A programmable bispecific nano-immuno-engager promotes T cell homing and reprograms tumour microenvironment


 Immune checkpoint blockade (ICB) therapy has revolutionized clinical oncology. However, the efficacy of ICB therapy is limited by the ineffective homing of T effector (Teff) cells to tumours and the immunosuppressive tumour microenvironment (TME). Here, we report a programmable tumour cells/Teff cells bispecific nano-immuno-engager (NIE) that can circumvent these limitations to improve ICB therapy. We have developed 28 nm non-toxic peptidic micellar nanoparticles (NIE-NPs) that bind α3β1 integrin on tumour cells membrane and undergo in situ transformation on surface of tumour cells into nanofibrillar network (NIE-NFs). The nanofibrillar network persistently facilitates cytotoxic T cells’ homing to the proximity of tumour cells via activatable α4β1 integrin ligands, and also allows sustained release of resiquimod to reprogram the TME. This bispecific NIE eliminates syngeneic 4T1 breast cancer and Lewis lung cancer models in mice, when given together with anti-PD-1 antibody. The in vivo structural transformation-based supramolecular bispecific NIE represents an innovative class of programmable receptor-mediated targeted immunotherapeutics to greatly enhance ICB therapy against cancers.


Introduction
Immune checkpoint receptor pathway blockade monoclonal antibodies such as anti-PD-1, anti-PD-L1, and anti-CTLA-4 can reverse T effector (T eff ) cell dysfunction and exhaustion, resulting in dramatic tumour shrinkage and sometimes complete remission in some patients, even with late stage metastatic diseases. 1,2 To date, the US Food and Drug Administration (FDA) had approved seven immune checkpoint blockade monoclonal antibodies (ICB-Ab): one CTLA-4 inhibitor (ipilimumab), three PD-1 inhibitors (nivolumab, pembrolizumab and cemiplimab), and three PD-L1 inhibitors (atezolizumab, durvalumab and avelumab), used either alone, or in combination with other therapies, against a range of tumour types. [3][4][5] However, the response rate varies greatly between tumour types: up to 40% in melanoma, 25% in non-small cell lung cancer, but <10% in most other tumour types.
The low response rate of ICB therapy was probably ascribed to defects in T eff cell homing to the tumour sites. 6 In addition, tumour microenvironment (TME), comprised of immune and stromal cells, vasculature, extracellular matrix, cytokines, chemokines, and growth factors, can all in uence tumour response to ICB therapies. Other mechanisms of ICB resistance include the presence of regulatory T (T reg ) cells, myeloidderived suppressor cells (MDSCs), and M2 tumour-associated macrophages (TAMs) at the immunosuppressive TME. 7 Elevated level of CCL5, CCL17, CCL22, CXCL8 and CXCL12 facilitates the recruitment of T regs and MDSCs to the TME, resulting in a diminished ICB response. In contrast, CXCL9 and CXCL10 promote homing of cytotoxic T-cells to the tumour sites, boosting anti-tumour immune response; transforming growth factor beta (TGF-b) does the opposite and upregulates T regs . Oncogenic or tumour suppressor pathways, such as mitogen-activated protein kinase (MAPK) and PI3K-g in the cancer cells can also in uence TME by altering the immune cell compositions and cytokine pro le. Toll-like receptors (TLRs) 7/8 agonists have been shown to be able to reprogram the TME and enhance the function of immune cells. 8,9 In an attempt to overcome ICB resistance, many combination therapeutic strategies have been tried preclinically and clinically. 7 These include the addition of the following drugs to a ICB-Ab: one other ICB-Ab (antibodies against CTLA-4, PD-1, PD-L1, LAG-3 and TIM-3), chemotherapeutic agents (paclitaxel, gemcitabine and carboplatin), radiation therapy, targeted therapy (inhibitors against PI3K, VEGF, BRAF/MEK, IDO, A2AR, FGFR, EGFR, PARP and mTOR), macrophage inhibitors (inhibitors against CSF1R and ARG1), cytokine/chemokine inhibitors (inhibitors against CXCR4, CXCR2 and TGF-b), epigenetic modulators (histone deacetylase inhibitors and hypomethylating agents), immunomodulatory agents (antibodies against OX40, 41BB, GITR, CD40 and ICOS), adoptive cell transfer therapy (car T, TIL and TCR), and modulation of gut microbiome. The development of a therapeutic modality that can reliably alter the complex tumour immune microenvironment favoring tumour regression is a big challenge, preclinically and clinically.
Advancement and optimization of nano-immunotherapy lie in the development of innovative approaches to enhance the speci city and controllability of immunotherapeutic interventions, and the targeting of desired cell types. Advanced bionanomaterials or approaches in a more controlled manner could enhance immunotherapeutic potency by increasing the accumulation and prolonging the retention of immunomodulatory and immune cell homing agents at the TME while sparing the normal tissues and organs, thus reducing off-target adverse effects such as systemic cytokine storm. [10][11][12][13][14][15] Especially, in situ modulation of nanomaterial in vivo has been demonstrated to improve the performance of bioactive molecules. [16][17][18][19][20][21] In order to improve ICB therapy, here we report on a programmable tumour cells/T eff cells bispeci c nanoimmuno-engager (NIE) that can enhance the homing of T eff cells to the tumour sites and sustainably release immunoagonist to reprogram the TME. This bispeci c NIE, initially in nanoparticle form (NIE-NPs), is self-assembled from two transformable peptide monomers (TPMs, Fig. 1a). TPM1, LXY30-KLVFFK(Pa), was comprised of three discrete functional domains: (1) the high-a nity and high-speci city LXY30 cyclic peptide (cdG-Phe(3,5-diF)-G-Hyp-NcR) ligand that targets α 3 β 1 integrin heterodimeric transmembrane receptor expressed by many epithelial tumours, 22 (2) the KLVFF b-sheet forming peptide domain originated from β-amyloid (Aβ) peptide, 16, 23-25 and (3) the pheophorbide a (Pa) moiety with uorescence property, serving as a hydrophobic core to induce the formation of micellar nanoparticles. TPM2, proLLP2A-KLVFFK(R848), was also comprised of three discrete functional domains: (1) proLLP2A, the "pro-ligand" version of LLP2A ligand that is a high-a nity and high-speci city peptidomimetic ligand against activated α 4 β 1 integrin of lymphocytes, 26 (2) the same KLVFF b-sheet forming peptide domain, and (3) R848 (resiquimod), a hydrophobic TLRs 7/8 agonist, grafted to TPM2 main chain via an esterbond. In proLLP2A, the carboxyl group of LLP2A is blocked by 3-methoxy-1-propanol through esteri cation such that it will not interact with normal lymphocytes and mesenchymal stem cells during blood circulation.
Under aqueous condition and in blood circulation, TPM1 and TPM2, at a ratio of 1:1, would co-selfassemble into nanoparticles, NIE-NP, in which KLVFFK(Pa) and KLVFFK(R848) hydrophobic domains were in the interior of the NIE-NPs, while relatively hydrophilic LXY30 and proLLP2A ligand peptides were on the surface of the NIE-NPs. NIE-NPs would be preferentially taken up by tumours through the leaky tumour vasculatures (enhanced permeability retention effects). Upon interaction with α 3 β 1 integrin receptor protein displayed on the tumour cell membrane, the NIE-NPs would undergo in situ transformation into nano brillar structural network (NIE-NFs) on the surface of tumour cells, thus maintaining a prolonged retention of the nano brillar network of NIE at the tumour sites (at least 7 days).
With the abundant esterase in the TME and on the tumour cells, proLLP2A would quickly be converted to LLP2A (T cell ligand) against activated a 4 b 1 integrin. LLP2A displayed on the brils would facilitate the homing and retention of T eff cells (e.g. CD8 + T cells) at the TME and adjacent to the tumour cells. The homing and retention of large numbers of T eff cells would greatly enhance ICB therapy. Anti-PD-1 antibody would greatly activate the NIE homed cytotoxic T cells for ICB therapy. 15,[27][28][29] Besides, the sustained release of R848 from the nano brillar network would improve immunosuppressive tumour microenvironment, e.g., activate antigen-presenting cells, promote immune cells to produce anti-tumour response factors, and re-educate the phenotype of macrophage from M2 to M1. 8 This in vivo structural transformation-based supramolecular bispeci c NIE represents an innovative class of programmable receptor-mediated targeted immunotherapeutics against cancers via enhancing T cell homing to the tumours and improving the TME (Scheme 1).

Results
Self-assembly, brillar transformation and esterase hydrolysis of programmable bispeci c NIE Two transformable peptide monomers (TPM1 and TPM2) were synthesized and characterized ( Fig.  1a and Supplementary Fig. 1). As the proportion of water in the mixed solvent (water and DMSO) of the TPM1 and TPM2 mixture solution (the ratio of 1:1) was increased, there was a gradual decrease in uorescence peak at 675 nm due to the aggregation-caused quenching (ACQ) properties of Pa dye (Fig.  1b), re ecting the gradual formation of NIE-NPs via self-assembly. Concomitantly, there was a modest decrease in the absorption peak at both 405 and 680 nm ( Supplementary Fig. 2a). Nanoparticles were analyzed by transmission electron microscopy (TEM) and dynamic light scattering (DLS). TPM1 and TPM2 each alone were able to self-assemble to form nanoparticles (NPs TPM1 and NPs TPM2 ) at 18 and 55 nm, respectively. NIE-NPs, assembled from 1:1 mix of TPM1 and TPM2 into nanoparticles at around 28 nm, which fell between the sizes of NPs TPM1 and NPs TPM2 (Supplementary Fig. 2b). The critical aggregation concentration (CAC) of NIE-NPs was determined to be 8 µM ( Supplementary Fig. 2c). We have also demonstrated that NIE-NPs could maintain good serum stability and proteolytic stability over 7 days at 37 o C ( Supplementary Fig. 2d).
To verify the receptor-mediated brillar transformable process of NIE-NPs in vitro, soluble α 3 β 1 integrin protein (receptor for LXY30) was added to NIE-NPs solution. After 24 h of incubation at room temperature, a brillar network (NIE-NFs, width diameter about 8 nm) with a broad size distribution was clearly detected (Fig. 1c,f). No transformation was observed in the NIE-NPs preparation without the addition of α 3 β 1 integrin protein, even after 24 h (Supplementary Fig. 2e). The CAC of NIE-NFs was determined to be 5 µM, which was lower than that of NIE-NPs (8 µM), indicating that NIE-NFs has higher propensity to form nanostructures than NIE-NPs ( Supplementary Fig. 2f). The process of NIE-NPs transformation could be monitored by the uorescent intensity of Pa (Fig. 1d). Addition of α 3 β 1 integrin protein triggered a gradual decrease in uorescence intensity of Pa over time, re ecting an increase in uorescence quenching in NFs1, indicating that the packing of Pa in NIE-NFs was denser. In addition, a 5 nm uorescence red shift (from 675 nm to 680 nm) with uorescence peak reversal (from 680 nm to 725 nm) were observed, indicating that the disordered arrangement of Pa in the core of NIE-NPs was transformed into J-aggregate form in NIE-NFs. [30][31][32] We have also investigated the responsiveness of proLLP2A displayed on the NIE-NPs surface to soluble α 4 β 1 integrin protein in the presence and absence of esterase (Fig. 1e,f and Supplementary Fig. 3). Soluble α 4 β 1 integrin protein alone was not able to alter the structure of NIE-NPs displaying proLLP2A, even after 24 h of incubation. In contrast, successive addition of esterase, followed by soluble α 4 β 1 integrin protein was able to elicit conversion of NIE-NPs to brillar network after 24 h of incubation. This expected result con rmed that esterase was able to convert pro-ligand proLLP2A to ligand LLP2A, which in turn was able to trigger receptor-mediated transformation of NIE-NPs to NIE-NFs. We also monitored the conversion of proLLP2A ligand to LLP2A ligand by HPLC. As shown in Supplementary Fig. 4, majority of proLLP2A were found to be converted to LLP2A after incubation with porcine liver esterase for 8 h at pH 7.4 and 37 o C.
Circular dichroism (CD) spectroscopic analysis of the transformation process of NIE-NPs showed a gradual progression of a negative signal at 216 nm and a positive signal at 195 nm upon incubation with α 3 β 1 integrin protein or combination esterase/α 4 β 1 integrin protein, indicative of β-sheet formation ( Fig.   1g) and consistent with TEM results shown in Fig. 1c,e. 23,33 In vitro release behaviour of R848 from NIE-NFs was studied at pH 6.5 with addition of esterase to simulate TME condition. As shown in Fig. 1h, about 45% of R848 was released in the rst 24 h, after which the release rate gradually slowed down and about 86% cumulative release was observed by 168 h, indicating that prolonged and sustained release of R848 could occur at the TME, which may be due to the slow hydrolysis of the ester bond to R848 within the brillar structure. To demonstrate the unique transformable property of NIE-NPs, we designed a related negative control nano-immuno-engager nanoparticle (CNIE-NP) formed by assembly of two negative control TPMs without b-sheet forming KLVFF peptide sequence, at a ratio of 1:1 (CTPM3: LXY30-KAAGGK(Pa) and CTPM4: proLLP2A-KAAGGK(R848), Supplementary Fig. 1). As expected, α 3 β 1 integrin protein was unable to transform CNIE-NPs to brillar structures even after 24 h, indicating that the b-sheet peptide was required for the transformation of NIE-NPs to NIE-NFs ( Supplementary Fig. 5a,b). CD spectroscopic analysis of CNIE-NPs remained the same with or without the addition of α 3 β 1 integrin protein, indicating that β-sheet was not formed (Supplementary Fig. 5c).

In vitro evaluation of bispeci c NIE facilitating T eff cells homing and reeducating macrophages
We investigated the binding a nity and selectivity of LXY30 to a 3 b 1 integrin and LLP2A to a 4 b 1 integrin via ow cytometry. To ensure the speci city and accuracy of binding analysis, K562 human myeloid leukemia cell line stably transfected with a 3 b 1 or a 4 b 1 integrins were used. As shown in Supplementary   Fig. 6a, biotinylated LXY30 ligand displayed little a nity for K562 (a 4 b 1 +) cells, whereas biotinylated LLP2A ligand showed high a nity for these cells. As expected, proLLP2A ligand showed no signi cant a nity to these a 4 b 1 integrin expressing cells, but the a nity was restored after proLLP2A ligand was treated with esterase ( Supplementary Fig. 6b). We also analyzed the binding a nity of LXY30 and LLP2A ligand to K562 (a 3 b 1 +) cells. As shown in Supplementary Fig. 6c, LXY30, but not LLP2A, was found to have high a nity against these a 3 b 1 integrins expressing cells. To further characterize the interaction between NIE-NPs and α 3 β 1 integrin receptors on the surface of living cells, we chose α 3 β 1 integrin expressing 4T1 murine breast cancer cell. As expected, ow cytometry analysis con rmed that LXY30, the high a nity α 3 β 1 integrin ligand, did bind to 4T1 tumour cells ( Supplementary Fig. 7). We have also found that NIE-NPs was slightly cytotoxic against 4T1 tumour cells, with 84.5% cell viability at 50 mM ( Supplementary Fig. 8). We investigated the distribution of nanoparticles by tracking the red uorescent signal emitted by Pa using confocal laser scanning microscopy (CLSM). Six hours after incubation of 4T1 tumour cells with NIE-NPs, a strong red uorescence signal was observed on the cell surface and its vicinity but not inside the cells (Fig. 2a). In contrast, the uorescent signal of Pa in CNIE-NPs-treated group was found to be concentrated primarily in the cytoplasm of the cells. To study the retention and stability of formed nano brillar network on the surface of tumour cells, unbound nanoparticles were washed off after 6 h of incubation and fresh medium without nanoparticles was added to incubate cells for another 18 h. As expected, NIE-NPs treated cells still retained strong red uorescence signals on the cell surface at 24 h, displaying that the capacity of prolonged retention of bispeci c NIE (Fig. 2b). In sharp contrast, only weak uorescence signal was observed inside the cells treated with CNIE-NPs after 24 h. This is probably due to the enzymatic degradation of the already endocytosed CNIE-NPs after 18 h of incubation, but without any new endocytic uptake during that time period. TEM images con rmed the presence of nano brillar network on the surface of, and between 4T1 tumour cells after incubation with NIE-NPs for 24 h, but absence of such nano brillar structures on cells treated with CNIE-NPs (Fig. 2c). The brillar structures further away from the tumour cell surface were probably induced by the secreted tumour exosomes displaying α 3 β 1 integrin proteins. 16, 34 We also investigated the effect of esterase on the interactions between NIE-NPs and T-cell surface a 4 b 1 integrin, after converting pro-ligand pro-LLP2A to LLP2A displayed on the surface of NIE-NPs. Live GFP transfected Jurkat T-lymphoid leukemia cells with high expression level of constitutively activated a 4 b 1 integrin protein were used to mimic T cells. As shown in Fig. 2d, after 6 h incubation of Jurkat cells with NIE-NPs (pre-treated with esterase), a luxuriant red uorescent layer was found surrounding the Jurkat cells, indicating that the conversion of pro-ligand to LLP2A ligand was successful. Scanning electron microscopy (SEM) con rmed the presence of brillar network on the surface of NIE-NPs-treated 4T1 tumour cells and esterase pre-treated NIE-NPs-treated Jurkat cells (Fig. 2e).
To simulate the processes of initial brillar transformation of NIE-NPs on the 4T1 tumour cells surface followed by T cell homing, we incubated 4T1 tumour cells with NIE-NPs for 6 h, unbound NIE-NPs were then washed off, followed by addition of fresh medium containing esterase but without NIE-NPs. After 1 h of incubation, Jurkat cells were added and incubated with 4T1 tumour cells for 2 or 4 h. After that, unbound Jurkat cells were gently removed prior to CLSM imaging (Fig. 2f). As expected, a brillar structure layer with red uorescence was detected surrounding 4T1 tumour cells surface, and Jurkat cells (GFP+) were found to interact with the red uorescent brillar network (bispeci c NIE) and in close proximity to 4T1 tumour cells, after 2 h of incubation. As the incubation time was increased to 4 h, many more Jurkat cells were found around the 4T1 tumour cells, which was consistent with our notion that nano brillar-based bispeci c NIE would facilitate the homing of immune cells. SEM imaging provided critical evidence that the bispeci c NIE had played a signi cant role in direct physical contact between 4T1 tumour cells and Jurkat cells through nano brillar network (Fig. 2g). In addition to using Jurkat cells as the surrogate T-cells, we also examined the effects of NIE on the killing of 4T1 tumour cells by activated murine CD8 T cells (a 4 b 1 integrins pre-activated by Mn 2+ ). As shown in Supplementary Fig. 9, NIE-NP transformed into nano brillar network on the tumour cell surface, and as a result of esterase treatment, the displayed LLP2A captured activated T-cells, leading to about 50% tumour cell kill (Supplementary Fig. 10), con rming that the nano brillar network did not dampen immune synapse formation and killing e ciency of CD8 T cells. Similar studies were also performed on the negative control CNIE-NPs, which were taken up by 4T1 tumour cells, instead of forming nano brillar network on the cell surface, and as expected, no bound Jurkat cells were detected ( Supplementary Fig. 11).
The conversion of TAMs from an immunosuppressive M2-polarized phenotype to an anti-tumourigenic M1-polarized phenotype is one of the major immunotherapeutic strategies for reprograming the immunosuppressive TME. Macrophage polarization states demonstrate hallmark morphology, e.g., elongated projections for M2-like cells as opposed to a round and attened morphology for M1-like cells ( Supplementary Fig. 12). IL-4 has been used to induce bone marrow derived macrophages (BMDM) to M2-polarized macrophages, as re ected by the increase in expression level of the metabolic checkpoint enzyme arginase-1 (Arg1) and mannose receptor-1 (Mrc1). R848 has been reported to be a powerful driver of the M1-phenotypes in vitro, resulting in elevated level of interleukin 12 (IL-12) and nitric oxide synthase (Nos2) produced by these cells. 8 Here we investigated the possibility of using NIE-NFs to reeducate macrophages from M2 phenotype to M1 phenotype. Not unexpected, incubation of M2-state macrophages with NIE-NFs, preformed from NIE-NPs with soluble a 3 b 1 integrin protein, did not have signi cant change in morphology and expression level of Arg1 and Mrc1 even after 12 h (Fig. 2h), which can be explained by the lack of R848 released from NIE-NFs. In contrast, addition of esterase to the culture medium followed by 12 h incubation resulted in morphological change of M2-state macrophages towards M1-state, a decrease in Arg1 and Mrc1, and an increase in IL-12 and Nos2 expression as measured by qPCR. These changes were even more pronounced after 24 h, at which time the macrophages were completely transformed to a round and attened morphology, with further decrease in Arg1 and Mrc1, and increase in IL-12 and Nos2 expression. We believe the ability of bispeci c NIE to anchor on the surface of tumour cells, affording the sustained release of R848 from the brillar network, would generate a durable anti-cancer immunoactive TME. Interestingly, we have also found a signi cant change in macrophage morphology from M2-state towards M1-state after 24 h treatment with CNIE-NPs, which was perhaps because the particulate CNIE-NPs were phagocytosed by M2-macrophages, releasing R848 intracellularly, resulting in conversion of these macrophages from M2 to M1 phenotype ( Supplementary Fig. 13).
In vivo evaluation of bispeci c NIE targeting tumour cells and nano brillar-transformation to facilitate T eff cell homing and activate immunity NIE-NPs was found to be non-toxic: blood counts, platelets, creatinine and liver function tests obtained from normal Balb/c mice treated with eight consecutive q.o.d. intravenous (i.v.) doses of NIE-NPs (13 mg/kg) were within normal limits (Supplementary Figs. 14 and 15). In vivo blood pharmacokinetics (PK) studies in rats showed that NIE-NPs possessed a long circulation time (T-half (α): 2.866 h and T-half (β): 23.186 h), indicating its stability during circulation ( Supplementary Fig. 16). For biodistribution studies, NIE-NPs were tail vein injected once (13 mg/kg) into Balb/c mice bearing syngeneic orthotopic 4T1 breast cancer; 10, 24, 48, 72, 120 and 168 h later, tumour and main organs were excised for ex vivo uorescent imaging (Fig. 3a,b). Signi cant uorescent signal of Pa was found to persist in tumour tissue for over 168 h, while uorescent signal in normal organs began to decline after 10 h and was almost undetectable in the main organs at 72 h. In sharp contrast, uorescent signal of Pa at tumour tissue treated by CNIE-NPs was found to gradually decline over time after peaking at 24 h (Fig. 3c). By 168 h, less than 2.91% of the peak uorescent signal for CNIE-NPs remained in the tumour, whereas for NIE-NPs, over 60.49% signal remained in the tumour (Fig. 3d). We also found that the red uorescence signals from pheophorbide a (Pa) in NIE-NPs were detected throughout the entire tumour tissue, not just in the tumour periphery ( Supplementary Fig. 17). Most of the red uorescence (Pa) was observed in the extracellular space in close proximity to the tumour cell membranes, as indicated by the green arrows. Prolonged retention of uorescent signal in NIE-NPs-treated mice could be attributed to in situ receptor-mediated transformation of NIE-NPs into NIE-NFs at the TME. TEM studies on excised tumour sections, 72 h after i.v. administration, showed abundant bundles of nano brils in the extracellular matrix while no such nano brils were observed in negative CNIE-NPs-treated and untreated mice (Fig. 3e). Fluorescent micrographs of tumour and overlying skin revealed intense uorescent signal in tumour region but negligible signal in normal skin (Fig. 3f). This is consistent with our notion that (1) NIE-NPs would leak into the TME through leaky tumour vasculatures (EPR effect), followed by interaction with a 3 b 1 integrin on tumour cells and tumour associated exosomes to generate NIE-NFs, and (2) blood vessels are not leaky in normal skin. The tumour tissue distribution of R848 over time was also determined with high pressure liquid chromatography-mass spectroscopy (HPLC-MS, Fig. 3g). We found R848 uptake by tumour was quite high at 24 h (3.62 μg per g tissue), and about one-third of R848 was found to retain at the tumour site (1.14 μg per g tissue) even at 7 days after injection of NIE-NPs. Although CNIE-NPs could also deliver signi cant amount of R848 to the tumour site (83% of what NIE-NPs could deliver), retention of R848 at the tumour site was much lower than that of NIE-NPs. Prolonged retention and release of R848 in tumour site indicates that a sustained immune-active TME could be achieved with NIE.
To evaluate if the bispeci c NIE displaying LLP2A and R848 at the TME could facilitate T eff cells homing to the tumour sites and reeducate TAMs in vivo, tumours from NIE-NPs-treated mice were excised on day 15 after a single i.v. injection of NIE-NPs (13 mg/kg), and the immune cell populations within the tumours were analyzed by ow cytometry, immunohistochemistry (IHC) and qPCR. Experiment using CNIE-NPs as an untransformable/endocytic negative control group was also performed at the same time. We found that tail-vein injection of NIE-NPs resulted in greatly increase concentration ratio of cytotoxic T cells and a sustained immunoactive TME. First, NIE-NPs was found to signi cantly stimulate the production of chemokine CXCL10 at the tumour site (Fig. 3h), which was known to facilitate T eff cells recruitment. 35,36 This should be attributed to R848 which is known to induce the production of chemokines. 37,38 R848 covalently linked to NIE-NPs via an ester bond was released from the nano brillar network anchored at the TME by endogenous esterases. As a result, released R848 would interact with a variety of cells in the TME, such as tumour cells, endothelial cells, and some immune cell subsets (dendritic cells, monocytes, macrophages, and B cells) to produce a range of chemokines and cytokines.
Indeed, we observed that the proportion of CD45 + CD3 + and CD45 + CD3 + CD8 + T cells in the NIE-NPstreated tumour tissue was substantially higher than those from mice treated with endocytic CNIE-NPs or saline (Fig. 3i, Supplementary Fig. 18 and 19a). More speci cally, the percentage of CD3 + CD8 + T effector cells in tumours was found to be 18 and 4-fold increase, relative to that of saline and CNIE-NPs-treated mice, respectively.
Second, we found that the relative abundance of CD4 + Foxp3 + T regs at the tumour site was substantially lower in mice that received NIE-NPs treatment than those in mice treated with CNIE-NPs, i.e. (4.97% versus 13.0%) or saline (4.97% versus 14.6%). The ratio of tumour-in ltrating CD8 + killer T cells to immunosuppressive T regs (CD3 + CD4 + Foxp3 + ), which could be an indicator of anti-tumour immune balance, was found to be the highest in NIE-NPs treated group. We also analyzed the expression (CD49 + ) and activation (CD29 + ) of a 4 b 1 integrin in tumour-in ltrating and blood circulating CD8 and CD4 T cells ( Fig. 3i and Supplementary Fig. 19b,c). Although both blood circulating and tumour in ltrating CD8 and CD4 T cells in mice bearing 4T1 tumour were found to be both CD49+ and CD29+, their expression levels were higher in the tumour in ltrating cell populations. This may be because there was more chemokine and cytokine secretion in tumour tissues, therefore inducing more a 4 b 1 integrin expression and activation.
More excitingly, we found that after NIE-NPs treatment, there were more CD8 + CD49 + T cells and CD8 + CD29 + T cells in both circulating and tumour in ltrating cell populations. This data indicates that NIE-NP treatment, probably via chemokine induction, could promote CD49 expression and CD29 activation of T cells, and therefore homing and retention of these cells by the displayed LLP2A at the TME. We also evaluated T cell activation and exhaustion makers (CD69 and LAG-3) in tumour tissues, which could better access the killing effect of T cells against 4T1 tumour cells (Supplementary Fig. 20). Compared to CNIE-NPs and saline treated groups, mice treated with NIE-NPs exhibited higher CD69 expression and lower LAG-3 (an immune check point receptor) expression, the desirable anti-tumour immune phenotype of the TME. IHC staining of tumour tissue sections also con rmed an increase in CD8/CD4 and decrease in Foxp3 positive T reg cells (Fig. 3j).
Third, IHC staining of tumour sections demonstrated an increase in M1-polarized macrophage marker CD68 and a decrease in M2-polarized macrophage marker CD163 in the NIE-NPs treated group, compared to the tumour tissue excised from mice treated by CNIE-NPs. This could be explained by the sustained release of R848 at the tumour site, causing the phenotypic reeducation of TAMs. Fourth, gene expression level of cellular immune related markers (IFN-g, TGF-b) and macrophage markers (IL-12, IL-10, Nos2 and Arg-1) were also evaluated by qPCR. As shown in Fig. 3k, the high expression level of IFN-γ and low expression level of TGF-b in the tumour tissue con rmed that a strong tumour-speci c immune response had been elicited. Furthermore, the secretion of IL-12 and Nos2 was found to be signi cantly upregulated, while the secretion of IL-10 and Arg-1 was signi cantly down-regulated, indicating a signi cant phenotypic conversion of TAMs from M2 state to M1 state, with NIE-NPs treatment, but not CNIE-NPs treatment nor saline control. To elucidate the mechanism of immunotherapeutic effects induced by bispeci c NIE, we collected the tumour tissues and thoroughly evaluate the pro le of tumour-in ltration immune cells via ow cytometry analysis ( Supplementary Fig. 21). We found that among the six treatment regimens, NIE-NPs (regimen 6) was the most potent inducer for increase in cell numbers of the following cell populations: CD45+, T cells, macrophages, dendritic cells, neutrophiles, B cells and monocytes. Compared to saline control, NK cells and eosinophil cells numbers were also increased with NIE-NPs treatment, but regimens 4 and 5 were equally effective in increasing these cell populations. Additional analysis (Fig. 4e and Supplementary Fig.   22) showed that only the treatment regimens capable of in situ brillar transformation and LLP2A display (regimen 5 and 6) were able to greatly increase the frequency of CD3 + and CD8 + T cells within the tumours, particularly in combination with immune adjuvant R848 (regimen 6, NIE-NPs), which was consistent with the observed strongest anti-tumour effects in NIE-NPs. The percentage of M1-phenotype macrophage in total macrophages was 60%, which was much higher than that of other control groups ( Supplementary Fig. 23). Tumour sections (H&E) obtained from mice treated with NIE-NPs revealed a marked decrease in Ki-67 expression, an increase in CD8 + T cells, and a decrease in Foxp3 (T reg cells), compared with other control groups ( Supplementary Fig. 24). There was an increase in CD68 and a decrease in CD163, indicating that the phenotype of macrophages was reversed after 8 doses of NIE-NPs.
It is known that CD8 + T cells secrete cytokines IFN-g and TNF-a to kill tumour cells. 15,39 The expression levels of IFN-g and TNF-a in the tumour tissue were further evaluated by qPCR. As shown in Fig. 4f, treatment regimen 6 (NIE-NPs) was the most e cacious in restoring the immunoactive state of the tumour microenvironment, with the highest expression levels of IFN-g and TNF-a. In addition, NIE-NPs also signi cantly induced expression of IL-12, IL-6 and Nos2, and suppressed expression of TGF-b, IL-10 and Arg-1, leading to the suppression of the T reg cells recruitment and reeducation of M2-like macrophages to M1 phenotype. To better clarify the contribution of CD8 T cell recruitment to the tumour in NIE-NPs immunotherapy, we performed an CD8 T cell depletion experiment with an antibody (Fig. 4g).
Syngeneic orthotopic 4T1 breast cancer-bearing mice were randomly divided into three groups: (1) saline; (2) regimen 6 (NIE-NPs) plus anti-CD8; (3) regimen 6 (NIE-NPs). Anti-CD8 depletion antibody (i.p. injection) was given at an initial dose of 200 μg 3 days before treatment and then given four doses on day 1, 5, 9 and 13. When tumour volume reached about 100 mm 3 , NIE-NPs were injected via tail vein consecutively eight times q.o.d. (13 mg/kg each dose) and the mice were continuously observed for 21 days. As expected, the therapeutic e cacy of NIE-NPs was greatly diminished in mice given CD8 T celldepleting antibody (lesser tumour inhibition and shorter survival), con rming the importance of tumour homing of CD8 T cells in the effectiveness of NIE-NP immunotherapy.

Bispeci c NIE enhances ICB therapy against breast cancer and lung cancer
Since the receptor-mediated bispeci c NIE could signi cantly mount systemic anti-tumour response by facilitating T eff cells homing and reprogramming of TME, we believe it could enhance the e cacy of ICB therapy. It is well known that tumour cells hijack PD-1 receptors of T cells by overexpression of PD-L1, which can activate PD-1, leading to inhibition of T cell proliferation, activation, cytokine production, altered metabolism and cytotoxic T lymphocytes killer functions, and eventual death of activated T cells.
Most remarkably, mice treated with regimen 6 (NIE-NPs) plus anti-PD-1 resulted in gradual shrinkage and eventual complete elimination of tumours within 21 days, and without any sign of recurrence during the observation period of 90 days (Fig. 5c), validating the synergistic effects of our bispeci c NIE with ICB therapy. In addition to i.v. administration, we have also investigated the therapeutic e cacy of NIE-NPs via intratumoural (i.t.) injection plus anti-PD-1 antibody given i.p. As shown in Fig. 5 and Supplementary  Fig. 25, both administration routes for NIE-NPs (i.v. or i.t.) yielded excellent immunotherapeutic e cacy. However, regimen 4 (i.t.) plus anti-PD-1 and regimen 5 (i.t.) plus anti-PD-1 groups were still unable to eliminate the tumour after ve cycles of treatment, while i.t. injection of regimen 6 (NIE-NPs) plus anti-PD-1 group eventually eliminated tumour and maintained 100% survival rate within 90 days.
Unlike traditional chemotherapy or targeted therapy in clinical oncology, immunotherapy can potentially induce an adaptive response with capacity for memory. 40,41 Immune memory is crucial to achieving durable tumour responses and preventing recurrence. To assess whether the synergistic therapy of bispeci c NIE plus anti-PD-1 Ab could induce an immune memory response, we re-challenged the "cured" mice from previous experiment (regimen 6 plus anti-PD-1 treatment, Fig. 5a-c) with 4T1 tumour cells on the opposite mammary fat pad on day 90; naive mice of the same age were used as a negative control (Fig. 5d). The tumour volume of all the naïve mice increased rapidly within 30 days (Fig. 5e). However, either no tumour growth or signi cant delay in tumour growth was observed in mice previously treated successfully with NIE-NPs plus anti-PD-1 (Fig. 5f), con rming the presence of an excellent immune memory response mounted by these previously treated mice. Survival curves of this experimental group correlated well with tumour growth results (Fig. 5g). All mice remained alive during the 60-day observation period (day 90-150). In addition, the serum levels of cytokines such as TNF-a and IFN-g in this experimental group were found to be much higher than those in the control same age naïve mice group after re-challenged with 4T1 tumour cells for 6 days (Fig. 5h,i). These results suggest that a durable and robust T cell memory response was generated by regimen 6 (NIE-NPs) plus anti-PD-1 given previously.
Although the up to 40% response rate of ICB therapy in non-small cell lung cancer (NSCLC) is considered a clinical breakthrough, the overall survival of patients remains only 15% at 5 years and declines to less than 2% in patients with metastatic disease. 42 We speculated that like 4T1 breast cancer model, NIE would also synergize ICB treatment of NSCLC. As expected, complete tumour regression and prolonged survival was obtained for the treatment of Lewis lung syngeneic subcutaneous murine tumour model using NIE-NPs plus anti-PD-1 (5 cycles , Fig 5j-l). No systemic toxicity and weight loss were detected.

Discussion
In spite of the clinical success of ICB therapy, only a fraction of cancer patients bene ts from it. Defects in T eff cells homing to the tumour sites is probably one of the main reasons why many patients remain refractory to such treatment. 6 Development of approaches to convert an immunologically "cold" tumour to a "hot" tumour is undergoing intense investigation around the world. We believe the receptor-mediated bispeci c NIE reported here can provide a relatively simple solution to this challenge. By incorporating pro-ligand LLP2A and R848 to the in vivo transformable nano brillar networks, we have already demonstrated in syngeneic 4T1 breast cancer and Lewis lung cancer model that this non-toxic treatment can (1) facilitate the homing of T-cells to the tumour sites, (2) promote retention of T-cells at close proximity to the tumour cells, and (3) provide sustained release of R848 at the TME, resulting in the reeducation of TAMs to M1 phenotype in vivo, signi cantly enhancing the e cacy of anti-PD1 antibody based ICB. We have also clari ed the role of T-cell capturing ligand (LLP2A) and TLR7/8 immunoagonist (R848) in NIE-NPs, in chemokine secretion and T cells homing to the TME. As shown in Supplementary   Fig. 26a, R848 (group ii) was more effective in inducing chemokine secretion than LLP2A ligand (group iii), but the population of CD8 T cells in group iii was slightly higher than that in group ii, and the ratio of CD8 to T reg cells was also higher (Supplementary Fig. 26b). This suggested that LLP2A ligands played a signi cant role in capturing and retaining CD8 T cells at the TME, more so than R848. However, LLP2A or R848, each alone was found to be insu cient to induce a robust anti-tumour immune response. It is clear that group iv, with both LLP2A and R848, was needed to mount a robust anti-tumour immune response, resulting in higher level of chemokine expression, more CD8 T cells and less T reg cells, when compared to either LLP2A or R848 alone.
To demonstrate that the receptor-mediated bispeci c NIE not only works well in murine tumours but also human tumours, we used MCF-7 breast cancer and A549 lung cancer cell lines and human primary T cells to con rm the formation of nano brillar networks and capturing of human CD8 T-cells around these target tumour cells (Supplementary Fig. 27), indicating that the NIE strategy has great translational potential. Although NIE was found to be equally e cacious whether it was given i.t. or i.v. in 4T1 syngeneic tumour model with single subcutaneous tumour ( Fig. 5 and Supplementary Fig. 25), i.t. administration is not as useful clinically since many patients do have multiple tumour lesions, many of which may not be easily accessible for i.t. injection. Because NIE is relatively non-toxic, intravenous administration is the preferred route, so that each and every tumour lesion will be treated and be converted from immunologically "cold" to "hot".
Since the programmable bispeci c NIE is modular, we have the options of combinatorially incorporating various different ligands, pro-ligands, or immunomodulators to generate a series of nano-engagers that may be applied for capturing other bene cial immune cells, including natural killer cells. Other potent immunomodulators against other pathways such as the stimulator of IFN genes (STING) pathway may also be tried. One unique feature of NIE therapy is that the nano brillar network formed at the TME is durable, which may explain its remarkable in vivo anti-tumour immune response and memory effects but without any sign of systemic immunotoxicity. This programmable bispeci c NIE is highly robust. Each transformable peptide monomer is chemically well-de ned, and the nal immune-nanoparticle can be simply assembled from two or more monomers under aqueous condition. Scale-up production for clinical development should not be a problem. Work is currently undergoing in our laboratory on using other ligands and immunomodulators against other cancers besides breast and lung.
First, NIE-NPs or NIE-NFs was diluted to different concentrations (0.01, 0.05, 0.1, 0.5, 1, 5, 10, 20, 30 and 50 µM), then 999 µL of NIE-NPs or NIE-NFs of each dilution was incubated with 1 µL of pyrene acetone solution (0.1 mM) at 37°C for 2 h. The uorescence spectra of pyrene (excitation, 335 nm) in different NIE-NPs or NIE-NFs dilutions were recorded. The uorescence intensity ratio (I 3 /I 1 ) of the third and rst emissive peaks was measured for CAC calculation. Stability of NIE-NPs in the presence of human plasma and protease. The stability of NIE-NPs was studied in 10 % (v/v) plasma from healthy human volunteers and protease solution. The mixture was incubated at physiological body temperature (37 °C) followed by size measurements at predetermined time intervals up to 168 h. Accumulated drug release of R848 from NIE-NFs was triggered by acidic pH 6.5 and porcine liver esterase (100 U mL −1 ). Considering the experimental design (R848 was released from NIE-NFs in the tumour microenvironment), we rstly prepared NIE-NP solution (100 µM), and then transformed into NIE-NFs after the interaction with α 3 β 1 integrin soluble protein. Then pH value in NIE-NF solution was set up to 6.5 and porcine liver esterase was added into NIE-NF solution. The mixture was incubated at 37 o C. At timed intervals, 20 µL solution was taken out and added into 80 µL DMSO for HPLC. Based on the preestablished calibration curves, free R848 concentration was calculated according to the integral area of free R848. Each value was reported as the means of the triplicate samples. Statistical analysis.
Data are presented as the mean ± standard deviation (SD). The comparison between groups was analyzed with the student's t-test (two-tailed). The level of signi cance was de ned at *p < 0.05, **p < 0.01 and ***p < 0.001. All statistical tests were two-sided. Figure 1 Assembly and brillar transformation of the programmable bispeci c NIE, as well as esterase-induced LLP2A ligands exposure and R848 release. a, Schematic illustration of the molecular structure and function of TPM1 (LXY30-KLVFFK(Pa)) and TPM2 (proLLP2A-KLVFFK(R848)  Experiments were repeated three times. g, Representative SEM images of 4T1 tumour and Jurkat cells after treatment with NIE-NPs (see f). Experiments were repeated three times. h, Representative images of M2-like murine macrophages and subsequent reeducation by NIE-NFs, NIE-NFs plus esterase, or R848 at different time points. Scale bar in h is 20 μm. Experiments were repeated three times. Statistical signi cance was calculated using a two-sided unpaired t test; *P < 0.05, **P < 0.01, ***P < 0.001.  e, Representative ow cytometric analysis images of CD3+CD8+ T cell within the 4T1 tumours excised from treated mice on day 21. f, The expression levels (analyzed by qPCR) of IFN-γ, TNF-α, IL12, IL6, TGFβ, IL10, Nos2 and Arg-1 in 4T1 tumours excised from mice on day 21 (data were mean ± s.d.). g, Observation of tumour growth inhibitory effect, weight change and cumulative survival of mice bearing orthotopic 4T1 tumour treated by regimen 6 plus anti-CD8 antibody (CD8 T cell depletion) and regimen 6 (n = 8 per group). Data are presented as mean ± s.d. Statistical signi cance was calculated using a twosided unpaired t test compared to regimen 6; *P < 0.05, **P < 0.01, ***P < 0.001.

Figure 5
Bispeci c NIE greatly enhances ICB therapy in mice bearing 4T1 breast tumour and Lewis lung tumour. a, Experimental design: orthotopic tumour inoculation and treatment protocol (4 treatment arms; regimens 4, 5 and 6 are the same as those shown in Fig. 4a). b, Tumour response in mice bearing orthotopic 4T1 tumour over 21 d of treatment (n = 8 per group). Data are presented as mean ± s.d. c, Cumulative survival of the four treatment groups. d, Experimental design: Mice previously treated with regimen 6 plus anti-PD-1 Ab were re-challenged with re-inoculation of 4T1 breast cancer cells on day 90. Same age naïve mice as a negative control group were carried out same operation. e, No anti-tumour immune memory effect was observed in same age naïve mice. f, Anti-tumour immune memory effect was observed in mice previously treated with regimen 6 and anti-PD-1 Ab. g, Cumulative survival of naïve mice and previously regimen 6 plus anti-PD-1 treated mice. h,i, IFN-γ (h) and TNF-α (i) level in mice sera 6 days after mice were rechallenged with 4T1 tumour cells. j,k, Observation of tumour inhibitory effect (j) and weight change (k) of mice bearing subcutaneous murine Lewis lung tumour over 21 d after initiation of treatment (n = 8 per group); Treatment protocol followed experiment design in a, 5 cycles (i.v. regimen 4-6 and i.p. anti-PD-1). Data are presented as mean ± s.d. l, Cumulative survival of different treatment groups of mice bearing subcutaneous murine Lewis lung tumours. Statistical signi cance was calculated using a two-sided unpaired t test; *P < 0.05, **P < 0.01, ***P < 0.001.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.