Nanoalbumin–prodrug conjugates prepared via a thiolation‐and‐conjugation method improve cancer chemotherapy and immune checkpoint blockade therapy by promoting CD8 + T‐cell infiltration

Abstract Protein–drug conjugates are emerging tools to combat cancers. Here, we adopted an indirect thiolation‐and‐conjugation method as a general strategy to prepare protein–drug conjugates. We found for the first time that this method led to the formation of nanometric conjugates, probably due to the formation of intermolecular disulfide bonds, which facilitated enhanced uptake by cancer cells. As a proof‐of‐concept application in cancer therapy, a nanometric albumin–doxorubicin prodrug conjugate (NanoAlb‐proDOX) was prepared. The nanometric size promoted its uptake by cancer cells, and the prodrug characteristic defined its selective cytotoxicity toward cancer cells in vitro and reduced side effects in vivo. In multiple tumor xenograft models, nanometric NanoAlb‐proDOX showed superior antitumor activity and synergy with immune checkpoint blockade, probably due to the synergistically enhanced tumor CD8+ T‐cell infiltration and activation. Hence, the thiolation‐and‐conjugation strategy may serve as a generally applicable method for preparing drug conjugates, and the proof‐of‐concept nanometric albumin–doxorubicin conjugate may be a good choice for antitumor therapy with the ability to co‐stimulate the efficacy of immune checkpoint blockade.

chemotherapy and immune checkpoint blockade, which may synergistically elevate the therapeutic efficacy. 13 However, the strategy still suffers from adverse side effects induced by chemotherapy.
One emerging strategy to reduce the side effects and to enhance the therapeutic efficacy of chemotherapy is to conjugate smallmolecule chemotherapy drugs to functional proteins to yield proteindrug conjugates, 14,15 such as antibody-drug conjugates (ADCs). [16][17][18] Protein-drug conjugates or antibody-conjugates usually consist of three components: a protein (antibody), a cytotoxic payload (drug), and a chemical linker that connects the two components. 19 Thus, to produce protein-drug conjugates, specific conjugation methods are essential. 20 Currently, many conjugation strategies are available.
Among them, site-specific labeling methods, due to their homogeneous production of protein-drug conjugates, have attracted increasing interest. 21,22 However, the necessity of genetic engineering of the proteins hinders their application to a certain extent. 23 In contrast, random labeling of reactive amino acid residues, such as lysine or cysteine, with the advantage that no genetic manipulations are needed, is more widely applied. 20,24 ADCs with linker and payloads directly conjugated to either lysine residues or cysteine residues have been approved. 24 In addition to direct conjugation of drug payloads to protein, indirect conjugation methods involve a first step of functionalization of the protein of interest with chemical molecules and a second step of conjugation on the functionalized groups. These methods are also widely used because some linker-payloads are not suitable for direct one-step conjugation to protein. 25 A typical example of these indirect conjugation approaches is the thiolation-and-conjugation method F I G U R E 1 Thiolation-and-conjugation do not alter the tumor targeting of the protein of interest but enhance cancer cellular uptake. (a) Schematic illustration of the synthesis of protein-drug conjugates via the thiolation-and-conjugation approach. POI, protein of interest. (b) SDS-PAGE analysis of albumin-sulfo-cy5 conjugate (HSA-TC-Sulfo-Cy5). HSA-Sulfo-Cy5 was prepared through direct conjugation of Sulfo-Cy5 to lysine residues on HSA. In gel Cy5 fluorescence indicated that HSA was successfully labeled with Sulfo-Cy5 using two different methods. CBB, Commassie brilliant blue staining; Cy5, Sulfo-Cy5 fluorescence. (c) In vivo and ex vivo imaging of the tumor targeting of HSA-TC-Sulfo-Cy5. HSA-Sulfo-Cy5 or HSA-TC-Sulfo-Cy5 was intravenously injected into mice bearing MDA-MB-231 subcutaneous tumors, and fluorescence imaging was performed 24 h post-injection using a living animal imaging system. No obvious differences in tumor accumulation were observed between HSA-Sulfo-Cy5 or HSA-TC-Sulfo-Cy5. (d) Enhanced cellular uptake of HSA-TC-Sulfo-Cy5 by cancer cell lines. Two cancer cells were incubated with the indicated concentrations of protein-dye conjugates for 18 h, and flow cytometry analysis showed that the uptake of HSA-TC-Sulfo-Cy5 was higher than that of HSA-Sulfo-Cy5 in both MDA-MB-231 and HeLa cells. The marked concentration represents the concentration of Sulfo-Cy5 added to the cell culture. The percentages of Sulfo-Cy5-positive cells were gated, and fold changes between the two conjugates are marked. Data are presented as the mean ± SEM. n = 3 technical replicates. ****p < 0.0001 enabled by the application of the bifunctional crosslinking molecule N-succinimidyl-3-(2-pyridyldithio) propionate (SPDP). 25,26 The protein of interest is first modified with sulfhydryl groups, and further conjugation of the protein with linker-payload is accomplished using maleimide-thiol-based chemistry. Although the SPDP-based method has been reported for the production of protein-protein conjugates, its application and versatility as a method to generate protein-drug conjugates for cancer therapy have not been extensively studied.
Whether this conjugation method interferes with the properties of the proteins, 27,28 especially in terms of the tumor targeting ability and internalization ability by cancer cells for application in cancer therapy, 29 should be well studied. Generally, a conjugation method that either improves the tumor targeting or enhances the uptake of the protein by cancer cells would be greatly preferred. 30 Herein, we applied this SPDP-based, indirect thiolation-andconjugation method to construct protein-drug conjugates ( Figure 1a) and proved its versatility in cancer therapy. Although this approach has been previously reported, we revealed that during the synthesis procedures, the conjugates self-assembled to form nanometric particles. The nanometric protein-drug conjugates preserved good tumor targeting ability and showed enhanced uptake by cancer cells in vitro. As proof-ofconcept application in cancer therapy, the nanometric protein-drug conjugate strategy was combined with an acid-sensitive prodrug strategy to produce an albumin and doxorubicin conjugate: NanoAlb-proDOX. This nanoconjugate showed preferential cytotoxicity to cancer cells over noncancer cells in vitro and reduced side effects in vivo. In multiple mouse xenograft models, NanoAlb-proDOX also showed superior antitumor activity over doxorubicin. In a combination therapy model, NanoAlb-proDOX synergistically elevated the efficacy of immune checkpoint blockade with enhanced infiltration of cytotoxic CD8 + T cells into the tumor microenvironment. Overall, the results showed that the thiolationand-conjugation approach can serve as a general method for constructing protein-drug conjugates that self-assemble into nanoparticles and that the proof-of-concept nanoalbumin-doxorubicin conjugate is a possible choice for cancer therapy.

| Preparation of protein-drug conjugates via a thiolation-and-conjugation method
To prepare protein-drug conjugates, a thiolation-and-conjugation approach was adopted ( Figure 1a). As illustrated, the protein of interest was first modified with SPDP on lysine residues and selectively reduced with dithiothreitol (DTT) to first install free sulfhydryl groups.
The reduction was performed under slightly acidic conditions that left the native intramolecular disulfide bonds unaffected (see detailed procedure in Section 5.2). Then, sulfhydrylated protein was conjugated with a drug of interest bearing a maleimide warhead. As a proof of concept, human serum albumin (HSA) and the fluorophore Sulfo-Cyanine 5 (Sulfo-Cy5) conjugate (HSA-TC-Sulfo-Cy5) were prepared and characterized with sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE, Figure 1b). HSA-Sulfo-Cy5 with direct Cy5 conjugation was also prepared for comparison. In gel Cy5 fluorescence indicated that HSA was modified by Sulfo-Cy5 with both methods (Figure 1b).
The effects of the thiolation-and-conjugation approach on the tumor targeting ability of HSA were first validated by comparison with HSA-Sulfo-Cy5. Breast cancer mouse xenograft models were constructed and utilized. As shown in Figure 1c, both HSA-Sulfo-Cy5 and HSA-TC-Sulfo-Cy5 showed good accumulation in tumors, showing that the thiolation-and-conjugation method did not affect the tumor targeting ability of the protein of interest and indicating that the thiolation-and-conjugation may be a feasible approach to prepare protein conjugates. HSA can be internalized by different cancer cells ( Figure S1), which facilitates its application in cancer therapy. Therefore, we further tested whether the thiolation-and-conjugation approach affects protein uptake by cancer cells. A flow cytometry assay was performed to quantify the internalization of both HSA-Sulfo-Cy5 and HSA-TC-Sulfo-Cy5 by two cancer cell lines. The time-and concentration-dependent uptake of the two HSA conjugates was validated. As shown in Figure 1d and Figure  , indicating that the conjugate prepared by the thiolation-and-conjugation approach may exhibit improved internalization by cancer cells, which is beneficial to tumor therapy. As another example, we also prepared albumin conjugates with nonsulfonated cyanine 5 (Cy5) using the two methods ( Figure S3A). In a similar uptake assay, HSA-TC-Cy5 showed enhanced uptake in HeLa cells compared with that of HSA-Cy5 after different incubation times (1.6-to 1.9-fold, calculated from the mean fluorescence intensity, Figure S3B,C). To prove that the enhanced cellular uptake was universal to other proteins, we further prepared another conjugate (RBD-TC-Sulfo-Cy5) using the recombinant receptor binding domain (RBD) of SARS-CoV-2 due to fast accessibility to this protein ( Figure S4). Flow cytometry analysis indicated that uptake of RBD-TC-Sulfo-Cy5 by HeLa cancer cells was also enhanced (2.3-to 24.6-fold in mean fluorescence intensity) compared to that of directly conjugated RBD-Sulfo-Cy5 ( Figure S4), indicating that the enhanced cellular uptake of the conjugate prepared via thiolation-and-conjugation may be universal. Albumin is reported to enter cancer cells through the gp60 receptor, and the RBD of SARS-CoV-2 enters cells through the ACE2 receptor; therefore, we speculated that the enhanced cellular uptake may be independent of their receptors. Overall, the above data showed that the thiolation-and-conjugation approach can be a universal strategy to generate protein-drug (small molecule) conjugates with the advantage of increasing uptake by cancer cells.

| Synthesis and characterization of the NanoAlb-proDOX conjugate
Encouraged by the data shown above, we next decided to prepare a real protein--drug conjugate and validate its application in cancer therapy. HSA, due to its biocompatibility, was selected as the protein of interest. In addition, a doxorubicin prodrug molecule was utilized to further reduce the side effects. 31 Therefore, HSA was first introduced with sulfhydryl groups as described above (Figure 2a The nanoparticle was speculated to assemble via the spontaneous oxidation of the excess sulfhydryl groups to form intermolecular disulfide bonds. Intracellular disulfide bonds and residual sulfhydryl groups may also exist. Red circles represent the proDOX molecules. HSA-TC-proDOX is hereafter named NanoAlb-proDOX. Data are presented as the mean ± SEM. n = 3 technical replicates. **p < 0.01, ***p < 0.001 with pH values of 7.5 or 5.5 were selected to simulate either the pH during circulation or the pH in the tumor microenvironment (or in the lysosome), respectively. As shown in Figure 2c, the conjugate was relatively stable at pH 7.5 (less than 40% doxorubicin released from the protein conjugate in 24 h), while it was quickly triggered to release active doxorubicin at pH 5.5 (approximately 50% doxorubicin released from the protein conjugate within 2 h). These results indicated that the conjugate preferentially releases the drug in the acidic tumor microenvironment and may thus reduce unexpected drug release during circulation, helping to minimize side effects.
As mentioned earlier, the HSA-TC-Sulfo-Cy5 conjugate prepared via the thiolation-and-conjugation approach not only did not alter the tumor targeting ability of HSA but also enhanced the uptake of HSA by cancer cells. Similar properties were also observed in the RBD-TC-Sulfo-Cy5 conjugate, showing that it may be a universal phenomenon. We therefore  Figure S7). Looking into the synthesis procedures (detailed in Section 5.2), we reasoned that the introduced sulfhydryl groups were much more likely to tend to form intermolecular disulfide bonds. Because, first, the aqueous buffer used was not deoxidized, which provided an oxidative environment, and the pH of the reaction buffer was slightly basic (pH 7.5), which both favored the formation of disulfide bonds via the mechanism of air oxidation. 32 In addition, in the conjugation step, dimethyl sulfoxide (DMSO) was added as a cosolvent. It has been reported that DMSO can serve as an oxidation reagent to facilitate the formation of disulfide bonds. 33 Moreover, sulfhydryl groups were installed on the surface exposing lysine residues, which made them accessible to each other and thus provided favorable conditions for disulfide bond formation. Therefore, we speculated that the excess amount of free sulfhydryl groups present on HSA may form intramolecular or intermolecular disulfide bonds during the overnight reaction. Based on this assumption, we further analyzed the particle sizes of HSA-TC-Sulfo-Cy5 and RBD-TC-Sulfo-Cy5 using TEM, and the images showed that HSA-TC-Sulfo-Cy5 and RBD-TC-Sulfo-Cy5 also formed nanometric particles with average sizes of 20 and 8 nm, respectively ( Figures S8 and S9).
Hence, we proposed the mechanism of nanoparticle assembly during the thiolation-and-conjugation approach via the formation of intermolecular disulfide bonds (Figure 2g). To prove the presence of intermolecular disulfide bonds, reducing and nonreducing SDS-PAGE was performed. As shown in Figure S10, larger molecular weight bands clearly existed under nonreducing conditions and disappeared under reducing conditions, proving the presence of intermolecular disulfide bonds. Therefore, these data further supported the existence of nanometric particles. To better illustrate the nanoscale structure of the conjugate, HSA-TC-proDOX is referred to hereafter as NanoAlb-proDOX.
Previous research has shown that the cellular uptake of gold nanoparticles increases with increasing particle size. 34 Our own previous study also proved that nanosized protein nanoparticles facilitate cellular uptake. 35 Therefore, the formation of albumin nanoparticles should be a reasonable explanation for the enhanced cellular uptake of NanoAlb-proDOX. Nanomaterials can be internalized by target cells through different mechanisms. For example, micrometer-sized particles usually enter target cells through phagocytosis or micropinocytosis, while nanometer-sized particles typically enter target cells through other types of endocytosis, such as clathrin-dependent, caveolin-dependent or receptor-mediated endocytosis. 36 The nanoparticles described in this study may undergo the mechanism for nanometer-sized particles.
Data acquired for HSA-TC-Sulfo-Cy5 and RBD-TC-Sulfo-Cy5, which were both prepared through the thiolation-and-conjugation method, showed that both displayed enhanced cellular uptake by cancer cells.
However, the two proteins enter cells through different receptors.
We therefore reasoned that the enhanced cellular uptake may be independent of their binding receptors. Clathrin-dependent or caveolin-dependent endocytosis may be the potential mechanism for the enhanced cellular uptake of albumin nanoparticles prepared by the thiolation-and-conjugation method. We therefore tested this hypothesis by using two inhibitors, Pistop 2 and genistein, which target clathrin-and caveolin-dependent endocytosis, 37 respectively. The results showed that both inhibitors reduced the endocytosis of HSA-TC-Cy5 in HeLa cells, indicating that the conjugate may be internalized through both endocytosis pathways ( Figure S11).

| Selectivity and toxicity of NanoAlb-proDOX
The selective cytotoxicity of anticancer drugs to cancer cells is critical to define their safety profiles. We therefore first evaluated the cytotoxicity of NanoAlb-proDOX in vitro. When internalized by cancer cells, doxorubicin localizes to the nucleus to cause DNA damage ( Figure S12). Similarly, when incubated with cancer cells, NanoAlb-proDOX also localized doxorubicin to the cell nucleus ( Figure S12). However, only slightly higher (for AST) or even no obvious change (for the other five markers) was observed for the NanoAlb-proDOXtreated group, revealing reduced side effects and better safety profiles for NanoAlb-proDOX in vivo. In contrast, tumor growth was strongly inhibited by treatment with NanoAlb-proDOX (78% tumor suppression), and only moderate inhibition was observed for the group treated with doxorubicin (35% tumor suppression, Figure 4c). Individual tumor volumes were plotted and also showed that mice treated with NanoAlb-proDOX exhibited more shrunken tumors (Figure 4d). After treatment, bioluminescence F I G U R E 4 Antitumor activity of NanoAlb-proDOX in breast cancer xenografts. (a) Schematic representation of the design of animal experiments. Drugs were administered intravenously every 3 days for a total of six doses. (b) Body weight curves of the treated mice. Body weights were measured every 3 days. (c) Average tumor volume curves of the treated mice. Tumor volumes were measured every 3 days. Significantly reduced tumor proliferation was observed in mice treated with NanoAlb-proDOX. (d) Individual tumor growth curves of the mice treated with Vehicle, DOX or NanoAlb-proDOX. (e) Representative bioluminescence images of the mice posttreatment. Images were taken at Day 21. The weakest bioluminescence signal was observed in mice treated with NanoAlb-proDOX. Data are presented as the mean ± SEM. n = 6 mice for the vehicle group and n = 5 mice for the DOX and NanoAlb-proDOX groups. ***p < 0.001, ****p < 0.0001 imaging (MDA-MB-231 cells stably transfected with the firefly luciferase gene) was performed to further confirm the antitumor activity of NanoAlb-proDOX. As shown in Figure 4e, mice treated with NanoAlb-proDOX had the weakest bioluminescence intensity, indicating the strongest antitumor activity of NanoAlb-proDOX. All these data proved that NanoAlb-proDOX possessed better antitumor activity than doxorubicin against triple-negative breast cancer xenografts.

| Antitumor activity of NanoAlb-proDOX
Breast cancer is the most common type of tumor in women. 39 While effective treatment and good prognosis can be expected for breast cancer patients, the treatment and prognosis are usually not satisfying for triple-negative breast cancer subtype patients, who make up 15%-20% of all breast cancer patients. 40 The NanoAlb-proDOX introduced here showed great anti-triple negative breast cancer activity in a mouse xenograft model, which may make NanoAlb-proDOX an effective alternative for the treatment of triplenegative breast cancer.
In addition to breast cancer, ovarian cancer is the sixth most common and most lethal cancer in women. 41 Ovarian cancer develops ascites during the late stage, 42 which exacerbates the patient's condition. Therefore, we next validated whether NanoAlb-proDOX can address the situation of ascites in ovarian cancer. We constructed an ascites model of ovarian cancer using the ID8-Luc cell line (a mouse ovarian cancer cell line stably transfected with firefly luciferase, Figure S13A). ID8-Luc cells were directly injected into the peritoneal cavity of female C57BL/6 mice to allow tumor dissemination and ascites formation in approximately 4 weeks. The establishment of tumors in the peritoneal cavity was monitored using in vivo live animal bioluminescence. Available xenograft mice were divided into three groups, which were administered PBS, 1.5 mg/kg doxorubicin (DOX) or NanoAlb-proDOX (NanoAlb-proDOX, equivalent to 1.5 mg/kg doxorubicin) intravenously every 3 days for a total of five doses ( Figure S13A). The body weights of the mice in each group were monitored every 3 days during the administration period. As indicated, the body weights of the mice in the PBS and DOX groups were slightly higher, mainly due to heavily developed ascites masses, while the body weights of the mice in the NanoAlb-proDOX-treated group remained stable with reduced ascites masses ( Figure S13B). The progression of ascites and tumor dissemination was monitored by in vivo bioluminescence imaging ( Figure S13C). Images showed that both groups of mice treated with either doxorubicin or NanoAlb-proDOX displayed reduced bioluminescence signals compared with PBS-treated mice, indicating the effectiveness of both drugs in the ascites xenograft model. Quantitative analysis of the bioluminescence imaging data from Day 0 and Day 14 revealed that doxorubicin treatment still resulted in slight tumor proliferation, while NanoAlb-proDOX treatment led to stable or even slightly reduced tumor proliferation ( Figure S13D). Afterward, survival of the mice was monitored. Survival curves revealed prolonged median survival times for both doxorubicin-and NanoAlb-proDOX-treated mice, with NanoAlb-proDOX treatment leading to greater benefits (median survival times: 100 days for NanoAlb-proDOX versus 81 days for doxorubicin, Figure S13E).
In brief, the NanoAlb-proDOX invented here showed improved antitumor activity in both triple-negative breast cancer and ovarian ascites xenograft models. Taking into consideration its selective cytotoxicity toward cancer cells in vitro and reduced side effects in vivo, NanoAlb-proDOX may be a good choice for cancer therapy.

| Synergy with immune checkpoint blockade
Although immune checkpoint blockade has been proven powerful to battle cancer, the overall responses are low, [43][44][45] and great improvement is therefore still desired. One promising solution is combining chemotherapy with immune checkpoint inhibitors. 13 The NanoAlb-proDOX invented in the manuscript showed improved safety profiles and superior antitumor activity; therefore, we combined NanoAlb-

| Enhanced tumor infiltration of CD8 + T cells
The efficacy of immune checkpoint blockade is limited by the infiltration and activation of T cells in tumors, especially cytotoxic CD8 + T cells. NanoAlb-proDOX had a synergistic effect with α-PD-L1 treatment, so we asked whether this synergy was related to different T-cell infiltrations. Therefore, the tumor tissues collected above were subjected to immunohistochemistry (IHC) analysis. Tumor tissue slices were stained with anti-mouse CD8 antibody ( Figure 6a). As shown in Figure 6a, the infiltration of CD8 + cells into the untreated tumor tissue was rare, as shown by the IHC images from the Vehicle group.
Monotherapies with α-PD-L1, doxorubicin or NanoAlb-proDOX and To further support the findings from IHC and flow cytometry, we also conducted an in vitro coculture assay using purified CD8 + T cells to validate whether cotreatment with NanoAlb-proDOX and α-PD-L1 stimulates the proliferation of CD8 + T cells. 46

| DISCUSSION
In this study, we described the validation of a thiolation-andconjugation method to prepare protein and drug conjugates. The  Figure 1 and Figure S4). The thiolation-and-conjugation method did not affect the in vivo tumor targeting ability of albumin in a human triple-negative breast cancer model (Figure 1c). Moreover, the conjugates prepared through this method possessed enhanced cancer cellular uptake in vitro (Figure 1d and Figure S4), suggesting that the protein-drug conjugates prepared via this strategy may be useful in cancer therapy.
As a proof of concept, we prepared an albumin doxorubicin conjugate with the thiolation-and-conjugation method (Figure 2). In addition to the conjugation method, the doxorubicin drug molecule was designed to contain an acid-sensitive linker to facilitate the specific release of the drug molecule from the conjugate (Figure 2a,c). As part of the validation, we revealed that the albumin and doxorubicin conjugate (HSA-TC-proDOX) exhibited good tumor targeting in vivo and enhanced cancer cellular uptake in vitro (Figure 2d,e). In exploring the mechanism behind the enhanced cancer cellular uptake of the prepared conjugates, we thought that the native receptors of the proteins may play very limited roles, as albumin and RBD bind to different receptors (e.g., SPARC/gp60 for albumin and ACE2 for RBD). Using TEM and DLS technologies, we discovered that the HSA-TC-proDOX, HSA-TC-Sulfo-Cy5 and RBD-TC-Sulfo-Cy5 all formed a nanosized structure (Figure 2f, Figure S7, S8 and S9), and we proposed a possible mechanism involving the formation of intermolecular disulfide bonds (Figure 2g). We therefore speculated that the nanoparticle structure may be a key reason for the enhanced cellular uptake, as nanoparticles tend to accumulate in cells. To further determine the endocytosis pathway of the conjugate prepared via the thiolation-and-conjugation method, we performed an endocytosis inhibition assay with clathrin and caveolin inhibitors. The results showed that the cellular uptake of the albumin conjugate was reduced by both inhibitors, indicating that the conjugate may adopt both clathrin-dependent and caveolin-dependent endocytosis pathways ( Figure S11).
To validate the properties of the NanoAlb-proDOX conjugate, we showed that NanoAlb-proDOX preferentially killed cancer cells in vitro (Figure 3a) and had reduced side effects in vivo compared with doxorubicin (Figure 3b-d). These data indicated that NanoAlb-proDOX could become a safer antitumor drug. The effectiveness of NanoAlb-proDOX was validated in two different xenograft models, in which both revealed that NanoAlb-proDOX was more effective than doxorubicin (Figure 4 and Figure S13). In another combination therapy trial, NanoAlb-proDOX was coadministered with α-PD-L1 ( Figure 5).
The results showed that the two molecules induced significant synergistic effects and greatly inhibited tumor proliferation in a colorectal cancer xenograft model ( Figure 5). In exploring the mechanism behind the synergistic effects, we found that the tumor infiltration of CD8 + T cells was enhanced (Figure 6a,b). With an in vitro assay, we provided evidence that the enhanced infiltration may be a result of enhanced CD8 + T-cell proliferation, which was stimulated by the coadministration of NanoAlb-proDOX and α-PD-L1 (Figure 6c). These data indicated that NanoAlb-proDOX can be a costimulator of immune checkpoint blockade. In addition, with a preliminary trial, we found that NanoAlb-proDOX induced a small amount of tumor cell pyroptosis in vitro (data not shown in this manuscript, but available if required). It has also been reported in the literature that cytotoxic lymphocytes induce tumor cell pyroptosis, and tumor cell pyroptosis in turn induces T-cell infiltration. 47,48 We therefore reasoned that in combination, NanoAlb-proDOX and α-PD-L1 may induce potent tumor cell pyroptosis, which then induces cytotoxic CD8 + T-cell infiltration/proliferation. This hypothesis will be tested in our following study.

| CONCLUSIONS
In summary, we reported the validation of a generally applicable thiolation-and-conjugation approach to generate protein-drug conjugates. The resulting protein-drug conjugates were not affected by the conjugation method in terms of the tumor targeting ability in vivo and exhibited enhanced cancer cellular uptake. As a proof of concept of the utility of the method, an albumin-doxorubicin prodrug conjugate NanoAlb-proDOX was prepared. NanoAlb-proDOX self-assembled into nanoparticles, probably via the formation of intermolecular disulfide bonds, which were believed to be responsible for its enhanced cancer cellular uptake. NanoAlb-proDOX exhibited selective cytotoxicity toward cancer cells in vitro and reduced side effects in vivo. In HSA and anti-PD-L1 antibody were kindly gifts from Zhejiang Hisun Pharmaceutical Co., Ltd. Anti-CD8 antibody was purchased from Cell Signaling Technology (Cat. # 98941S). RBD protein was expressed and purified from Pichia pastoris by our collaborators. 35 All mice (BALB/c nude and C57BL/6) were purchased from Charles River.

| Characterization of NanoAlb-proDOX
To determine the conjugation efficiency of HSA-proDOX, a method that plots the fluorescence intensity of doxorubicin was adopted. A MC38 cells were cultured in RPMI 1640 culture medium supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin at 37 C under 5% CO 2 .

| In vitro cytotoxicity assay
Cells were seeded in 96-well plates at a density of 5000 cells per well.
Twenty-four hours later, NanoAlb-proDOX or doxorubicin at the indi- For the animal experiment shown in Figure 4, drugs were admin-

| Detection of serum or tissue markers
Forty-eight hours after doxorubicin or NanoAlb-proDOX injection, the mice were sacrificed. Blood was collected from eyeballs, and serum was separated. Serum levels of ALT, AST, LDH, CK-MB, cTnI, and cTnT were determined using ELISA. For the determination of IFN-γ and IL-2 in tumor tissues, the supernatant of the tumor tissue homogenates in PBS buffer was analyzed using ELISA.

| Flow cytometry analysis of the infiltrated T cells
The mice shown in Figure 5  markers were gated and quantitatively analyzed.

| In vivo fluorescence and bioluminescence imaging
All in vivo images were taken with a PerkinElmer IVIS III system or IVScope 8200 (CLINX). For fluorescence imaging, proteins were first labeled with the indicated fluorophore (Cy5.5 or Cy5) and injected intravenously into mice. Images were taken in the cy5 channel. For bioluminescence imaging, mice were first injected intraperitoneally with D-luciferin (1 mg/ml, 100 μl). Ten minutes later, images were taken with the bioluminescence settings.

| Histology analysis
After the mice were sacrificed, the major organs were recovered from the necropsy and fixed with 10% neutral buffered formalin. Afterward, the organs were embedded in paraffin and sectioned at 5 mm, and hematoxylin and eosin (H&E) staining was performed for histological examination.

| IHC analysis
After the mice were sacrificed, the tumor tissues were harvested and fixed with 4% paraformaldehyde. Formalin-fixed, paraffin-embedded samples were cut into 4 μm sections. IHC was performed on the Leica Bond automated staining platform. The CD8 antibody (CST, #98941) was run at 1:400 dilution using the Leica Biosystems Refine Detection Kit with EDTA antigen retrieval. Images were captured using a Zeiss

| Statistical Analyses
All the values are presented as the mean ± S.E.M. The number n represents technical replicates or the number of mice used. All statistical analyses were performed with GraphPad Prism 7 software. Significance tests between two groups were performed using two-tailed unpaired t tests. Significance was defined as p > 0.05 n.s. not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.