tumor sites while also blocking immunosuppressive PD-1/PD-L1 interactions. This bivalent nanobody carrier for covalently conjugated STING agonists stimulated robust antigen-specific T cell responses and long-lasting immunological memory, conferred enhanced therapeutic efficacy, and was effective as a neoadjuvant treatment for improving responses to adoptive T cell transfer therapy. Albumin-hitchhiking nanobodies thus offer an enabling, multimodal, and programmable platform for systemic delivery of STING agonists with potential to augment responses to multiple immunotherapeutic modalities.

scienti c advisory board member, and stockholder of Sitryx Therapeutics, a scienti c advisory board member and stockholder of Caribou Biosciences and holds stock options for Nirogy Therapeutics.He has consulted and received speaker fees from Merck, P zer, and Abbie.He has received research support from Incyte Corp. within the past three years.J.M.B. receives research support from Genentech/Roche and Incyte Corporation, has received advisory board payments from AstraZeneca and Mallinckrodt and is an inventor on patents regarding immunotherapy targets and biomarkers in cancer.

Introduction
Immune checkpoint inhibitors (ICIs) targeting CTLA-4 and PD-1/PD-L1 have revolutionized the treatment of an increasing number of cancers but are still only effective for a relatively small fraction of patients (~ 15%). 1 For many cancers, this can be attributed, in part, to poor tumor immunogenicity and an immunosuppressive (i.e., "cold") tumor microenvironment (TME) that restricts the in ltration and/or function of antitumor T cells. 2,3 he innate immune system plays a critical role in cancer immune surveillance, 4 with clinical evidence linking activation of certain pattern recognition receptor (PRR) signaling pathways to increased T cell in ltration and responses to ICIs in cancer patients. 5,6 ordingly, the relationship between innate and adaptive antitumor immunity has motivated the clinical exploration and continued development of agonists targeting PRRs, including toll-like receptors (TLRs), RIG-I-like receptors (RLRs), and stimulator of interferon genes (STING).Activation of these pathways can induce a coordinated antitumor immune response by triggering the production of type-I interferons (IFN-Is), proin ammatory cytokines, chemokines, costimulatory molecules, and other mediators that potentiate T cell responses and enhance the e cacy of ICIs. 4,7,8 PR agonists have typically been administered intratumorally as an "in situ vaccine" with the intent to stimulate a systemic adaptive immune response that mediates distal tumor regression and/or immune memory to protect against disease recurrence. 2,9 hile promising, intralesional therapy may not be feasible or practical for patients with metastatic, poorly accessible tumors, particularly for repeated dosing. 10Importantly, intratumoral administration has thus far yielded underwhelming outcomes in the clinic, 11 motivating a need for systemically administered therapies targeting PRR agonists.Amongst the PRRs, STING has emerged as one of the most promising targets for stimulating antitumor innate immunity, [12][13][14] with remarkable e cacy in preclinical models leading to clinical trials of a growing arsenal of STING-activating therapeutics. 15,16 owever, clinical exploration of STING agonists has been primarily restricted to intratumoral administration of cyclic dinucleotides (CDN) and, unfortunately, has yielded disappointing results. 17This can be partially attributed to both the aforementioned limitations of intratumoral administration as well as the poor drug-like properties of CDNs (i.e., anionic small molecules) that limit their activity and e cacy for systemic administration. 15This challenge has prompted the development of several promising nanoparticle-based drug carriers for CDNs 15,[18][19][20][21][22] as well as small molecule STING agonists with improved chemical properties for systemic administration. 15,23,24 Hwever, therapeutic targeting of STING remains a considerable challenge owing to multiple intertwined pharmacological barriers, including suboptimal pharmacokinetics and poor tumor accumulation, that limit e cacy and increase the risk of in ammatory toxicities. 15,25 ence, there is a need for drug delivery technologies that afford increased spatiotemporal control over the delivery of systemically administered STING agonists for the treatment of advanced and metastatic disease.
8][29] Albumin and albumin-binding chaperones have been widely employed to improve the delivery of chemotherapeutics, exempli ed by albumin-bound paclitaxel (Abraxane®) 30 , as well as protein 31 , peptide 32 , and nucleic acid therapeutics 30 .Inspired by this previous work that motivates the unexplored potential of albumin as a carrier for STING agonists, we engineered a high-a nity anti-albumin nanobody (i.e., single-domain antibody) for site-selective enzymatic bioconjugation of STING agonists via biorthogonal chemistry.Employing a conjugatable diamidobenzimidazole (diABZI) STING agonist as a clinically relevant example, we demonstrate that nanobody hitchhiking of STING agonists on serum albumin dramatically improves their pharmacological properties and increases tumor accumulation, leading to a reduction in tumor burden and improved therapeutic outcomes in multiple mouse tumor models.We further demonstrate the programmability of the platform for integrating tumor targeting and additional immunoregulatory functions through the development of a bispeci c nanobody-diABZI conjugate that binds to both albumin and the immune checkpoint ligand PD-L1.We demonstrate that use of this bivalent nanobody carrier for STING agonist delivery further increases tumor accumulation while also inhibiting immunosuppressive PD-1/PD-L1 interactions, resulting in a reprograming of the tumor microenvironment (TME) to a more immunogenic "hot" milieu and a priming of antitumor T cells that further potentiate responses to multiple immunotherapeutic modalities.Collectively, our study positions albumin-hitchhiking, nanobody-STING agonist conjugates as an enabling, multimodal, and programmable platform for cancer immunotherapy with high translational potential.

Synthesis of albumin-hitchhiking nanobody-STING agonist conjugates
We hypothesized that conjugation of a STING agonist to an albumin binding chaperone would extend blood circulation half-life and increase accumulation in cancerous tissue, enriching the production of cytokines and chemokines that facilitated the recruitment, proliferation, and activation of leukocytes to the TME, which promotes cancer cell death (Fig. 1a).While several promising albumin-binding molecules have been described, including small molecules, fatty acids, and peptides, 27,29 we elected to build our platform from a nanobody with high a nity for albumin because nanobodies are modular and programmable via genetic engineering, are molecularly well-de ned, are amenable to scalable industrial manufacturing, and are components of approved and clinically-advanced therapeutics, including ozoralizumab, which contains an anti-albumin nanobody domain. 31Additionally, we sought to avoid the potential risk of accelerated albumin clearance that can occur due to direct covalent drug conjugation strategies 27,29 and to minimize the liver accumulation associated with the use of lipid-based albumin binders 30 , a challenge also faced by many promising nanoparticle-based STING agonists. 19,21,33,34 Wetherefore recombinantly expressed a previously described nanobody domain -termed nAlb -that binds with nanomolar a nity to serum albumin (Fig. 1b). 35We modeled the binding of the nanobody domain to human serum albumin (HSA) using RoseTTAFold to generate the nAlb nanobody and RosettaDock to predict the binding site of the nanobody to the serum protein albumin.We found that the nAlb nanobody reached an optimal energy conformation through binding at domain IIB of HSA, indicating that nAlb does not compete with albumin binding to FcRn, which facilities its long serum half-life (PDB: 4N0F).The binding a nity of nAlb was veri ed using isothermal calorimetry (ITC) both at physiological pH (7.5) and at endosomal pH (5.5), where nAlb maintained nanomolar a nity to both HSA and recombinant mouse serum albumin (rMSA) (Fig. 1c, Fig. S1).
To enable site-selective ligation of STING agonists, we cloned the C-terminal of the nAlb nanobody to present a selective ligation tag (LPETGGHHHHHHEPEA) that acts as a substrate for an engineered pentamutant of sortase A designed to selectively ligate any primary amine-containing small molecule to the C-terminal of a protein, 36 offering high programmability in the design.Using this approach, we ligated an amino-PEG 3 -azide linker, which conferred a single azide functional handle on the nAlb nanobody and can be used to ligate cargo via strain-promoted azide-alkyne cycloaddition (Fig. 1d,f).While this strategy is amenable to ligation of diverse classes of STING agonists, we selected a diABZI compound since ongoing clinical trials are exploring similar agents as a systemically administered immunotherapy (e.g., NCT03843359).To enable covalent conjugation to the nanobody, we synthesized a diABZI variant that was functionalized with an azide-reactive DBCO group and a PEG 11 spacer (DBCO-PEG 11 -diABZI) at the 7 position of one of the benzimidazole groups (Fig. 1e, Fig. S2-4), a modi cation that is not predicted to interfere with diABZI binding to STING.We then used copper-free click chemistry to install a single DBCO-PEG 11 -diABZI STING agonist or a DBCO-functionalized sulfo-Cy5 (referred to herein as Cy5) dye onto the nanobody and veri ed precise 1:1 conjugation by electrospray ionization mass spectrometry (ESI-MS) (Fig. 1f) and sodium dodecyl sulfate polyacrylamide electrophoresis (SDS-PAGE) (Fig. 1g).
We evaluated the activity of the nAlb conjugated STING agonist (nAlb-diABZI) as well as the parent DBCO-PEG 11 -diABZI compound and a previously optimized diABZI 23 (compound 3; referred to henceforth as diABZI) in two human reporter cell lines for type-I interferon (IFN-I) production: monocytes (THP1-Dual) and lung carcinoma cells (A549-Dual) (Fig. 1h,i).We found that the DBCO-PEG 11 -diABZI variant retained a near identical EC 50 value to the original diABZI agonist from literature, while, as expected, the in vitro activity of the nAlb-diABZI conjugate was reduced but nonetheless maintained high sub-100 nM activity for IFN-I production.Further, we tested the activity of the nAlb-diABZI conjugate in murine bone marrow derived macrophages (BMDM), demonstrating that nAlb-diABZI stimulated the expression of STING-driven cytokines Ifnb1, Tnf, and CXCL10 after 4 hours (Fig. 1j).

Albumin-hitchhiking nanobodies exhibit tumor tropism and enrich cargo delivery
While still incompletely understood and variable across cancers and patients 29 , albumin can accumulate in tumor tissue through several interrelated mechanisms, including passive diffusion through leaky tumor vasculature, active transport via endothelial cell transcytosis, binding to SPARC (secreted protein acidic and rich in cysteine) produced by cancer cells, and cellular uptake and catabolism by cancer and tumor-associated immune cells such as macrophages. 27,29 lbumin has been reported to enter cancer cells and tumor-associated myeloid cells (e.g., macrophages) through both albumin-dependent, receptormediated pathways as well as by micropinocytosis. 27,29 hough mechanisms of cellular albumin internalization may vary between tumor and cell types, we sought to gain insight into how nAlb-diABZI enters cells and activates STING.First, we rst validated that intracellular uptake of nAlb-Cy5 was abrogated at 4°C indicating an active endocytotic mechanism (Fig. 2a, Extended Data Fig. 1); by contrast diABZI can enter cells by passive transport across the plasma membrane. 23We next assessed if albumin binding enhanced nAlb internalization by EMT6 and myeloid cells.To test this, we rst used ow cytometry compare the cellular uptake of nAlb-Cy5 to a negative control nanobody targeting GFP, nGFP-Cy5 (Fig. S5), in serum containing media, nding insigni cant or minor differences in cellular uptake between nAlb-Cy5 and nGFP-Cy5 (Extended Data Fig. 1).While eliminating serum from culture media decreased nAlb-Cy5 uptake, this occurred to the same extent for nGFP-Cy5, again indicating that cellular uptake occurs predominantly in an albumin receptor-independent manner in these cell types.Albumin can also be internalized by cancer and tumor-associated immune cell populations through non-receptormediated micropinocytosis.To evaluate this, we inhibited micropinocytosis in EMT6 cells, RAW264.7 macrophages, and BMDMs using 5-[N-ethyl-N-isopropropyl] amiloride (EIPA), which signi cantly reduced nAlb-Cy5 uptake (Fig. 2b).Given that macropinosomes often tra c to lysosomes, we next assessed colocalization of nAlb-Cy5 with lysotracker and found that a substantial and similar fraction (> 50%) of nAlb-Cy5 and nGFP-Cy5 was colocalized with lysosomes or late endosomes in EMT6 and RAW264.7 cells (Fig. 2c,d).As expected, nAlb-diABZI did not mediate endosomal disruption as assessed using a previously described galectin 8 (Gal8) endosomal recruitment assay (Extended Data Fig. 1) 41 .
To gain insight into how amide-linked diABZI is released from the nanobody upon cellular internalization, we incubated nAlb-diABZI with lysosomes isolated from rat liver (Tritosomes), which are used to investigate stability and catabolism of molecules tra cked to an endosome-lysosome pathway, and used MALDI mass spectroscopy to assess the emergence of a PEGylated diABZI adduct that would be predicted due to amide bond cleavage by lysosomal proteases (Fig. S6).We observed the presence of this peak as early as 1 h following incubation with Tritosomes, suggesting that a fraction of nAlb-diABZI is lysosomally degraded to release a PEGylated diABZI variant.We synthesized this compound (Fig. S7-9) and evaluated in vitro activity in THP1 IFN-I reporter cells, nding that it had a similar EC 50 value to the previously described diABZI molecule, which can enter cells through passive transport 23 (Fig. S10).While in vivo mechanisms of albumin transport and cellular uptake are complex and still not fully understood 29 , taken together our data suggests that nAlb that accumulates at tumor sites is macropinocytosed, primarily by tumor-associated myeloid cells, resulting in lysosomal degradation and release of a diABZI variant that activates STING.
We next evaluated the pharmacokinetics and biodistribution of nAlb site-selectively conjugated to Cy5 as described for diABZI above (nAlb-Cy5) compared to an analogous control anti-EGFR nanobody (nEGFR) that we cloned and Cy5-labeled using the same strategy (Fig. S5).To assess the pharmacokinetic pro le achieved by using anti-albumin nanobody hitchhiking, we intravenously (I.V.) administered free DBCO-Cy5, nEGFR-Cy5, and nAlb-Cy5 in healthy female C57BL/6 mice and collected blood at discrete time points over several days (Fig. 2e).By measuring the concentration of Cy5 in the serum using uorescence spectroscopy, we determined the elimination half-life of both the free dye and the nEGFR-Cy5 conjugate to be approximately 5 minutes, matching the expected half-life of a typical nanobody that is rapidly cleared via renal excretion due to its small size (~ 15kDa). 37However, the nAlb-Cy5 conjugate had an elimination half-life of approximately 55 hours, consistent with in situ binding to and hitchhiking on serum albumin, which has a half-life of ~ 35-40 h in mice. 38By comparison the reported half-life of diABZI is ~ 90 minutes, 23 while that of CDNs is typically < 5 min. 33We next tracked the biodistribution of DBCO-Cy5, nEGFR-Cy5, and nAlb-Cy5 in female Balb/c mice with orthotopic EMT6 (EGFR + ) breast tumors inoculated in the mammary fat pad.At 24 hours post-administration, mice were euthanized, and major organs and tumors were imaged with an in vivo imaging system (IVIS) instrument to evaluate Cy5 biodistribution (Fig. 2f,g) and tissue was homogenized for quanti cation of Cy5 using uorescent spectroscopy (Fig. 2h).We observed minimal Cy5 uorescence in major organs for both nEGFR-Cy5 and nAlb-Cy5 conjugates, but substantial tumor accumulation of only the nAlb-Cy5 conjugate, corresponding to ~ 11% injected dose/gram tissue, signi cantly higher than other organs; similar ndings were observed in a B16.F10 tumor model (Fig. 2i).Immuno uorescent staining of excised and cryosectioned tumors (Fig. 2j) further con rmed nAlb-Cy5 accumulation at the tumor site, with the highest Cy5 uorescence observed proximal to CD31 + tumor vasculature and with Cy5 signal also observed within the tumor stroma (e.g., colocalizing with CD45 + immune cells).Albumin binding to SPARC expressed in tumor tissue has also been implicated in increased accumulation of albumin-binding therapeutics, 27 and we found that SPARC is expressed in both EMT6 and B16.F10 tumors (Fig. S11) and may therefore contribute to nAlb accumulation.
Based on the signi cant tumor accumulation of nAlb-Cy5, we next used ow cytometry to determine which tumor-associated cell populations internalized the conjugate (Fig. 2k,l, Fig. S12).At 24 h after I.V. injection of nAlb-Cy5, we found that ~ 8% of all live cells in the tumor were Cy5 + (Fig. S13) and, amongst Cy5 + cells, the majority were CD45 − CD31 − cells, which are primarily cancer cells, and tumor-associated CD11b + F4/80 + macrophages (Fig. 2k).Cancer cells (CD45 − CD31 − ) and macrophages are the most prevalent cell populations in the EMT6 tumor model and have been reported to endocytose albumin in tumors. 39,40 valuating nAlb-Cy5 uptake within speci c cell populations, we found that ~ 5-10% of cancer cells (CD31 − CD45 − ), macrophages (CD11b + F4/80 + ), and dendritic cells (CD11c + ) were Cy5 + with a higher (~ 15-20%) frequency of Cy5 + CD45 − CD31 + endothelial cells and neutrophils (Fig. 2l).As assessed by Cy5 median uorescent intensity (MFI), the cell populations with the highest degree of nAlb-Cy5 uptake were CD45 − CD31 + endothelial cells, neutrophils, dendritic cells, macrophages, and cancer (CD45 − CD31 − ) cells (Fig. S12).To determine if this cellular uptake pro le was in uenced by STING activation, we concurrently administered nAlb-Cy5 with nAlb-diABZI and found that the addition of nAlb-diABZI primarily impacted the myeloid cell composition of the tumor at 24h, resulting in an increased frequency of neutrophils and MDSCs and a reduction in macrophages (Fig. 2k, inset) while slightly biasing nAlb-Cy5 uptake towards macrophages, dendritic cells, and neutrophils.We also evaluated cellular uptake of nAlb-Cy5 in the spleen (Fig. S12), which, while not a major organ of distribution for nAlb-Cy5, is a potentially important secondary lymphoid organ for generating systemic antitumor immunity, nding that ~ 5-10% of macrophages and dendritic cells were Cy5 + .Taken together, these data demonstrate that nanobody albumin hitchhiking can increase tumor accumulation to allow for endocytosis of cargo by multiple tumor-associated cell types.

nAlb-diABZI potently stimulates STING activation in the TME to inhibit tumor growth
Based on the ability of nAlb to enhance cargo distribution to tumor sites, we next performed a doseresponse response study to evaluate the therapeutic e cacy of nAlb-diABZI conjugates in an established non-or low-immunogenic (immunologically "cold") B16.F10 tumor model that is resistant to ICB (Fig. S14). 42Using a treatment regimen that we and others have employed for evaluation of STING agonists, 33,43 we intravenously administered nAlb-diABZI to mice bearing ~ 75 mm 3 B16.F10 tumors at doses ranging from 5-0.05 µg diABZI content, nding that all doses signi cantly (P < 0.0001) inhibited tumor growth and extended survival time.Notably, the 5 µg dose signi cantly (P < 0.0001) enhanced e cacy relative to a 3x higher dose of diABZI, demonstrating the enhancement in potency enabled through albumin-hitchhiking.While the 5 µg dose resulted in ~ 10-12% weight loss, this was transient and occurred only after the rst injection (Fig. S14a).Nonetheless, towards maximizing the safety pro le of the treatment, we selected a dose of 1.25 µg, con rmed antitumor e cacy of both a single and three-dose regimen in the B16.F10 model (Fig. S14d-g; Fig. S15), and performed a preclinical analysis of nAlb-diABZI toxicity (Extended Data Fig. 2).Healthy mice were administered vehicle (PBS) or nAlb-diABZI (1.25 µg diABZI) intravenously three times spaced three days apart, weight loss was monitored daily, and blood was collected 4 and 24 hr after the rst injection for analysis of serum cytokines (Extended Data Fig. 2).In response to nAlb-diABZI, mice experienced only a mild (~ 5%) and transient weight loss similar to that described for nanoparticle-based delivery of STING agonists 18,19,21 with elevated plasma levels of STING-driven cytokines with antitumor functions (e.g., type I IFN, IL-12 4 h following injection, which returned to near baseline by 24 h.Mice were euthanized a week following the last injection, blood was collected for biochemistry analysis (Extended Fig. 2d), and major organs were isolated for histological evaluation (Extended Fig. 2e) by a board certi ed veterinary pathologist, who observed no clinically signi cant changes between the untreated control mice and nAlb-diABZI treated mice, consistent with insigni cant or minor changes in blood biochemistry and cellular composition.Based on this favorable safety pro le at a therapeutically effective dose in a challenging B16.F10 tumor model, we selected a dose of 1.25 µg for all subsequent studies.
Given the signi cant tumor accumulation of nAlb observed in orthotopic EMT6 breast tumors -and considering that only approximately 20% of breast cancer patients bene t from PD-1/PD-L1 ICB 44 -we next evaluated the capacity of nAlb-diABZI to create a TME that inhibited tumor growth.Female Balb/c mice were inoculated with EMT6 cells in a mammary fat pad (MFP) and treated with nAlb-diABZI, free diABZI, or vehicle (PBS) at a tumor volume of ~ 75 mm 3 (Fig. 3a).Interestingly, treatment with nAlb-diABZI strongly suppressed tumor growth whereas the free diABZI STING agonist did not confer a therapeutic bene t (Fig. 3b,c).Consistent with accumulation of nAlb at tumor sites, we found a notable increase in the expression of genes associated with STING pathway activation, including Ifnb1, Cxcl10, Cxcl9, and Tnfa (Fig. S16).
To gain insight into the immunological mechanisms by which nAlb-diABZI inhibited tumor growth, we used multispectral ow cytometric immunophenotyping to quantify changes in key myeloid and lymphocyte populations and their phenotypes (Fig. 3d-j, Extended Data Fig. 3) in EMT6 tumors and in the spleen 24 h following the third nAlb-diABZI administration.We found that administration of nAlb-diABZI increased the in ltration of CD8 + T cells with considerably elevated markers of activation (CD69) and proliferation (Ki67) -as well as the frequency of Ki67 + PD-1 + CD8 + T cells -which have been correlated with favorable responses to immunotherapy in patients. 45While there was a reduction in the overall frequency of CD4 + T cells this was associated with an increased frequency of CD69 + Ki67 + and Ki67 + PD-1 + CD4 + T cells.There was also a signi cant increase in the frequency of NK cells and Ki67 + NK cells in the TME; interestingly, the levels of splenic CD69 + and Ki67 + NK cells were also elevated, potentially suggesting mobilization of NKs from the spleen to the tumor (Fig. S17). 46Trends towards increased frequency of MDSCs (Fig. 3d,e), a signi cant increase in the frequency of FoxP3 + CD4 + regulatory T cells (Fig. 3e,f), and elevated MHC-II and PD-L1 on macrophages (Fig. 3g,h) was also observed.Similar ndings have been described for other STING agonists, which may act as counter regulatory mechanisms that may contribute to resistance to nAlb-diABZI as a monotherapy.In particular, MDSCs have been reported to reduce the e cacy of some STING-activating therapies [47][48][49] and we therefore evaluated nAlb-diABZI in combination with orally administered SX-682, which inhibits CXCR1/2 chemokine receptors involved in MDSC recruitment 50 , but, surprisingly, found that SX-682 tended to reduce nAlb-diABZI e cacy (Fig. S18).We also used anti-Gr1 antibodies to deplete MDSCs (primarily gMDSCs) 51 and again found a modest reduction in nAlb-diABZI e cacy (Fig. S19).Similar ndings have been reported by others, 52 re ecting the potentially complex roles of MDSCs in response to immunotherapy given their capacity to differentiate into mature antitumor effectors.Nonetheless, our data suggests that MDSCs do not strongly restrict the e cacy of nAlb-diABZI, at least in the EMT6 breast tumor model.
In addition to immunological resistance mechanisms, the e cacy of nAlb-diABZI may also be inhibited through generation of anti-nAlb or anti-diABZI antibodies that may lead to accelerated blood clearance. 53ile albumin has been described to generate immune tolerance to antigenic cargo 54 and nanobodies typically have low immunogenicity 55 , we addressed this possibility by intravenously administering healthy wild-type C57BL/6 mice with nAlb-diABZI on day 0, 3, and 6 and evaluated anti-VHH antibody titer in serum on day 14 and also compared the plasma half-life of nAlb-Cy5 to untreated mice.We did not detect an anti-VHH antibody response in serum (Fig. S20) and observed a similar nAlb-Cy5 half-life between untreated mice (~ 59 h) and treated mice (~ 64 h) (Fig. S21), suggesting that antibody-mediated nanobody clearance was unlikely to reduce nAlb-diABZI e cacy in our studies; however, this possibility cannot be discounted in humans where dose and treatment regimen, amongst other variables, will be different and therefore will need to be further investigated.

Engineering an albumin-binding, bivalent nanobody fusion for combined STING agonist delivery and immune checkpoint inhibition
Having demonstrated the potent antitumor effects of our albumin hitchhiking STING agonist, we next sought to leverage the modularity of nanobody engineering to confer additional immunotherapeutic functionality and demonstrate the programmability of the platform.As a translationally-relevant example, we introduced a second previously described nanobody domain that binds to PD-L1 (anti-programmed cell death ligand 1). 56Our rationale for selecting PD-L1 was two-fold.First, we, and others, have demonstrated synergy between STING agonists and PD1/PD-L1 ICB in suppressing tumor growth, including evidence that STING activation can directly upregulate PD-L1 expression. 43,57 econd, PD-L1 can be expressed by both cancer cells and immunosuppressive myeloid cells in solid tumors, 58 providing a molecular target for increasing tumor accumulation; indeed, anti-PD-L1 nanobodies have been employed previously in imaging applications with high selectivity for tumor tissue. 56We therefore hypothesized that an anti-albumin/anti-PD-L1 nanobody fusion would increase tumor targeting, while inhibiting immunoregulatory PD1/PD-L1 interactions that restrain responses to STING agonists.Thus, we generated a fusion protein that uses a genetic linker to connect both nanobody domains and maintained the C-terminal sortase ligation tag to generate an anti-albumin/anti-PD-L1 (AP)-STING agonist conjugate, termed AP-diABZI (Fig. 4a).We characterized the synthesis and generation of both anti-PD-L1 nanobody (nPD-L1) and AP conjugates to Cy5 and diABZI, showing that a single, homogeneous product that contained all three functional elements was formed (Fig. 4b-c, Fig. S1, Fig. S10).The in vitro activity of nPD-L1-diABZI and AP-diABZI was tested in A549-Dual and THP1-Dual IFN-I reporter cells (Fig. 4d,e) and by qPCR for analysis of STING-associated cytokines/chemokine gene expression in primary BMDMs and BMDCs (Fig. 4f, Fig. S22).We found that all nanobody-diABZI conjugates were potently active in both reporter cell lines without evidence of cytotoxicity (Fig. S23) and that nanobody-diABZI conjugates were more active than the parent DBCO-diABZI in BMDMs and triggered STING-associated gene expression with similar kinetics (Fig. 4f); both nAlb-diABZI and AP-diABZI were also active in murine bone marrowderived dendritic cells (BMDC; Fig. S22).Additionally, we showed using ow cytometry that the incorporation of the PD-L1 targeting domain enhanced binding and internalization in B16.F10 (PD-L1 low ) and EMT6 (PD-L1 high ) cells (Fig. 4g-h) relative to the albumin binding nanobody domain alone, which we further con rmed by comparing internalization by wild-type and PD-L1 KO EMT-6 cells (Fig. 4i).
We next tested the hypothesis that integrating a PD-L1 binding domain would increase tumor accumulation.We administered 2 mg/kg of Cy5-conjugated nEGFR, nPD-L1, nAlb, and AP nanobodies to healthy Balb/c mice I.V. and collected blood at discrete time points to evaluate pharmacokinetics (Fig. 4j).We also administered Cy5-conjugated nanobodies to mice with orthotopic EMT6 breast tumors and euthanized mice at 48 h to quantify nanobody-Cy5 conjugate biodistribution to major organs and tumors using IVIS (Fig. 4k-l).While the AP-Cy5 conjugate had a shorter elimination half-life than nAlb-Cy5 (17 h to 55 hours, respectively), likely due to binding of target PD-L1 in tissue and removal from circulation, both carriers maintained an increased elimination half-life and AUC relative to either targeted nanobody (nEGFR and nPD-L1) alone, which were cleared rapidly from circulation (Fig. 4i).While AP is approximately twice the size (~ 28 kDa) of the anti-PD-L1 nanobody, both are below the threshold for renal clearance 37 and, therefore, the increased circulation time of AP can be primarily attributed to the albumin hitchhiking functionality.Further, while the nPD-L1-Cy5 conjugate was observed at similarly low levels in major organs (liver and kidneys) and the tumor at 48 h (Fig. 4j), the AP-Cy5 conjugate demonstrated signi cant tumor accumulation (corresponding to 2.19 ± 0.43%ID/gram tumor) relative to major organs (Fig. 4k) and signi cant increase over nAlb alone (Fig. 4l-m).To further demonstrate increased tumor targeting, we compared the relative tumor accumulation of AP-Cy5 in breast tumors established using parental or PD-L1 knockout EMT-6 cells and found a signi cant decrease in tumor accumulation in the PD-L1 knockout model (Fig. 4m).It should be noted that PD-L1 was only knocked out of cancer cells in this model and that in ltrating myeloid cells can also express PD-L1 which may explain the modest < 2-fold decrease in AP-Cy5 accumulation.Nonetheless, these studies support our hypothesis that integrating a PD-L1 binding domain further improves delivery to tumor tissue.

AP-diABZI reprograms the TME to eliminate breast tumors and generate immunological memory that prevents recurrent disease
We next investigated the anti-tumor effects of systemically administered AP-diABZI fusion in the orthotopic EMT6 tumor model, comparing effects to those elicited by the constitutive components nAlb-diABZI and nPD-L1-diABZI (Fig. 5a-d).All nanobody carriers were administered I.V. at 1.25 µg of agonist.Additionally, mice were treated with commercially available anti-PD-L1 immune checkpoint blockade (ICB) IgG antibody to model an FDA-approved anti-PD-L1 ICI (e.g., Atezolizumab).A standard preclinical dose of 100 µg ICI was delivered intraperitoneally, which is a near equivalent molar dose of administered nanobody based on antigen binding domains (i.e., single domain for nanobody and two domains for antibody).Remarkably, treatment with AP-diABZI completely eliminated observable EMT6 tumors, resulting in a 100% complete response (CR) rate (10/10 mice) whereas treatment with nAlb-diABZI, while still very effective, yielded a 30% CR rate (3/10 mice); nPD-L1-diABZI only modestly inhibited tumor growth, though to slightly greater extent than the conventional anti-PD-L1 IgG ICB, which conferred only minimal activity in this model.Importantly, no additional toxicity was observed for AP-diABZI relative to nAlb-diABZI (Extended Data Fig. 2).To further assess this, we compared AP-diABZI to a combination regimen of nAlb-diABZI and ICB (i.e., anti-PD-L1 IgG) and observed comparably effective antitumor responses, suggesting that the improved e cacy of AP-diABZI over nAlb-diABZI can largely be attributed to immune checkpoint inhibition.Mice treated with AP-diABZI and nAlb-diABZI + ICB that exhibited complete responses were rechallenged 80 days after the initial tumor inoculation with the injection of EMT6 cells in a distal MFP and tumor growth monitored without additional treatment.In both groups, mice were largely resistant to tumor re-challenge with only 1/9 (AP-diABZI) or 1/8 (nAlb-diABZI + ICB) mice developing a tumor with the others remaining cancer free until at least day 100, demonstrating induction of memory lymphocytes that recognize EMT6 tumor antigens (Fig. 5e,f).We next evaluated the antitumor e cacy of AP-diABZI in mice inoculated with parental or PD-L1 knockout EMT6 cells and found that it was less effective (100% vs. 60% CR rate) when PD-L1 was not expressed by breast cancer cells (Fig. 5g,h, Fig. S24), potentially due to decreased tumor accumulation and/or reduced checkpoint inhibition.We also evaluated AP-diABZI in a MMTV-PyMT model of spontaneous breast cancer, nding that systemic administration of AP-diABZI signi cantly reduced tumor burden without evidence of increased lung metastasis (Extended Data Fig. 4), which has been implicated as a potentially deleterious consequence of STING signaling in some preclinical models. 59,60 gain insight into the mechanism underlying the increased e cacy of AP-diABZI, we treated mice bearing orthotopic EMT6 tumors with AP-diABZI, nAlb-diABZI, or PBS, collected serum at 4h following the rst dose for analysis of serum cytokines (Fig. S25), and euthanized mice 24 h after the third dose for gene expression analysis of tumor tissue using the NanoString PanCancer IO 360™ panel (Fig. i-m, Extended Data Fig. 5).Administration of nAlb-diABZI and AP-diABZI increased serum levels of antitumor IFN-I (IFN-α, IFN-β) and Th1 cytokines (e.g., IL-12, TNF-α) whereas nPD-L1-diABZI did not stimulate response, consistent with its low therapeutic e cacy; interestingly, only AP-diABZI notably increased levels of IFN-γ, a cytokine with an established role in antitumor immunity.Likewise, both nAlb-diABZI and AP-diABZI mediated considerable shifts in the gene expression pro le, with transcript signatures associated with increased immune cell in ltrate (immune cell tra cking, CD8 + T cells, NK cells, Th1 cells), tumor immunogenicity (antigen presentation, T cell priming, T cell recognition, costimulation, cytokine/interferon signaling), and cancer cell death/apoptosis, with AP-diABZI tending to exert a stronger effect relative to nAlb-diABZI (Fig. 5k-m, Fig. S26).
To further understand how AP exerts potent antitumor effects, we performed ow cytometric immunophenotyping of EMT6 tumors 48 h following the rst dose of nAlb-diABZI and AP-diABZI (Extended Data Fig. 6).We observed a decreasing frequency of live cancer cells (CD45 − ) within the tumor and found a signi cant (P < 0.0001) decrease in proliferating (Ki67 + ) cancer cells, consistent with the potent antitumor effects induced by AP-diABZI as well as gene expression analysis supporting increased cancer cell death.Interestingly, there was also an observed trend towards a decrease in PD-L1 expression within cancer cells.We found that a single dose of either nAlb-diABZI or AP-diABZI increased the in ltration of neutrophils and NK cells; more granulocytic MDSCs were also present, potentially contributing as an immunoregulatory mechanism to acute STING activation.However, as observed with nAlb-diABZI treatment, inhibition of MDSCs using SX-682 or anti-GR1 antibody depletion reduced AP-diABZI treatment e cacy (Fig. S18, S19, S27).While no change in the overall frequency of CD8 + T cells was observed at this early time point, tumor in ltrating CD8 + T cells tended to display a more activated phenotype (i.e., CD69 + ), which was also re ected in the splenic T cell population (Fig. S28 Motivated by these data, we studied the tumor and spleen immune cell dynamics after treatment with one, two, or three doses of AP-diABZI (Fig. 6, Fig. S29-30).We found that AP-diABZI increased the frequency of CD4 + T cells, CD8 + T cells and NK cells expressing markers of activation and proliferation, with a trend towards a stronger response after two and three doses, where a robust antitumor effect was observed (Fig. 6a-e).
Consistent with observations following a single dose and the potent anti-tumor e cacy of AP-diABZI, the frequency of CD45 − Ki67 + cancer cells was also reduced (Fig. 6a-d, Fig. S31).This is also consistent with gene expression pro ling (Fig. 5j-l) indicating increased NK and T cell in ltration and tumoricidal activity.
Within the tumor in ltrating T cell compartment, the percentage of CD8 + T cells increased with similar trends towards a more activated phenotype, and importantly, the ratio of CD8 + cells to FoxP3 + regulatory T cells was increased (Fig. 6b,c), indicative of a more immunogenic "hot" immune pro le within the TME.
Further, within both CD8 + and CD4 + T cells -both within the tumor and spleen -we observed a shift towards Ki67 + CD69 + and Ki67 + PD-1 + cells, indicating the prevalence of both proliferating and activated lymphocytes in response to AP-diABZI (Fig. 6d,e, Extended Data Fig. 7).Together, these data demonstrate that AP-diABZI increases the in ltration of CD8 + T cells and NK cells with an activated phenotype and that this effect is enhanced over the use of nAlb-diABZI alone, potentially implicating CD8 + T cells and NK cells as the primary antitumor effectors.To test this, we antibody depleted NK cells, CD8 + T cells, and CD4 + T cells and evaluated antitumor responses elicited by AP-diABZI treatment.Again, we observed a 100% CR rate for AP-diABZI, but treatment e cacy was almost completely inhibited with CD8 + T cell NK cell depletion, with CD8 + T cell depletion having a slightly stronger effect (Fig. 6f-h); no effect of CD4 + T cell depletion was observed.Therefore, both NK cells and CD8 + T cells are essential to the potent e cacy of AP-diABZI in an EMT-6 breast tumor model.
AP-diABZI inhibits B16.F10 tumor growth and primes an antigen-speci c memory CD8 + T cell response in situ We next assessed the e cacy of AP-diABZI in a more challenging and immunosuppressive B16.F10 melanoma model, initiating the three-dose treatment regimen when subcutaneous tumors reached an average size of ~ 75 mm 3 .As expected in this model, anti-PD-L1 ICB exerted no therapeutic bene t, whereas both nAlb-diABZI and AP-diABZI suppressed tumor growth and elongated median survival, with AP-diABZI conferring the most survival bene t, consistent with ndings in the EMT6 model (Fig. 7a-d).
We also found that AP-diABZI was more effective than free diABZI administered at 24 times (30 µg) the dose in the B16.F10 model (Fig. S32).We again evaluated cytokine levels in plasma 4 h following the rst injection using a multiplexed ELISA and found that anti-PD-L1 ICB increased only IL-1α levels, while nAlb-diABZI and AP-diABZI stimulated the production of cytokines associated with antitumor immunity, including IFN-α, IFN-β, IFN-γ, IL12p70, and CXCL10 (Extended Data Fig. 8).To determine the primary cellular effectors to AP-diABZI in the B16.F10 model, we antibody depleted CD4 + and CD8 + T cells and NK cells again nding that the antitumor response was mediated predominantly by CD8 + T and NK cells (Extended Data Fig. 9).STING activation can prime the immune system to stimulate a systemic, antigen-speci c, antitumor T cell responses with potential to lead to generation of T cell memory. 19,61 iven evidence of increased antigen presentation, cancer cell killing, and T cell priming (Fig. 5), as well as protection from tumor rechallenge in mice with EMT6 tumors treated with AP-diABZI, we next assessed the capacity of AP-diABZI to stimulate a de novo tumor antigen-speci c CD8 + T cell response.To test this, we inoculated C57BL/6 mice with B16.F10 melanoma cells expressing ovalbumin (B16.F10-OVA) as a model antigen and treated mice with either PBS or AP-diABZI on a three-dose regimen once tumors reached a size of 75-100 mm 3 (Fig. 7e-m).24 h after the nal dose, mice were euthanized for ow cytometric evaluation of splenic T cell response.Consistent with results in mice with parental B16.F10 tumors, AP-diABZI treatment signi cantly (P < 0.0001) reduced tumor burden (Fig. 7f).Treatment with AP-diABZI resulted in a signi cant (P < 0.0001) increase in activated CD69 + CD4 + and CD8 + T cells (Fig. 7h) and effector memory (CD44 + CD62L − ) CD4 + and CD8 + T cells, with a reduction in CD4 + central memory (CD44 + CD62L + ) T cells.

Albumin-hitchhiking STING agonists inhibit lung metastatic disease
Based on the evidence that AP-diABZI can stimulate an effective antitumor immune response in the immunologically "cold" B16.F10 model, we extended our investigations to evaluate therapeutic e cacy in an aggressive model of lung metastatic melanoma induced through intravenous inoculation of luciferase-expressing B16.F10 (B16.F10-Luc) cells (Fig. 8a).A week following inoculation, we used the three-dose combination therapy regimen described previously.On day 17 post-inoculation, mice were euthanized and lungs were harvested for quanti cation of tumor burden via measurement of lung mass, immunohistochemistry, and bioluminescent imaging (Fig. 8b-e, Fig. S33).High metastatic tumor burden was evident in mice receiving anti-PD-L1 ICB alone, but signi cantly (P < 0.0001) reduced in mice receiving nAlb-diABZI and nearly eliminated in mice receiving AP-diABZI.Importantly, these data show that albumin-hitchhiking STING agonists are effective against metastases in the lung, one of the most common metastatic sites for many cancers.This also suggests a potential to treat micrometastases, which typically lack the leaky vasculature required for tumor accumulation via the enhanced permeation and retention effect; 62 by contrast, albumin-binding molecules have been shown to accumulate in micrometases. 63-diABZI opens a therapeutic window for adoptive T cell transfer therapy Finally, we sought to demonstrate the versatility of the platform by extending the application of AP-diABZI to the setting of adoptive cellular immunotherapy, 64 which includes tumor in ltrating lymphocyte (TIL) therapy, CAR T cells, and TCR-engineered T cells that face major barriers to tumor in ltration and function, which continues to limit their clinical impact in the treatment of solid tumors.65,66 Founded on data demonstrating that nAlb-diABZI and AP-diABZI enhance the in ltration of endogenous antitumor T cells, we hypothesized that the approach could be used to pre-condition the TME to generate a therapeutic window for adoptive T cell therapy.To test this, we inoculated female C57BL/6 mice with subcutaneous B16.F10-OVA cells and allowed the tumors to reach approximately 75 mm 3 (Fig. 8f).We then treated mice with either one or three doses of AP-diABZI, followed by a single intravenous dose of activated OVA-speci c activated CD8 + T cells (OT-I T cells).Treatment with OT-I cells only (no STING agonist) on day 9 resulted in marginal therapeutic bene t (Fig. 8g-h), consistent with the highly immunosuppressive B16.F10 TME that restricts T cell in ltration and effector function.However, treatment with OT-I T cells 48 h after either one or three AP-diABZI doses conferred signi cant (P < 0.0001) reduction in tumor growth and prolonged mouse survival (Fig. 8i).Importantly, the treatment regimen of three doses of AP-diABZI prior to one dose of OT-I T cells resulted in a 25% complete response rate (3/12 mice).This provides additional evidence that albumin-hitchhiking STING agonists can establish an in ammatory milieu that supports T cell in ltration and function. Whle here we employed a simpli ed model of an adoptive T cell therapy, these studies highlight the potential to leverage nanobody-STING agonist conjugates to enhance responses to multiple T cell-based immunotherapies, including autologous TIL therapy, CAR T cells, and cancer vaccines.

Discussion
8][69] However, the development of innate immune agonists targeting speci c PRRs has been limited by pharmacological barriers that have largely restricted their application to intralesional administration, 4 which has yet to deliver on its clinical promise. 11This challenge has been recently exempli ed by the clinical exploration of STING agonists, which have demonstrated impressive results when administered intratumorally in mouse models but have not yet proven effective in patients.To address this, we developed a platform for systemic delivery of STING agonists based on an albumin hitchhiking nanobody (nAlb) engineered for precisely de ned and site-selective ligation of a DBCO-functionalized "clickable" diABZI cargo that we synthesized.Our data demonstrates that intravenously administered nAlb conjugates bind to circulating albumin in situ, increasing nanobody half-life from minutes to days and harnessing the capacity of albumin to accumulate in tumors to delivery of cargo to cancer cells and tumor-associated myeloid cells in the TME.This triggered potent STING activation at tumor sites, initiating an in ammatory program that increased the in ltration of activated NK cells and CD8 + T cells with antitumor function.Accordingly, nAlb-diABZI conjugates exhibit improved e cacy in mouse models of breast cancer and melanoma relative to a leading free diABZI agonist.
An appealing feature of anti-albumin nanobodies, and other protein-based albumin hitchhiking agents (e.g., a bodies, humabodies) over other albumin binders (e.g., lipids, Evans blue) is the high degree of molecular programmability that can be achieved through protein engineering.Here we illustrate this modularity by recombinantly integrating a PD-L1 binding domain to create a bi-valent fusion protein for covalent conjugation of diABZI.This yielded a single, well-de ned, multifunctional STING agonist that increased tumor accumulation in a PD-L1-dependent manner, while also blocking an important immune checkpoint, resulting in spontaneous induction of tumor antigen-speci c T cells that inhibited tumor growth and provided immunological memory that protected against tumor rechallenge.While we selected PD-L1 on translational considerations, the bivalent nanobody platform is readily amenable to integration of other immunoregulatory features and/or molecular targeting ligands.[77] Though there are vast future opportunities for bivalent nanobody-agonist conjugates, it is also notable that nAlb-diABZI was highly effective as a single agent, which may be advantageous for cancers that lack a de ned cell surface target (e.g., triple negative breast cancer, melanoma).However, clinical imaging has demonstrated that albumin accumulation varies amongst cancer types and patients 29 and the implications of this for the activity and e cacy of nAlb-diABZI must be considered and further investigated.To date, most research on albumin-based drug carriers for cancer (e.g., Abraxane, aldoxorubicin) has focused on delivery of chemotherapy drugs that target cancer cells.By contrast, immunostimulatory agents such as STING agonists can stimulate complex antitumor immunological programs that may be more dependent on immunological variables (e.g., neoantigen load, immune status of the TME) than on the e ciency of drug accumulation in tumor tissue or delivery to cancer cells.
For example, in our analysis of nAlb-Cy5 biodistribution, we found ~ 11% ID/gram tumor in the EMT6 model and a comparable ~ 8.4% ID/gram tumor in B16.10 model (Fig. 2), yet a substantial difference in response to both nAlb-diABZI alone and in combination with anti-PD-L1 that may be attributed to the relatively low immunogenicity to B16.F10 tumors.Importantly, the e cacy of nAlb-diABZI was enhanced when delivered in combination with anti-PD-L1 ICB and therefore may hold promise when combined with other ICIs and as an adjuvant therapy for patients with acquired resistance to ICIs.Additionally, nAlb-diABZI was much more effective than nPD-L1-diABZI, which was cleared rapidly with minimal tumor accumulation despite a capacity to activate STING, bind PD-L1, and inhibit immunoregulatory PD-L1/PD-1 signaling. 78][81] Indeed, anti-albumin nanobodies have been engineered with variable a nity 82 and this may afford a future opportunity for precisely modulating plasma half-life to establish immunopharmacological principles to optimize systemic innate immune agonist delivery.Also critical to the e cacy of our technology was the design and synthesis of a diABZI STING agonist functionalized with a DBCO group for biorthogonal conjugation to azide-presenting nanobodies.Despite being stably linked to the nanobody via an amide bond, the STING agonist exhibited high potency in vitro and in vivo, which we attribute to lysosomal degradation of endocytosed diABZI-nanobody conjugates and release of an active species (Fig. S6, Extended Data Fig. 1).While there may be an advantage to using such stable linkers to minimize premature drug release into the circulation 83 , our strategy also opens the possibility of installing cleavable linkers (e.g., enzyme cleavable, ROS cleavable) that enable environmentally responsive, "logic-gated" drug release with potential to further improve tumor speci city and reduce systemic exposure. 84,85 dditionally, while our selection of a diABZI agonist was largely motivated by their recent advancement into clinical trials, the platform is also amenable to conjugation to other STING agonists (e.g., recently described conjugatable CDNs) 18,22 as well as agonists targeting other PRRs. 86,87 summary, we have integrated synthetic biology tools to engineer precisely de ned nanobody-STING agonist conjugates as a platform for cancer immunotherapy.We leveraged albumin binding nanobodies as a scaffold from which diverse immunomodulatory components can be readily integrated via recombinant and chemical design.We demonstrated that albumin-hitchhiking nanobodies enhanced the potency and e cacy of a diABZI STING agonist and we showcased the versatility of the system by introducing an PD-L1 binding nanobody that affords increased tumor targeting and immune checkpoint inhibition to further potentiate antitumor immunity and e cacy.We found nanobody-diABZI conjugates to be highly effective in an orthotopic breast cancer model and an aggressive model of lung metastatic melanoma and we further demonstrated their utility an adjuvant therapy to improve responses to adoptive T cell transfer.Importantly, nanobody-diABZI conjugates were well-tolerated with a favorable preclinical toxicity pro le and are amenable to established scalable manufacturing work ows and translational pipelines.Collectively, our study establishes a preclinical foundation for future development of nanobody-and other protein-STING agonist conjugates as an enabling platform for cancer immunotherapy.
Cloning of Proteins.Gene cassette was purchased from IDT in the form of a gene block, with cloning restriction sites placed on both anking regions (BsaI -GGTCTC).In the case of a fusion protein, a genetic sequence was placed between the two domains (XTEN -SGSETPGTSESA).For sortase mediated bioconjugation of nanobodies, a C-terminal sequence was incorporated (LPETGGHHHHHHEPEA).The gene fragment was digested with BsaI-HF v2 in a golden gate master mix (New England Biolabs) and ligated into a pET28-b(+) plasmid.We transformed the construct into chemically competent DH5  (New England Biolabs) E. coli and plated on LB agar with Kanamycin.We transformed the sequence-veri ed pET28b plasmid into T7 Shu e Express (New England Biolabs) E. coli as the expression strain.Glycerol stocks were maintained at -80 o C of every successfully transformed bacterial strain.Expression and Puri cation of Proteins.5 µL of Kanamycin (stocked at 50 mg/mL) was added to a culture tube containing 5 mL 2xYT media and inoculated with a stab of protein (cloned into a NEB T7 Shu e Express cell line).The culture was incubated at 30 ºC, with shaking at 250 RPM, for 16 hours.

Engineered
Each culture was transferred to a 2 L ba ed ask containing 500 mL of autoclaved 2xYT media and 500 µL of Kanamycin (25 mg) and shaken at 30 ºC in an Innova 42R (New Brunswick Scienti c) incubator for 4.5-5 hours (until the OD 600 reached ~ 0.8).The cultures were then cooled to ~ 16°C and induced with IPTG (2.5 mM nal concentration).The induced cultures were shaken overnight (20-24 hours) at 16 ºC.The bacteria were harvested the next day by centrifugation (3900 rpm for 10 min) and the pellet was reconstituted in 1x PBS with Dnase I and a tablet of protease inhibitor cocktail (EDTA free).The cells were lysed by sonication on an ice bath in 5 second increments over 10 minutes.The resulting bacterial lysate was centrifuged (11000 rpm for 20 min) to remove cellular debris.The supernatant was added to a 50 mL Kontes Flex column (Kimbal Kontes Glassware) containing 3 mL of Nickel-NTA histidine binding resin that was preequilibrated with 1x PBS buffer.This column was placed on a rotating shaker at room temperature for 1-2 hrs.After this period, the supernatant was drained from the column using gravity and the column washed with 1x PBS buffer twice.Weakly bound proteins were rst washed off the resin using a low concentration elution buffer (2x 10 mL, 10 mM imidazole, 1x PBS pH 7.4 @ 25 ºC).The bound protein was then eluted from the resin using elution buffer (15 mL, 150 mM imidazole, 1x PBS pH 7.4 @ 25 ºC).The eluate was then concentrated to 0.5 mL in a 15 mL Microcon 10 kDa Centrifugal Filter Unit (Millipore) and subsequently puri ed by size exclusion chromatography (SEC) via an Akta FPLC (Cytiva), on a Hi-Load 16/60 Superdex 200 column using 1x PBS, pH 7.4 as the running buffer at 4 ºC.
Pure fractions were determined by SDS-PAGE, pooled together with buffer exchange to 1x PBS, and stocked at either − 20 ºC or 4 ºC.
Enzymatic Bioconjugation and Click Chemistry Reactions.Bioconjugation reactions occurred in mild conditions (20 mM HEPES at pH 7.4, 150 mM NaCl, and 10 mM CaCl 2 ) between eSrtA (100 µM) and a nanobody containing a C-terminal ligation tag (75 µM) using a primary amine containing functional group (20 mM).Reactions occurred with mixing by a rotary shaker overnight (16 h) and were quenched by the addition of a 1:1 volume of a chelating agent EDTA containing solution (20 mM HEPES at pH 7.4, 300 mM NaCl, and 10 mM EDTA) under rotation for one hour.After the reaction was stopped, the solution was concentrated and buffer exchanged to 1x PBS (without NaCl or MgCl 2 ) three times by centrifugal dialysis.The protein solution was then immobilized to Nickel-NTA histidine binding resin at least 2 hours, and unbound protein was collected by washing the resin with 1x PBS.For nanobodies that contain a histidine in the native sequence, proteins were eluted in mild conditions (10 mM Imidazole in 1x PBS).Collected protein was concentrated and buffer exchanged to 1x PBS by centrifugal dialysis and veri ed by ESI-MS and SDS-PAGE.Click chemistry reactions proceeded by the addition of 5 eq.(molar) of the complementary handle (e.g. if an azide was placed on the nanobody, the click chemistry reaction would proceed with the addition of 5 eq. of DBCO-containing moiety).For Cy5 conjugations, sulfo-Cy5 was used from Sigma Aldrich, and Cy5 was also used from Broadpharm.After 48 hours of reaction between the protein-azide and the DBCO-moiety under rotation at room temperature, the mixture was puri ed by centrifugal dialysis four times with 1x PBS, and veri ed for purity by UV-VIS, ESI-MS, and SDS-PAGE.
Sodium Dodecyl Sulfate-Polyacrylamide Gel Electrophoresis (SDS-PAGE).Protein samples were diluted in 1x PBS to 10 µM before analysis.10 µL of the protein sample was mixed with 10 µL of reducing Laemmli buffer.Samples were boiled at 95 ºC for 5 minutes, and 15 µL of each sample was loaded into a 15-well, 4-15% Tris-glycine precast polyacrylamide gel (Biorad) and ran at a constant 150 V with 343 mA for 30 minutes.The gel was then either rst imaged for uorescence on a UV-transilluminator or directly stained using Coomassie-B-250.
Electrospray Ionization Liquid Chromatography Mass Spectrometry (ESI-MS).Proteins were buffer exchanged into ammonium acetate (pH 5.5) and concentrated to approximately 100 µM.ESI-MS data were collected using an Agilent 6210A time-of-ight (TOF) mass spectrometer at a range of 50 − 20,000 m/z over a period of two minutes.Data were analyzed with Agilent MassHunter IM-MS Acquisition Data software to reveal m/z data, where les were condensed across the two-minute run.These m/z data were deconvoluted using a maximum entropy deconvolution calculation using UniDec to give the deconvoluted mass spectra using background subtraction between a range of 1,000-5,000 m/z and with an export range of 5,000-50,000 Da.
Computational Modeling and Analysis of nAlb Nanobody.nAlb was modeled in silico using RoseTTAFold (GitHub; RosettaCommons) and binding between HSA (PDB: 1AO6) and nAlb was predicted using RosettaDock through ROSIE (Rosetta Online Server that Includes Everyone; Pittsburgh).After an initial screening for best ts of the docking between HSA (receptor) and nAlb (ligand), the best t model was then returned for rescreening to con rm an optimal energy conformation between the structures.The nal structures of nAlb and the bound nAlb-HSA complex were exported to PyMOL for generating a gure of the structure.
Isothermal Titration Calorimetry.All proteins used were equilibrated in buffer at indicated pH values by titration using HCl (aq.) or NaOH (aq.) in PBST, 150 mM NaCl, 3 mM EDTA, 0.05% Tween 20.Albumin (HSA and rMSA) and PD-L1 receptor titrations were run on the TA Instruments A nity ITC instrument (TA Instruments, New Castle, DE).350 µL of albumin or the PD-L1 receptor was added to the sample cell (10-20 µM), and either the nAlb nanobody (125-250 µM) or nPD-L1 nanobody (250 µM) was loaded into the injection syringe, respectively.The reference cell contained ultrapure water and was changed after each titration experiment.All runs used the following instrument settings: cell temperature 298 K, reference power 10 µCal/second, initial delay 240 seconds, stirring speed 75 rpm, feedback mode/gain high, and injection volume 2 µL for 10 seconds, titration spacings at 120 second intervals, and a lter period of 10 seconds.Data were analyzed using the provided NanoAnalyze software for the instrument to determine thermodynamics of binding from an independent model.
Tritosome Degradation Assay and MALDI-TOF MS.Tritosomes (BioIVT) were prepared and activated by combining 70 µL of nuclease-free water, 10x of catabolic buffer (K5200, BioIVT), and 100 µL of pure lysosomes (H0610.L, BioIVT) and incubating the mixture at 37 ºC for 15 minutes.Samples for lysosomal degradation were added at 0.5 µM (10 µL) with the tritosome mixture and incubated at 37 ºC over a period of 48 hours.Aliquots were taken from the reaction mixture at distinct time points and ash frozen with liquid nitrogen and stored at -80 ºC to stop the reaction.Activity was determined by observing molecular weight shifts in the substrate using matrix assisted laser desorption and ionization time-ofight mass spectrometry (MALDI-TOF MS). 3 µL of matrix (15 mg mL-1 THAP in dry acetone) was combined with 1 µL of aliquoted sample and spotted on a stainless steel MALDI-MS plate (Bruker).Matrix was evaporated using compressed-air, read on a Bruker AutoFlex MALDI-TOF, and processed with the FlexControl software (Bruker Daltonics).The laser pulse rate was 500 Hz and spectra were obtained with a mass window of 400-4000 m/z at the highest resolution for the instrument (4.00 GS/s).
FlexAnalysis software (Bruker Daltonics) was used to obtain baseline spectra for all samples.Data were exported and plotted using MATLAB to generate gures showing m/z spectra at distinct time points.
Synthesis and NMR Veri cation of DBCO-PEG 11 -diABZI.Synthesis of the DBCO conjugated STING agonist (diABZI) is reported in Fig. S2, with NMR veri cation in Fig. S3-4.We rst generated a STING agonist that features a reactive amine handle, which was synthesized in four steps.Brie y: Aryl amination of an aryl chloride 1 with an amine 2 gave a di-nitro analog, compound 3.The di-nitro compound 3 was subjected to reduction using sodium dithionite in methanol, generating a di-amine moiety 4. Compound 4 was then treated with isothiocyanate, followed by EDC coupling, to reveal a bocprotected analog, compound 5. Next, the boc-group from compound 5 was deprotected by treating with TFA:DCM.To a stirred solution of amine 6 (100 mg, 0.089 mmol, 1 eq.) in 5 mL DMF, hunig's base was added (77 µL, 0.44 mmol, 5 eq.) under argon atmosphere at room temperature.After stirring for 5 min, a solution of activated NHS ester (98 mg, 0.098 mmol, 1.1 eq.) in DMF (5 mL) was added dropwise and stirred overnight (16 h).The solvent was evaporated to get crude product 7, which was puri ed by silica gel column chromatography using a mixture of Methanol/Dichloromethane as an eluent (5-25% MeOH) to get the desired product as a solid (70mg, 0.042mmol, yield 43%). (R f = 0.5 in 20% MeOH in DCM).In Vitro Reporter Cell Assays.Cell reporter assays were utilized in THP1-Dual and A549-Dual cell lines, as adapted from the manufacture protocols.Brie y, cells were plated at a density of 50,000 cells/well in a total volume of 180 µL of supplemented media in cell-culture treated 96-well plates overnight.After 24 h, cells were dosed with 20 µL of treatment groups (for a total volume of 200 µL/well in a 10:1 dilution, with either a 1:1 or 2:1 dilution down the plate) overnight.After 24 h of treatment, cells were pelleted at 1500 RPM for 5 minutes in the centrifuge, and 20 µL of supernatant was plated in a white-walled 96-well plate for analysis by QUANTI-Luc™ (InvivoGen) assay.After loading in a plate reader, 50 µL of QUANTI-Luc reagent was added to each well and luminescence was measured for determination of cell-based activity.To the remaining cells in the cell-culture treated 96-well plate, 30 µL of Cell-Titer Glo reagent (Promega) was added and the plate was incubated at 37 ºC for 1 h.After incubation, the plate was loaded into the plate reader and luminescence was measured to determine cell-mediated toxicity.Data were recorded in triplicate and analyzed in GraphPad PRISM (Version 10), with data reported with standard error of the mean (SEM).
Primary cells were seeded into 12-well plates for analysis by qPCR or 96-well plates for in vitro ow cytometry.
Quantitative RT-PCR (qPCR).RNA was extracted either from animal tissue (by TissueLyser II, Qiagen) or from in vitro cell cultures using the RNeasy® Plus Mini Kit (Qiagen) according to the manufacturer's protocol.cDNA was generated through a reverse transcriptase reaction using the iScript cDNA synthesis kit (Bio-Rad), by following the manufacturer's instructions.To run the qPCR, cDNA was mixed with TaqMan gene expression kits (primer and master mix) to a nal volume of 20 µL and run on the Bio-Rad CFX Connect Real-time System, with a threshold cycle number determination made by the Bio-Rad CFX manager software V.3.0.Primers used included: mouse Ifnb1 (Mm00439552_s1), mouse Tnf (Mm00443258_m1), mouse Cxcl10 (Mm00445235_m1), mouse Cxcl1 (Mm04207460_m1), and mouse Hmbs (Mm01143545_m1).Gene expression was rst normalized to the housekeeping gene, Hmbs, and then normalized to the PBS treatment within groups using the 2 − ddCt analysis method.
Colocalization Analysis.density 10 x 10 3 EMT6 and RAW 264.7 cells per well was plated in a glass bottom 96-well plate.After culturing for 12 h, cells were treated for 1.5 h.nAlb-Cy5 (2 µM) and nGFP-Cy5 (2 µM) were added to each well and further incubated for 4 h.Cells were treated with 50 nM of Lysotracker Green (Invitrogen) and 2 µM Hoechst (Invitrogen) for 10 min after the end of the incubation period.Wells were washed three times with phenol red free medium and visualized under confocal microscope (Zeiss LSM880).High-magni cation images were obtained using the 40x objective lens.
Manders' coe cient calculated by using Image J Software for colocalization analysis.
Nanobodies in Spontaneous Tumor A cohort of female FVB/N-Tg (MMTV-PyVT) 634Mul mice, bred in house, was used for these studies.Study animals were weighed, and mammary glands palpated twice weekly starting at 6 weeks of age.Tumor diameters in two dimensions were obtained using calipers.Treatment for the full cohort was initiated when rst palpable tumors appeared at approximately 8-10 weeks of age.The mice received the nanobody-STING agonist conjugate or vehicle control for a total of 3 treatments at 7 day intervals.The study was terminated 22 days after the rst treatment.At necropsy, tumors were removed and a wet weight obtained for each.One tumor was xed in 10% buffered formalin for histological analysis.All tumor measurements and analysis were performed by individuals blinded to treatment group.
Western Blot Analysis.Mice were euthanized and tumors (EMT6 and B16.F10) were harvested and 500 µl RIPA buffer (Sigma) supplemented with protease inhibitors (Sigma) was added in approximately 10 mg of tissue.Tissue was homogenized using a bead mill tissue homogenizer (TissueLyser II; Qiagen) and kept the on ice for 30 min.For in vitro analysis of STING expression in EMT6 cells were incubated in RIPA buffer for 10 minutes on ice.Protein concentration was measured using a BCA protein assay kit (Thermo Scienti c).Equal amount of protein (30 µg) was subjected to SDS-PAGE and transferred onto nitrocellulose membranes using the semi-dry transfer protocol (Bio-Rad).After transfer, membranes were probed with each respective primary antibody (anti-SPARC, anti-TBK1, anti-p-TBK1, anti-sting, anti Hsp-90 and anti-β-actin) overnight at 4°C.Following incubation, the membranes were probed with HRPconjugated secondary antibodies.All antibodies were purchased from Cell signaling.Protein bands were visualized using ECL western blotting substrate (Thermo scienti c).Images of immunoblots were obtained using a LI-COR Odyssey Imaging System.
Immuno uorescent analysis of EMT6 Tumors.5-micron Para n-embedded tissue sections were prepared for immuno uorescence and stained with anti-CD31 (cell signaling #77699; 1:500), anti-SPARC (cell signaling #5420, 1:500), and anti-CD45 (cell signaling #70257; 1:500).Tissue slides were depara ned in xylene and rehydrated in serial ethanol dilutions.Antigen retrieval was performed by heating slides for 17 minutes Tris EDTA buffer, pH 9 in a pressure cooker at 110 ºC.Slides were cooled to room temperature and then blocked with 2.5% horse serum (vector labs).After blocking, slides were incubated overnight at 4 ºC with primary antibody in horse serum.Slides were then incubated in antirabbit HRP secondary (vector labs) for 1 h at room temperature the following day and subsequently incubated in 1:500 Opal 520 (green) or Opal 570 (red) (Akoya) for 10 minutes.For serial staining, slides were stripped using Citric Acid buffer, pH 6.1 in a pressure cooker at 110 ºC for 2 minutes and then staining was repeated using different antibody and Opal uorophore.After the last Opal staining, slides were mounted using antifade gold mount with DAPI (Invitrogen).Stained images were acquired using a Keyence digital microscope system.Images were analyzed with Fiji software.Quanti cation of markers was done by measuring total amount of uorescence divided by total number of cells (DAPI).
Flow Cytometric Experiments and Analysis.EMT6 tumor bearing Balb/c and B16.F10-OVA bearing C57BL/6 mice were euthanized either 24 h or 48 h after nal treatment.Spleens and tumors were harvested, weighed, and placed on ice.Tumors were digested in RPMI 1640 media containing a tumor dissociation kit (collagenase III and deoxyribonuclease I, Miltenyi Biotech).Tumors were further dissociated using an OctoMACS separator (Miltenyi Biotech) and incubated for 30 min at 37 ºC for complete digestion.Tumors and spleens were mashed and separated into single cell suspensions using a 70 µm cell strainer (Fisherbrand™; Thermo Fisher Scienti c) and red blood cells were lysed twice using ACK lysis buffer (Gibco).Cells were resuspended in ow buffer (1x PBS supplemented with 2% FBS and 50µM dasatinib), counted, and stained with Fc-block (aCD16/32, 2.4G2, Tonbo) for 15min at 4 ºC, and then stained with the appropriate antibodies for 1hr at 4 ºC (found below and in Tables S1-S2).After staining, cells were then washed again with FACS buffer, xed with 2% paraformaldehyde for 10min, washed again with FACS buffer containing AccuCheck counting beads, and analyzed on a Cytek Aurora ow cytometer.All ow cytometry data were analyzed using FlowJo software (version 10; Tree Star; https://www.owjo.com/solutions/owjo).Representative ow cytometry plots and gating schemes are shown in Fig. S34-41.
Pharmacokinetics and Ex Vivo Imaging Experiments.Healthy (Balb/c or C57BL/6) and EMT6 tumor bearing (Balb/c) mice were injected with 100 µL of Cy5 (either as free dye or as a nanobody conjugate) at a dose of 2 mg/kg intravenously.For pre-treatment with the diABZI conjugated nanobody, a dose was prepared at 1.25 µg diABZI in 100 µL and injected 3 days prior to Cy5 dosing.Blood draws were taken using heparinized capillary tubes (DWK Life Sciences) at discrete time points up to ve days after injection.1 µL of blood was mixed with 50 µL of PBS, centrifuged, and the diluted plasma was collected for analysis.Prescence of Cy5 was determined by uorescence intensity using a plate reader, with an excitation wavelength of 645 nm and an emission wavelength of 675 nm.Pharmacokinetic analysis was performed in GraphPad Prism (V10) using either a one-phase decay or two-phase decay, in which the reported half-life is the second phase (elimination).Biodistribution studies were performed by excising and weighing hearts, lungs, livers, spleens, kidneys, and tumors.Tissues were washed in 1x PBS and transferred to the stage of the IVIS Lumina III (PerkinElmer).After IVIS, tissue were homogenized using cell disruption in a volume of 200 µL 1x PBS.Homogenized tissue were centrifuged and the supernatant containing the Cy5 dye was quanti ed for the tissue was determined by uorescence intensity using a plate reader.A standard curve was generated of free DBOC-Cy5 dye in 1x PBS and concentrations of Cy5 in tissue were calculated by tting the standard curve to a linear regression.Fluorescence (radiant e ciency) was measured with a maximum value of 1.56 x 10 10 , and a minimum of 8.21 x 10 8 , and areas were drawn manually for organs to generate average radiant e ciency values (per cm 2 ) using the Living Image software (version 4.5).For B16.F10-LUC studies, lungs were placed in black 12-well plates (Cellvis) and incubated for 5 min in a solution of 1 mg/mL Pierce™ D-Luciferin, Monopotassium Salt (Thermo Fisher Scienti c) in 1x PBS.Images were taken on the IVIS and luminescence was quanti ed as total radiant ux (p/s) for each set of lungs Serum Analysis for Anti-VHH Antibodies.Mice were pre-treated PBS or nAlb-diABZI (1.25 µg dose of diABZI) three times every three days, or treated once with nAlb-diABZI (1.25 µg dose of diABZI).14 days after the rst dose, blood was collected by cardiac puncture in and allowed to clot to extract serum.Tubes were centrifuged at 2000 x g for 15 min at 4 ºC, and the serum was then collected and diluted directly in PBS (1:4 to 1:8192) for analysis.MonoRab rabbit anti-camelid VHH antibody plates (GenScript) were used to determine anti-VHH antibodies in mouse serum.3 µg in 100 µL of anti-albumin nanobody were loaded into each well of the 96 well plate and allowed to incubate in the pre-coated antibody plate, sealed, and incubated at 37 ºC for 30 minutes.The plate was washed with 200 µL of PBST four times.Either the diluted mouse serum or a commercial Rabbit anti-Camelid VHH antibody (Genscript; A01860) were added in serial dilutions to the wells of the plate at a volume of 100 µL.The plate was sealed and incubated at 37 ºC for 30 minutes, followed by washing four times with 200 µL of PBST.A commercial secondary Goat anti-Mouse IgG-FITC conjugate (Invitrogen; 31547) or secondary anti-Rabbit IgG-FITC conjugate (Sigma; F9887) was added to the mouse serum or commercial anti-VHH, respectively, at 100 µL and incubated for 30 minutes at 37 ºC before washing with PBST four times (200 µL).The plate was quanti ed using the uorescence intensity of FITC (ex: 495 nm, em: 515 nm) using a plate reader.
Ex VivoPlasma Analyte Analysis.Blood was collected by either cheek bleed or cardiac puncture in K 2 EDTA-coated tubes (BD Biosciences).Tubes were centrifuged at 2000 x g for 15 min at 4 ºC, and the plasma was collected for analysis.Cytokine levels were evaluated using either the LEGENDplex™ Mouse Anti-Virus Response Panel (BioLegend) the LEGENDplex™ Mouse Cytokine Panel 2 (BioLegend), both with V-bottom plates, according to manufacturer's instructions, and data were collected using ow cytometry.Cytokine concentrations were interpolated from standard curves using an asymmetric sigmoidal 5-paramater logistic curve ts (GraphPad Prism V10).Bar plots comparing groups and heat maps of averaged values for groups were generated to analyze results.
NanoString nCounter Analysis of EMT6 Tumors.After three treatments of nAlb-diABZI (1.25 µg, n = 3-4), AP-diABZI (1.25 µg, or PBS (n = 3-4) in EMT6 bearing female Balb/c mice, tumors were isolated, digested, and 100 ng of RNA was isolated, as described in the qPCR section.RNA was hybridized to the IO360 PanCancer panel, as well as through a selected gene panel, of target-speci c uorescent barcodes and analyzed using NanoString nCounter MAX Analysis system.The fold change for genes within groups was calculated by comparing against the average normalized gene expression values within PBS treated mice.All statistical signi cance, and clustering analysis, was performed in R (http://cran.r-project.org)based on the genes provided in the IO360 PanCancer panel.
Safety Statement.All research performed in this study was done so with careful consideration of any risks that are inherent to the materials, instruments, and experiments performed.All research safety guidelines and considerations as provided by the safety data sheets (SDS) and university guides were adhered to for the duration of this study.
Statistics.All data were plotted and statistical analysis performed using Prism 10 (GraphPad) software.
Unless indicated in the gures, all data are presented as mean ± SEM.For comparisons between two groups, unpaired two-tailed Student's t-tests were performed as indicated.For multiple comparisons a one-way ANOVA was performed with post-hoc Tukey's correction for multiple comparisons.For tumor volume, statistically signi cance was examined through a two-way ANOVA followed by Tukey's adjustment for multiple comparisons.A Log-rank (Mantel-Cox) test was used to compare Kaplan-Meyer survival data.

Declarations Ethics statement
Studies involving the use of animals were completed under Animal Care Protocols approved by the Vanderbilt University Animal Care and Use Committee.The health assessment of animals was completed using a standard operating procedure also approved by the Vanderbilt University Animal Care and Use Committee.

Figure 1 Design
Figure 1

Figure 2 Anti
Figure 2

Figure 4 Design
Figure 4

Figure 6 AP
Figure 6

Figure 8 Albumin
Figure 8 ºC for 8 h and then lyophilized.Diethyl ether was added (×3) and vigorously shaken with diethyl ether.The mixture was decanted to remove excess amine-PEG3-azide.After three washes with diethyl ether, it was dried overnight in a vacuum chamber to obtain the desired compound (7 mg, 3.7 µmol, 77% yield).