A Chimera Antigen Receptor Containing TLR4 Signaling Domain Enhances CAR-iMACs Polarization and Potency against Solid Tumors

Chimeric antigen receptor (CAR)-T cell therapies have shown success in treating certain types of hematologic malignancies, but its therapeutic effect on solid tumors is unsatisfactory. Macrophages came to attention because of its phagocytosis function against tumor cells and its immunomodulation capacity. The �rst generation of engineered CAR-macrophages demonstrated that the CAR can stimulate macrophage phagocytosis function in an antigen dependent way. In this work, we genetically engineered induced pluripotent stem cell (iPSC) derived macrophages (iMACs) with a TLR4 intracellular TIR domain-containing CAR, and achieved enhanced anti-tumor effect. CD3 ζ -TIR-CAR, the second generation of TIR-based dual signaling CAR endowed iMACs the target engul�ng capacity against antigen-expressing tumor cells, as well as potency of antigen-dependent M1 polarization and resistance to M2 polarization in a NF-κ B dependent manner. Taken together, we established the next generation of CAR-iMAC equipped with concurrent antigen-dependent phagocytosis and polarization capacity for better anti-tumor functions.


Introduction
Cancer immune cell therapies illustrated by CAR-T cells have achieved great success in B cell leukemia/lymphoma and other hematologic malignancies [1][2][3][4][5][6][7] .Although the therapeutic cells were designed to precisely target solid tumor antigens, the treatment of solid tumors has not achieved substantial progress [8][9][10][11] .To date, the challenges such as exhaustion in the immunosuppressive tumor microenvironment (TME) 12,13 , limited in ltration into dense extracellular matrix [14][15][16][17] , off-target effects 18 , antigen escape and heterogeneity within the tumor etc 19,20 , have been thought to be the major obstacles for successful immune cell therapies against solid tumors.Macrophage, a terminally differentiated monocytic phagocyte, comes to the horizon of tumor immunotherapy eld because of its engul ng capacity and central role in the crosstalk between adaptive and innate immune systems [21][22][23][24] .Macrophages are found to reside in normal tissues and tumors, and play critical roles in defensing pathogens and combating cancer cells 21,23−26 .As a consequence, macrophages become a potential target of immunotherapy or a type of therapeutic cell for adoptive transfer 25,27−31 .Tumors have been found to consist of large number of immune cells besides neoplastic cells [32][33][34] .Among the tumor-residing leukocytes, tumor-associated macrophages (TAMs) account for the majority of the population 25,35 .However, these macrophages are reprogramed toward the alternatively activated M2 phenotype, and play tumorigenic role in breast, prostate, ovarian and other carcinomas, through suppressing other immune responses, inducing angiogenesis, etc [36][37][38][39][40][41][42][43] .Therefore, the effort to introduce more pro-in ammatory M1 polarization macrophages state or to convert M2 to M1 macrophages in the immunosuppressive TME would be a viable anti-tumor strategy [44][45][46] .Previous efforts mainly focused on targeting endogenous macrophages or TAMs 47,48 .Recent technology progress in genetically engineering macrophages enabled us to better harness macrophages as a type of therapeutic cells for adoptive transfer 29,31,49−55 .Currently, pre-clinical and clinical studies using macrophages that have been genetically engineered are in development.In consideration of the availability of large source, stability and standardization of the engineered macrophages, iPSCs differentiated iMACs have been utilized as a valuable alternative solution 29,56−58 .Previous works veri ed the feasibility of producing CAR-iMACs, and demonstrated the CAR-dependent activation of CAR-iMACs in the treatment of solid and blood cancers 29 .However, the 1st generation of CAR containing CD3ζ activating domain used in macrophages was borrowed from CAR-T cells (T-CAR), and was mainly responsible for effector cell function of phagocytosis, and in theory had no capacity to polarize macrophages toward the durable M1 pro-in ammatory state 29 .Thus, it's urgent to design a new macrophage speci c CAR (M-CAR) to confer both phagocytosis and polarization functions.
In this work, we developed the 2nd generation macrophage-speci c CAR (M-CAR) by integrating intracellular CD3ζ and TIR domains in tandem, to construct an EGFRvIII-targeting M-CAR.Genetically modi cation of iMACs with the 2nd generation M-CAR markedly improved antigen-dependent anti-tumor e cacy both in vitro and vivo.With further cellular and molecular analysis and comprehensive single cell RNA-seq assessment we proved the underlying mechanisms were that TIR promoted the M1 proin ammatory state and suppressed the M2 state in a NF-κB pathway-dependent manner.Thus, we established a proof-of-concept 2nd generation of CAR-iMACs for future cancer immune cell therapies in solid tumors.

Construction of macrophage speci c CARs and development of EGFRvIII-targeting CAR-iMACs
The CD3ζ signal transduction domain obtained from T cell speci c CAR has been proved to be capable of stimulating the rst generation of human CAR-iMACs 29,31,49 .However, the domain cannot be helpful to maintain the M1 state of macrophages in TME 29 .Therefore, we designed macrophage speci c CARs containing the intracellular TIR domain, whose activation can drive M1 polarization of macrophages via interplaying with TIR domain-containing adaptors.Meanwhile, we employed EGFRvIII-targeting scFv as the extracellular domain.We designated it as TIR-CAR to distinguish it from the previous CD3ζ-CAR (Figure 1a).Inspired by the evolutionary process of different generations of CARs in T cells (T-CARs), we simultaneously designed a second generation macrophage speci c CAR (M-CAR) by integrating intracellular CD3ζ and TIR domains in tandom designated as CD3ζ-TIR-CAR (Figure 1a).A truncated CAR without an intracellular domain was designed as the negative control (Figure 1a).We then transduced the above CARs into human iPSCs with lentivirus.CAR-expressing iPSCs were subsequently differentiated into CAR-iMACs through the periods of mesoderm induction, hematopoietic speci cation, myeloid expansion, and macrophage maturation (Figure 1b).Flow cytometric (FCM) analysis showed that all types of CAR-iMACs shared about 80% transgene expression (Figure 1c).Thus, we successfully developed four types of EGFRvIII-targeting CAR-iMACs.
We subsequently analyzed expression of TLR4 signaling pathway genes at different stages of myeloids/macrophages differentiation with RNA-seq data.We found that the key factors such as IRF7, NF-κB/P65, MYD88, TRAP6, TRAM, and TAB2/3 were highly expressed on day 18 and declined on day 28 (Figure 1d).This data suggested that most of the downstream modulators of TLR4 are higher when the iMACs are still in a monocyte-like state, compared with the later stage when the iMACs are more matured.Furthermore, in vitro M1 polarized iMACs expressed higher level of the same TLR4 signaling pathway genes compared to the naive or M2 polarized iMACs (Figure 1e).Together, the data suggested that the early differentiated iMACs resembled the M1 polarized iMACs in terms of higher expression of the TLR4 pathway adaptor and signal transduction genes, and therefore might possess more potent signaling effects from the activated TIR domain.
In order to make a target cell line for the EGFRvIII CAR expressing iMACs, we cloned the human EGFRvIII sequence into a lentivirus vector (Supplemental Figure 1a), and overexpressed the gene in U87MG cells to develop a stable EGFRvIII-expressing tumor cell line (U87MG EGFRvIII ).While basic expression of total EGFR was observed in both wild-type (WT) and transgenic U87MG cells, only the latter had strong expression of EGFRVIII protein (Supplemental Figure 1b and c).With immuno uorescence, we observed the location of EGFRVIII on the cell membrane, as well as in cytoplasm due to overexpression (Supplemental Figure 1d).Thus we have successfully constructed an U87MG EGFRvIII cell line for testing effector functions of EGFRvIII CAR-iMACs.

Both TIR and CD3ζ domains enhanced antigen-dependent functions of CAR-iMAC in vitro
Previous work demonstrated that the rst generation of CAR conferred antigen-dependent functions of CAR-iMACs 29,31,49 .In order to evaluate whether the TIR signal domain can also enhance phagocytosis of CAR-iMACs against tumor cells in an antigen-dependent manner, we rst performed experiments to incubate the single intracellular domain-containing CAR-iMACs cells with U87MG EGFRvIII cells in vitro.We rst stimulated WT-iMACs, truncated CAR-iMACs, CD3ζ-CAR-iMACs, and TIR-CAR-iMACs with IFN-γ and LPS respectively, and then co-cultured the cells with tdTomato-expressing (tdTomato + ) U87MG EGFRvIII cells with different effect/target ratios (E/T) of 3/1, 5/1, and 10/1.After 12 hours of co-culturing, we observed enhanced adhering property of CAR-iMACs to the antigen-expressing cancer cells compared to WT-iMACs (Supplemental Figure 1e), signifying that the CAR enabled iMACs with targeting capacity.Moreover, immuno uorescence assays showed that both CAR-iMACs and WT-iMACs engulfed tdTomato labeled tumor cells (Supplemental Figure 1f), demonstrating that the iMACs possess basic phagocytosis like primary macrophages, and overexpression of CAR did not impair this feature.Next, we collected all the cells and assessed the number of residual tumor cells by FCM analysis (Supplemental Figure 2a).
Compared to WT-iMACs or truncated CAR-iMACs, U87MG EGFRvIII cells co-incubated with CD3ζ or TIR-CAR-iMACs suffered more killing and remained fewer cells at the same condition of E/T ratio (Supplemental Figure 2a and b).At the same time, all types of CAR-iMACs showed strengthened cytotoxicity with increased E/T ratios (Supplemental Figure 2b).Moreover, we did not observe obvious disparity of cytotoxicity between CD3ζ-CAR-iMACs, and TIR-CAR-iMACs at the 12 hours incubating time point (Supplemental Figure 2b).
To demonstrate progressive cytotoxicity of CAR-iMACs against their target cells, we next performed the same experiments at 24 hours of co-incubating.Compared to 12 hours, 24 hours co-incubating signi cantly improved tumor cell killing e cacy of CAR-iMACs at all tested E/T ratios (Figure 2a and b).
Although both CD3ζ-CAR-iMACs and TIR-CAR-iMACs showed powerful anti-tumor competence compared to truncated iMACs, it appeared the TIR domain conferred stronger killing capacity than CD3ζ at this time point (Figure 2b).When we compared the trend of killing effect of the three types of CAR-iMACs from 12 hours to 24 hours co-incubating with U87MG EGFRvIII cells, we found that TIR-CAR-iMACs had greater killing persistence than CD3ζ-CAR-iMACs as time goes on (Figure 2c).To validate the above ndings, we overexpressed a re y luciferase (FFluc) gene in U87MG EGFRvIII cells, and performed more killing assays to further test the anti-tumor cell capacity of the CAR-iMACs.Consistently, the bioluminescence signal from tumor cells was evidently attenuated by both CD3ζ-CAR-iMACs and TIR-CAR-iMACs comparing to truncated CAR-iMACs, and the e ciency achieved optimum at the E/T ratio of 10/1 (Figure 2d).Furthermore, in order to investigate whether the pro-in ammatory activity of CAR-iMACs is relied on CAR activation, we measured release of TIR downstream cytokines.ELISA tests revealed that when encountering U87MG EGFRvIII cells for 24 hours at an E/T ratio of 10/1, TIR-CAR-iMACs secreted considerable pro-in ammatory cytokines including IL6 (Figure 2e), IL12A (Figure 2f) and IL23 (Figure 2g), indicating more pro-in ammatory activity of TIR-CAR-iMACs.Interestingly, the CAR containing T cellspeci c CD3ζ domain also promoted milder expression of these cytokines in CAR-iMACs (Figure 2e-g), and the expression of TNF-α, another TLR4 regulated immune factor, was unexpectedly higher in CD3ζ-CAR-iMACs than in TIR-CAR-iMACs (Figure 2h), suggesting combining the two might achieve the maximal pro-in ammatory activity.Taking together, these results demonstrated that both TIR and CD3ζ domains enhanced CAR-dependent cytotoxic activity of CAR-iMACs, but TIR-CAR-iMACs exhibited more potential of persistent and potent pro-in ammatory activity.
TIR-CAR-iMACs possess stronger anti-tumor potency than CD3ζ-CAR-iMACs in vivo We next investigated anti-tumor activities of CD3ζ-CAR-iMACs and TIR-CAR-iMACs in tumor-bearing animals.We rst established the tumor model in abdominal cavity of immunode cient mice with FFluc + U87MG EGFRvIII cells.IFN-γ/LPS polarized truncated CAR-iMACs, CD3ζ-CAR-iMACs, and TIR-CAR-iMACs were labeled by the DiR dye, and then injected intraperitoneally into the tumor-bearing mice with an E/T ratio of 10/1.Animal imagings were performed 2 hours after immune cells injection (day 0), and at the same time of day 1, day 3 and day 7 (Figure 3a).All the three types of CAR-iMACs exerted anticancer cell activity against the tumor at day 1 immediately (Figure 3b and c).However, comparing with truncated CAR-iMACs and CD3ζ-CAR-iMACs, TIR-CAR-iMACs exhibited stronger anti-tumor activity, and achieved optimum e cacy at day3 (Figure 3c).Of note, even though the tumor started to relapse at day 7, the tumor signals from TIR-CAR-iMACs-treated mice were still maintained at a lower level than that in PBS and CD3ζ-CAR-iMACs-treated mice, suggesting that introducing the TIR domain into CAR-iMACs might have conferred resistance to reprogramming by the tumor microenvironment over time.Further survival analysis showed that TIR-CAR-iMACs markedly prolonged the life time of tumor-bearing mice (Figure 3d).Collectively, the data demonstrated that although both TIR-CAR-iMACs and CD3ζ-CAR-iMACs showed antigen-dependent anti-tumor activity in vivo, the TIR-CAR conferred more persistent killing capacity of CAR-iMACs against cancer cells.

The combination of TIR and CD3ζ domains strengthened M1 polarization and anti-tumor potency of CAR-iMACs
In the above work, we have veri ed that TIR-CAR-iMACs exhibited the M1 phenotype and both TIR and CD3ζ domains enabled CAR-iMACs with antigen-dependent phagocytosis capacity.Based on the design of second generation of T-CAR, we hypothesized that integration of TIR and CD3ζ domains into one CAR structure could further promote anti-tumor potency of CAR-iMACs.Thus we developed a new M-CAR consisting of intracellular CD3ζ and TIR domains in tandom (Figure 1a), and then generated CD3ζ-TIR-CAR-iMACs (Figure 1c).We rst compared expression of pro-in ammatory or M1 polarization genes supposedly driven by TIR activation.As expected, qPCR analysis revealed that, when incubating the three types of CAR-iMACs with U87MG EGFRvIII cells for 24 hours, expression of IL1A/B, IL6, IL12A, IL23, CCL8, CXCL8 and TNF-α were elevated signi cantly in CD3ζ-TIR-CAR-iMACs compared to TIR-or CD3ζ-CAR-iMACs at the condition of no IFN-γ/LPS pretreatment (Figure 4a-h), revealing a synergistic effect of TIR and CD3ζ domains.Correspondingly, ELISA testings con rmed that CD3ζ-TIR-CAR-iMACs improved production of IL6, IL12A, IL23, and TNF-α (Figure 4i-l).These results further demonstrated that the TIR domain contributed to M1 polarization of CAR-iMACs.Subsequently, we examined cytotoxicity of the integrated CD3ζ-TIR-CAR-iMACs confronting FFluc + U87MG EGFRvIII cells.Measurement of bioluminescence from FFluc + U87MG EGFRvIII cells indicated that comparing to single CD3ζ or TIR domain-containing CAR-iMACs, 24 hours incubation with CD3ζ-TIR-CAR-iMACs led to the lowest number of residual tumor cells under the condition of E/T of 10/1(Figure 4m), revealing that CD3ζ-TIR-CAR-iMACs conferred much more lethality against EGFRvIII positive tumor cells than single domain-containing CAR-iMACs.
Previous reports have illustrated that the TME leads to exhaustion of T cells and promotes M2-like polarization of macrophages 12,38,72−75 .To explore whether the TIR based CAR-iMACs were equipped with the ability of withstanding immunosuppressive effects from tumor cells, we examined the polarization state of the four types of CAR-iMACs co-cultured with tdTomato + U87MG EGFRvIII cells under the condition of no IFN-γ/LPS pre-treatment and an E/T ratio of 3/1.Subsequent FCM analysis showed that 24 hours incubation with the tumor cells stimulated high expression of CD80 in all types of CAR-iMACs, and the two TIR-based CAR-iMACs showed more CD80-positive (CD80 + ) populations compared to those in truncated CAR-iMACs and CD3ζ-CAR-iMACs (Supplemental Figure 3a).Of note, prolonged periods of exposure to tumor cells resulted in marked reduction of CD80 + populations to about 40-50% for truncated CAR-iMACs and CD3ζ-CAR-iMACs (Supplemental Figure 3a, and Figure 4n and p), whereas the percentage of CD80 + cells in both TIR-CAR-iMACs and CD3ζ-TIR-CAR-iMACs persisted at a higher level of ~80% (Supplemental Figure 3a, Figure 4n and p), indicating that TIR contributed to stimulation of the M1 state.
On the other hand, we also determined M2 populations of the four types of CAR-iMACs by measuring CD163, a typical M2 marker, and found that long time exposure to the tumor cells led to various level of CD163 expression in the rst three types of CAR-iMACs, whereas CD3ζ-TIR-CAR-iMACs exhibited robust potency against M2 polarization (Supplemental Figure 3b, Figure 4o and q).Indeed, the percentage of CD163 + CD3ζ-TIR-CAR-iMACs declined to about 36.93% at day 3 of co-incubation, and remained to be low at day 7 (Supplemental Figure 3b and Figure 4o and q).Based on the above results, we concluded that the combination of the TIR and CD3ζdomains supported M1 polarization of CAR-iMACs exposing to tumor cells, and contributed to prevent reversion of M1 to M2 state.
Noteworthy, CD3ζ-TIR-CAR-iMACs showed the strongest anti-tumor potency in the living body environment (Supplemental Figure 4b and c), and prolonged survival time of the tumor-bearing mice most prominently (Supplemental Figure 4d).In order to assess whether TIR contributed to M1 polarization of CAR-iMACs in the course of treatment inside the living body, we isolated the immune cells from abdominal cavities of the mice after two days of treatment, and measured CD80 + and CD163 + populations by FCM.As observation in vitro, more TIR-CAR-iMACs and CD3ζ-TIR-CAR-iMACs exhibited the CD80 positive feature than that in CD3ζ-CAR-iMACs (Supplemental Figure 4e and f), and much lower CD163 positive M2 feature was present in CD3ζ-TIR-CAR-iMACs (Supplemental Figure 4e and  f).In summary, we con rmed that integration of intracellular CD3ζ and TIR domains into one CAR structure could enhance cytotoxic activity against tumor cells and potency of maintaining M1 polarization of CAR-iMACs.
The CAR promotes M1 polarization of CAR-iMACs via TIR domain-mediated canonical NF-κB/P65 transcription activity The above data demonstrated that the TIR domain enhanced M1 polarization and activation of CAR-iMACs in an antigen-dependent manner.TIR domain is responsible for signal transduction from TLR4, and the downstream signaling liberates NF-κB/P65 from IκB-α-mediated degradation followed by the translocation of NF-κB/P65 into nucleus to transcriptionally activate pro-in ammatory target genes 70,76 .In order to investigate whether NF-κB/P65 mediated the pro-in ammatory activity of TIR-CAR-iMACs encountering the antigen, we rst tested expression of NF-κB/P65 gene RELA with qPCR analysis and found its higher expression in both TIR domain-containing CAR-iMACs (Figure 5a).Whereas ERK, another gene involved in macrophage activation had no distinction of expression in all CAR-iMACs (Figure 5b).Next, we monitored nucleus translocation of NF-κB/P65 by immuno uorescence when the control truncated CAR-iMACs were treated with LPS for 20 minutes (Supplemental Figure 5a).Afterwards, we cocultured truncated CAR-iMACs, TIR-CAR-iMACs, CD3ζ-CAR-iMACs, and CD3ζ-TIR-CAR-iMACs with tdTomato + U87MG EGFRvIII cells respectively with an E/T ratio of 3/1, followed by immuno urescence to detect NF-κB/P65 subcellular localization.Confocal imaging showed that su cient contact between CAR-iMACs and U87MG EGFRvIII cells for 4 hours initiated expression of NF-κB/P65 in CD3ζ-TIR-CAR-iMACs (Supplemental Figure 5b).When the incubating time was prolonged to 10 hours, NF-κB/P65 exhibited evident expression in both TIR-CAR-iMACs and CD3ζ-TIR-CAR-iMACs, and weak expression in CD3ζ-CAR-iMACs, and it aggregated in the nucleus of the two types of TIR-based CAR-iMACs (Figure 5c and d).
Besides, we didn't detected obvious expression and nuclear localization of NF-κB/P65 in the TIR-CAR-iMACs that did not touch U87MG EGFRvIII cells (Figure 5c), consistent with an antigen-dependent activation mechanism.Together, these results demonstrated that the NF-κB signal pathway was activated in the TIR domain-containing CAR-iMACs.
To further con rm the role of NF-κB/P65 in CAR-iMAC, we treated the CAR-iMACs co-cultured with FFluc/tdTomato + U87MG EGFRvIII cells with a NF-κB/P65 inhibitor JSH23.As expected, blocking of nuclear location of NF-κB/P65 by JSH23 abrogated cytotoxicity of CAR-iMACs against FFluc/tdTomato + U87MG EGFRvIII cells (Figure 5e).Similarly, FCM analysis of residual tumor cells indicated that blocking NF-κB/P65 into nucleus increased residual tdTomato positive tumor cells in TIR-CAR-iMACs and CD3ζ-TIR-CAR-iMACs groups (Figure 5F).Collectively, we con rmed that immune activation and M1 polarization of TIR-based CAR-iMACs were through the NF-κB pathway.
Single cell RNA-sequencing revealed robust molecular features and mechanisms of M1 polarization of TIR-based CAR-iMACs We next attempted to examine to what extent the TIR-based CAR-iMACs could promote the proin ammatory state when exposed to tumor cells through single cell RNA-sequencing analysis (Figure 6a).In this experiment, we pretreated the CAR-iMACs with IFN-γ/LPS for 24 hours, and then incubated each of them with U87MG EGFRvIII cells for another 24 hours.After that, we performed 10×genome single cells RNAsequencing analysis with all the cells, and found that they could be clustered into six sub-populations (Supplemental Figure 6a, c and e).According to the in vivo single cell RNA-seq data of human hematopoietic development, C1, C2 and C3 clusters were mainly matched to the Macrophage cell phenotype (Supplemental Figure 6b, d and f).Meanwhile, C4 and C5 clusters were primarily matched with the U87MG cancer cell phenotype in all the three types of CAR-iMACs groups we examined.Interestingly, the C5 cluster also showed evident macrophage feature in the CD3ζ-TIR-CAR-iMACs group, indicating these were the cells that engulfed the tumor cells and thus assumed features of both macrophage and tumor cells (Supplemental Figure 6b, d and f).According to the single cell analysis, compared with truncated CAR-iMACs and TIR-CAR-iMACs groups, CD3ζ-TIR-CAR-iMACs treatment resulted in fewer remaining tumor cells (Supplemental Figure 6a-f, and Figure 6b).In addition, even though the C6 cluster of all the three groups were mainly matched with macrophage, C6 in TIR-CAR-iMACs and CD3ζ-TIR-CAR-iMACs showed more evident and incremental DC characteristics compared to truncate CAR-iMACs (Supplemental Figure 6b, d and f), implying that TIR-based CAR-iMACs might have acquired enhanced antigen-presenting competence as dendritic cells.
As in the above results, we have demonstrated the M1 polarization-promoting effect of TIR-CAR.Next, we performed more quantitative analysis of M1 or M2 markers using single cell RNA-seq data.Normalized gene expression showed that more TIR-CAR-iMACs exhibited high expression of CD80 and CD86, and CD3ζ-TIR-CAR further boosted the expression levels of the two M1 marker genes, (Figure 6c-e).Accordingly, expression of M2 marker CD206 was reduced in TIR-CAR-iMACs, and to the lowest level in CD3ζ-TIR-CAR-iMACs, compared to that in truncated CAR-iMACs (Figure 6c-e).Moreover, the percentage of CAR-iMACs that expressed other M1 state-associated genes including CD83, CCL8, CXCL9, and CXCL11, increased incrementally from truncated CAR-iMACs to TIR-CAR-iMACs and CD3ζ-TIR-CAR-iMACs (Figure 6f).Accordingly, the normalized expression level of these genes exhibited the same trend (Figure 6g-j).On the contrary, fewer percentage of TIR-CAR-iMAC and CD3ζ-TIR-CAR-iMACs expressed M2 stateassociated genes such as CCL13 and CD163 (Figure 6k), and even though only a handful of cells had these genes being detected, TIR-CAR-iMAC and CD3ζ-TIR-CAR-iMACs presented much lower expression compared to truncated CAR-iMACs (Figure 6l and m).Collectively, all the results illustrated that integration of the TIR and CD3ζ domains facilitated M1 genes expression and suppressed M2 gene expression in the same time.
Next we further investigated whether the global gene expression signature resembled that of the M1 macrophages.We divided CAR-iMACs to CD80-expressing and CD80-absent (non-CD80) populations, as the latter represented the iMACs that might not encounter tumor cells, and evaluated the discrepancy of their expression pro le.Principle component analysis (PCA) using M1-related genes identi ed in our previous study indicated that CD80-expressing TIR-CAR-iMACs and CD3ζ-TIR-CAR-iMACs were positioned closer to the IFN-γ/LPS-stimulated M1 iMACs 29 , compared to CD80-positive truncated CAR-iMACs and non-CD80 CAR-iMACs from all the types (Figure 6n).Furthermore, pathway analysis revealed that CD80expressing CAR-iMACs showed lower level of M2-related pathways such as Oxidative phosphorylation and Respiratory electron transport, than that in non-CD80 CAR-iMACs, but higher level of M1-related pathway Interferon gamma response (Figure 6o).More importantly, within the CD80-expressing groups, TIR-CAR-iMACs and CD3ζ-TIR-CAR-iMACs showed evidently higher Interferon gamma response and lower Oxidative phosphorylation (Figure 6o), indicating that TIR based CAR conferred the CAR-iMACs with strengthened pro-in ammatory activity.Lastly, via hierarchical clustering analysis based on a de ned set of M1 and M2-associated genes 29 , we found that both TIR-CAR-iMACs and CD3ζ-TIR-CAR-iMACs clustered with IFN-γ/LPS-polarized iMACs, whereas in contrast, truncated CAR-iMACs clustered with M2 polarized iMACs (Supplemental Figure 7a and b).The same trend of clustering results was observed when using the CD80-expressing populations (Supplemental Figure 7c and d).To sum up, the above single cell RNA-seq analysis strongly demonstrated that TIR-based CARs contributed to the M1 polarization of CAR-iMACs in an antigen-dependent manner when encountering tumor cells, and improved immune activity against target tumor cells.

Discussion
Recently, the macrophage and genetically engineered CAR-macrophage have drawn lots of attention because of its potential advantages in homing and in ltrating to solid tumor, and manifold roles in regulating the immunosuppressive tumor microenvironment 36,82 .More excitingly, the rst clinical study in patients with solid tumor has started 83 , and the proof-of-concept results will justify CAR-macrophage as a new weapon to combat solid tumor.
The above achievements are all based on the 1st generation of CAR-macrophage, in which the intracellular signaling is mediated by a CD3ζ domain or a substitute effector domain from macrophage 29,31,49,84,85 .Like the 1st generation of CAR in T cells, this CAR mainly confers the effector function of phagocytosis, or the "signal 1" for CAR-macrophages.In order to integrate a "signal 2" like the co-stimulatory CD28 or 4IBB intracellular domain (ICD) in CAR-T 1,86−92 , we added a TLR4 intracellular domain TIR into the CAR, and thus provided an orthogonal "signal 2" of macrophage polarization.We demonstrated that the iPSC-derived, 2nd generation of EGFRvIII CD3ζ-TIR-CAR-macrophage against antigen-expressing tumor cells exhibited enhanced anti-tumor activity both in vitro and in vivo, with a particularly strong M1 polarization phenotype.
It is well established that macrophages are equipped with receptors called pattern recognition receptors (PRRs) that recognize "non-self" pathogen-or damage-associated molecular patterns (PAMP or DAMP), and then release pro-in ammatory or anti-in ammatory cytokines to exert various functions in defense 59,65 .The TIR-containing chimera antigen receptor is a novel engineered PRR which recognizes "antigen-associated molecular patterns", and enables macrophage with antigen-dependent capacity of polarization, in this case to be more pro-in ammatory to aid immune cell therapies in cancer.The opposite way of anti-in ammation can also be exploited to combat autoimmune diseases.These novel engineered macrophages with T cell and B cell function modules further blur the boundary between different immune cell types, and with the iPSC platform of high engineering feasibility, more and more payloads can be added onto this synthetic immune cell type to achieve better anti-tumor or other functions.Stable overexpression of CARs in iPSCs.CAR sequence-containing lentiviral vector, psPAX2, and pMD2.G were transfected into HEK293T cells cultured in 10 cm dishes to produce Lentivirus using Lipofectamine 2000 (Invitrogen, #11668).Medium was removedafter 6 hours of transfection to remove transfection reagent.Afterwards, lentivirus containing medium was harvested at the time point of 24h, 36h, 48h, ad 60h post transfection, and ltered with a 0.45 µm lter.The ltered medium was mixed with half volume of 30% PEG8000, and concentrated at 4000 rpm for 30 min at 4˚C.The supernatant was discarded, and the pellet was resuspended with mTeSR medium (STEMCELL Technologies, #85851) containing 10ug/ml polybrene (Yeasen Biotech, #40804ES76).The resuspended lentivirus was then added to prepared iPSCs in 6 well plate.The medium was refreshed after 6 hours post infection with new mTeSR medium.The infected cells were cultured for 24 hours, and then treated with 0.5 µg/ml puromycin for at least 48 hours to select successfully transduced clones.Subsquently, the CAR-expressing iPSCs were then expanded.

Methods
Macrophage differentiation from iPSC.WT-iPSCs and CAR-expressing iPSCs were treated with TrypLE (Gibco, #12604-021) for 1 min to disaggregate into individual cells.The cells werer collected and transferred to low-attachment plates (Corning, #3471) to allow formation of embryoid bodies (EBs) in mTeSR medium (STEMCELL Technologies, #85851) supplemented with Rock inhibitor Y27632 (STEMCELL Technologies, #72304).The following process of formation of embryoid bodies and macrophages differentiation from embryoid bodies has been described in our previous work.
Construction of genetic tumor cells.We rst synthesized coding sequence of EGFRvIII in GenScript Biotech, and then cloned the sequence into the Lenti-EF1α-T2A-EGFP-Puro vector.Subsequently, lentivirus containing EGFRvIII coding sequence was harvested from 293T cells, and it was used to infect U87MG cells to develop the U87MG EGFRvIII cell line.
Detecting of cytokines.We detected the cytokines released by CAR-iMACs with ELISA Kit (Elabscience).
In vivo anti-tumor assay.Four-weeks old NOD.CB17Prkdc SCID IL12fg rm1 /Bcgen (B-NDG) mice were obtained from Jiangsu Biocytogen Co., Ltd., and maintained under pathogen-free conditions in the animal centre of Zhejiang Academy of Medical Sciences with Committee-approved protocols.5×10 5 or 4×10 5 re y luciferase (FFluc + ) gene-expressing U87MG EGFRvIII cells mixed with 0.1% BSA in phosphate buffered saline (PBS) were injected intraperitoneally into B-NDG mice.Two hours after the tumor cells injection, 5×10 6 or 6×10 6 DiR (Meilunbio, #MB12482)-dyed CAR-iMACs (Truncated CAR-iMACs, CD3ζ-CAR-iMACs, TIR-CAR-iMAC, and CD3ZETA-TIR-CAR-iMACs) were injected into the same position in abdominal cavity of tumor burdened mice after being mixded with 0.1% BSA contained PBS.Another two hours after CAR-iMACs injection, D-luciferin (GOLDBIO, #LUCK-1G) was injected intraperitonelly into the mice with 150mg/kg dosage, following with living imaging through animal imaging system (in-vivo Xtreme) to capture the bioluminescence signal.Following animal imaging was conducted at the same time point on day 1, day 3 and day 7 to monitor the tumor growth.During tumor therapy by CAR-iMACs, the survival times of the mice was recorded, and statistics of survival curves was determined.
Immuno uorescence.The immuno uorescence (IF) assay was conducted to detect expression of EGFRvIII in U87MG EGFRvIII cell line.After being infected by lentivirus expressing plasmid Lenti-EF1α-EGFRvIII-T2A-Puro, U87MG cells were passaged to 0.1% gelatin coated glass slides in 24-well cell culture plate.When the cells grew to cover about 80% of the area of well, discarding mediums and washing the cells with PBS slightly thrice.Afterwards, xing the cells with 4% paraformaldehyde (PFA) for 30 minutes, and then discarding the PFA.Repeating the operation three times.Cells were then permeabilized with 0.5% Triton X-100 for 30 min at room temperature.Subsequently, cells were incubated with primary antibody against EGFRvIII (CST, #D6T2Q.1:200) at 4°C overnight.Discarding the primary antibody and Washing the cells with PBS slightly thrice followed with incubating the cells with secondary antibody (Alexa Fluor® 647 donkey anti-rabbit IgG.abcom, ab150075.1:1,000) for 1 hour in 37℃.Afterwards, the glass slides were mounted with anti-fading mounting medium (with DAPI) (Solarbio, #S2110).Cells were detected by Zeiss LSM800 uorescence microscope at a 63×oil objective.LSM800 with the Airyscan module was used to capture high-resolution pictures.
For detecting of NF-κB/P65, we rst co-incubated EGFP-marked CAR-iMACs with tdTomato + U87MG EGFRvIII cells for 4 hrs, 10 hrs and 24hrs respectively in 0.1% gelatin coated glass slides in a 24-well cell culture plate.Afterwards, processing the slides as described above, and incubating the cells with Alexa Fluor®647 conjugated rabbit antibodies against NF-κB/p65 (CST, #D14E12.1:100) at 4°C overnight.After being washed with PBS slightly thrice, the glass slides were mounted with anti-fading mounting medium (with DAPI) (Solarbio, #S2110).Cells were detected by Zeiss LSM800 uorescence microscope at a 63×oil objective.LSM800 with Airyscan module was used to capture high-resolution pictures.
RNA isolation and quantitative real-time PCR (qRT-PCR).For the qRT-PCR, total RNA extraction was performed using Eastep Super Total RNA Extraction Kit (Shanghai Promega Bio, #LS1040) according to the manufacturer's descriptions.cDNA was prepared using HiScript II Q Select RT SuperMix for qPCR with gDNA wiper (Vazyme, #R223-01).Gene expression was analyzed in triplicate using HiScript II One Step qRT-PCR SYBR Green Kit (Vazyme, #Q221-04) and Bio-Rad PCR machine (CFX-96 Touch).The difference in cycle threshold values (ΔCT) of all the genes tested were normalized to the ΔCT of GAPDH (glyceraldehyde-3-phosphate dehydrogenase), and the fold change in expression was expressed relative to WT iMACs.The primer sequences used in qRT-PCR assay were listed in Supplemental table.
Immunoblotting (IB).The IB was performed with conventional SDS-PAGE.Wild-type and EGFRvIIIexpressing lentivirus infected U87MG Cells were collected and washed twice with PBS, followed with lysing by RIPA buffer (Beyotime, #P0013K) containing protease inhibitor (Roche, #04693124001) on a shaker at 4°C for 20 minutes.Cellular debris was removed by centrifugation at 4°C.The supernatant was harvested and mixed with protein loading buffer that contains SDS, and heated at 90-100°C for 10 minutes.Proteins were resolved by SDS-PAGE and transferred to polyvinylidene di uoride membranes.
Fluorescence-activated cell sorting (FACS) analysis.To verify the cytotoxicity of CAR-iMACs against tumor cells, we incubated EGFP-marked CAR-iMACs with tdTomato + U87MG EGFRvIII cells in the case of E/T ratio of 3/1, 5/1 and 10/1 for 10 hours and 24 hours respectively in vitro.The cells were digested using 0.25% trypsin-EDTA (meilunbio, #MA0233).The dissociated cells were resuspended with 0.1% BSA, and then ltered into individual cells using a 300-mesh.Flow cytometry was performed on Beckman CytoFLEX LX (Version 9).The PE uorescence channel was chosen to detect tdTomato + U87MG EGFRvIII cells.The FITC channel was used to recognize EGFP-marked CAR-iMACs.FACS data was collected using CytExpert (Version 2.3), and was processed using FlowJo (Version 9).
To make sure the dynamic change of polarization state of the CAR-iMACs in the presence of tumor cells.
We rst incubated EGFP-marked CAR-iMACs with U87MG EGFRvIII cells in the case of E/T ratio of 3/1 for 1 day, 2 days, 3 days and 7 days respectively in vitro.After that, we collected the cells as performed as above, and equally divided them into two groups.Cells from one group were dyed with PE uorophore conjugated antibodies recognizing human CD80 protein (PE-CD80) (Biolegend, #305208).Another group of cells was labelled with human CD163 speci c antibody conjugated with PE (PE-CD163) (Biolegend, #333605).Then Flow cytometry was performed to detect CD80 + and CD163 + cells on Beckman CytoFLEX LX.The FITC channel was used to recognize EGFP-marked CAR-iMACs.PE uorescence channel was chosen to detect CD80 + and CD163 + cells from CAR-iMACs.FACS data was collected using CytExpert, and was processed using FlowJo.
Analysis of polarization state of CAR-iMACs in vivo.To analyze the polarization state of CAR-iMACs in vivo, we rst established U87MG EGFRvIII cells derived tumor model mice using intraperitoneal injection.After 2 days of CAR-iMACs treatment, we isolated CAR-iMACs from bdominl cvity of tumor-bearing mice using dissociative buffer containing 0.02% collagenase IV (Solarbio, #C8160), 0.01% hyaluronidase (Meilunbio, #37326-33-3) and 0.002% DNAase (Meilunbio, #9003-98-9).We then prepared two equal groups of the cells and stained either of the groups of cells using PE-CD80 antibody and PE-CD163 antibody.Then Flow cytometry was performed to detect CD80 + and CD163 + CAR-iMACs on Beckman CytoFLEX LX.The FITC channel was used to recognize EGFP-marked CAR-iMACs.PE uorescence channel was choosed to detect CD80 + and CD163 + cells from CAR-iMACs.FACS data was collected using CytExpert (Version 2.3), and was processed using FlowJo (Version 9).
Generation of single cell GEMs and sequencing libraries.We established single cell GEMs and sequencing libraries using the methods in our previous work 28 .10000 cells (90-95% viability) were captured per sample using a 10X Chromium device using 10X V2 Single Cell 3' Solution reagents (10X Genomics, Inc) at a concentration of 1000 cells/µl.Experiment was performed according to manual instructions.After the GEM-RT incubation, barcoded-cDNA was puri ed with DynaBeads cleanup mix, followed by 10-cycles of PCR ampli cation (98°C for 3 min; [98°C for 15 s, 67°C for 20 s, 72°C for 1 min] x 10; 72°C for 1 min).The total cDNA of single-cell transcriptomes was then fragmented, double-size selected with SPRI beads (Beckman), followed by 12 cycles sample index PCR ampli cation (98°C for 45 s; [98°C for 20 s, 54°C for 30 s, 72°C for 1 min] x 10; 72°C for 1 min), subsequently another double-size selection with SPRI beads was performed before sequencing (Illumina NextSeq platform).
Public bulk RNA-sequencing data mapping and analysis.Publicly available bulk RNA-seq dataset of iPSC and macrophage cells were downloaded from two previous studies 93,94 .All publicly available bulk RNAseq reads were rst trimmed using Trimmomatic (Version 0.36) software with the parameters "ILLUMINACLIP: TruSeq3-PE.fa:2:30:10LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:36" and were further quality-ltered using the FASTX Toolkit (Version 0.0.13,http://hannonlab.cshl.edu/fastx_toolkit/)with the minimum quality score 20 and minimum percent of 80% bases that has a quality score larger than this cutoff value.The high-quality reads were mapped to the GRCh38 genome by HISAT2, a fast and sensitive spliced alignment program for mapping RNA-seq reads, with -dta parameter.PCR duplicate reads were removed using Picard tools (v2.18.2) and only uniquely mapped reads were kept for further analysis.The expression levels of genes were calculated by StringTie (Version v1.3.4d, with -e -B -G parameters) using Release 28 (GRCh38.p12)gene annotations downloaded from GENCODE data portal.To obtain reliable and cross-sample comparable expression abundance estimation for each gene, reads mapped to the reference genome were counted as TPM (Transcripts Per Million reads) based on their genome locations.We used "prcomp" and pheatmap function with default parameters for Principal Component Analysis (PCA) and hierarchical clustering, respectively.We used high con dent genes characterizing the M1 and M2 states (with the maximum of TPM values among different samples larger than 1) for PCA and hierarchical clustering analysis.
Single cell RNA-sequencing data analysis.The "cellrange count" program, a subcommand included in 10X single-cell gene expression analysis pipeline (https://support.10xgenomics.com/single-cell-geneexpression/),was used to produce gene-cell barcode expression matrix.The single-cell gene expression matrix further was analyzed with Seurat v3.2.1 (https://satijalab.org/seurat/).To guarantee those quality of genes and cells used for downstream analysis, we excluded the genes with expressed cell number smaller than 3 and those cells with expressed genes smaller than 200 or larger than 30000.We also ltered the cells with the expression percentages of mitochondrial genes larger than 0.2.We then adopted 10 Principal Components (PCs) for tSNE and clustering analysis with the cluster resolution of 0.25.
The known cell identities (types) of each cell cluster of gene expression data were further predicted by an entropy-based predictor with default parameters (http://scibet.cancer-pku.cn/)(SciBet v1.0).To perform a comprehensive annotation of well-known cell types, weusedd an atlas database of cell types during human fetal liver haematopoiesi 95 , and integrate them with 30 major human cell type databases from 42 single cell RNA-seq datasets 96 .To perform Principal Component Analysis (PCA) and hierarchical clustering analysis on bulk RNA-seq data, we have grouped and summed up the normalized expression levels of our single-cell gene expression data to produce the synthetic bulk RNA-seq dataset.We used "prcomp" and pheatmap function with default parameters for PCA and hierarchical clustering, respectively.All the bioinformatic data analyses and resulting visualization were performed in R software   The CARs mainly comprise an extracellular signal peptide, a single-chain fragment variable (scFv) recognizing EGFRvIII, a transmembrane (TM) domain from CD8α, and either without an intracellular domain (truncated CAR), or with an intracellular CD3ζ signaling domain (CD3ζ-CAR), or cytoplasmic TIR domain from TLR4 (TIR-CAR), or both CD3ζ and TLR4 domains in tandom (CD3ζ-TIR-CAR).Hinge was added between scFv and CD8α transmembrane domains to endow the CARs with exibility to target antigens.All the CAR transgenes were linked to EGFP expressing sequence via a T2A element.b, Overview of the differentiation process of CAR-iMACs from the CAR-expressing iPSCs.EGFP-labeled CAR-iPSCs were differentiated into CAR-iMACs by mimicking the in vivo process of mesoderm induction, hematopoietic stem cell speci cation and myeloid cell production.c, Flow cytometry analysis was performed to determine the transduction e ciency of CAR expression in iMACs through counting the percentage of EGFP-expressing cells.d, A heat map from RNA-sequencing shows dynamic changes of the expression levels of TLR4 signaling related modulator genes along with the differentiation process from iPSCs to mature iMACs.The samples were taken from undifferentiated iPSC population, EBs of day 2 and day 7, and iMACs at day 18 and day 28.e, Comparison of expression level of TLR4 pathway-related adaptor and signal transduction genes in the in vitro cultured unstimulated iMACs, LPS/IFN-γ-polarized M1 iMACs and IL-4/IL-10-polarized M2 iMACs Construction of CAR.We designed T-CAR or M-CARs according to the intracellular CD3ζ domain from T cell receptor or TIR domain derived from TLR4, integrating hinge and transmembrane region of human CD8α (amino acids 183-206).The intracellular fragment of the CAR is humanized scFv nucleotide sequence speci c to EGFRvIII (139 scFv nucleotide sequence.Patent No.: US 9,394,368 B2).Cytoplasmic part of the human CD3ζ (amino acids 52-164) was designed for T-CAR.The TIR domain derived from cytoplasmic portion of human TLR4 (amino acids 672-818) was designed for M-CAR.The integration of cytoplasmic parts of the CD3ζ and TIR were designed for the second generation M-CAR.All the CARs were synthesized by GenScript Biotech, and cloned into the EcoRI and BamHI sites of Lenti-EF1A-T2A-EGFP-Puro vector.

Figures
Figures

Figure 2 Both
Figure 2