C57BL/6J (B6) mice were obtained from Japan Slc Inc. (Hamamatsu, Japan). B6.SJL-ptprca (B6.SJL, Strain #: 4007) mice congenic at the CD45 locus (CD45.1⁺CD45.2⁻), CAG-EGFP mice (Strain#: 006567), Ccr2 ^-/- mice (Strain#: 004999) and Ccr2-creERT2 mice (Strain #: 035229) were obtained from Jackson Laboratories (Bar Harbor, ME, USA). Casp3^flox/flox mice and Casp7^-/- mice were generated as shown in Supplementary Figures 3A, B . Casp3-flox and Casp7 Knockout (KO) mice were generated using the CRISPR/Cas system as previously reported (19) with slight modifications. For Casp3-flox, two crRNAs were synthesized (Fasmac, Kanagawa, Japan) to target the 5’-CAACAACTCAAGTTAAGTAC-3’ and 5’-GTCTCATCTTGAGGCCAAGG-3’ sequences in introns flanking the Casp3 exon2 containing the start codon. The plasmid vector used to generate the floxed allele of Casp3 was constructed with the 5’- and 3’-homology arms of 1500 bases long loxP sequences (5’-ATAACTTCGTATAGCATACATTATACGAAGTTAT-3’) placed at the cleavage sites by Cas9 and restriction sites (SalI and NotI). For Casp7 KO, two crRNAs targeting the 5’-AAGCAGCCCACAGAACAGCT-3’ sequence in the intron between exons 4 and 5 and the 5’-TGTTTCCTTGCCCCGCCAAG-3’ sequence in the intron between exons 6 and 7 were synthesized (Fasmac) to delete exons 5 and 6. To generate the KO allele with a precisely designed DNA sequence, a 150-mer single-stranded oligodeoxynucleotide (ssODN) was synthesized (5’-CACCTCTTTATAGAGAGAGAAGAAAATGGAGGTTAGCCAACCTCTCTCTGCTGTCACAAAGCAGCCCACAGAACAAAGGGGACCTGGCTCAGTGGGCAAGCTACTTGCAGTACAAGCAATGAGGACCCAAGACCGAATCCCCAGCATCTA-3’ by Hokkaido System Science (Sapporo, Japan). A mixture of Cas9 protein (final concentration of 30 ng/μl, NEB), crRNAs (8.7 ng/μl each), tracrRNA (28.6 ng/μl, Fasmac), the plasmid vector (10 ng/μl) and the ssODN (15 ng/μl) was injected into the pronuclei of C57BL/6J zygotes produced by in vitro fertilization. Zygotes surviving after the injection were transferred into the ampullae of the oviducts of pseudopregnant female mice. After natural delivery or caesarean section, newborn pups were genotyped using PCR primer pairs (5’-GTAAGCTAACCGAGCCAATG-3’ and 5’-AGTCAGGTAGATCAGAGGTC-3’ for Casp3-flox, 5’-GACGGAGCACACCTTTAATCC-3’ and 5’-CAGTCTGTCGGAATTTGGAGC-3’ for Casp7 KO), restriction enzymes and Sanger sequencing. Founder mice with the expected genetic modifications were used for further experiments.

We performed an analysis of patients with relapse/refractory diffuse large B cell lymphoma (DLBCL), B-cell lymphoid leukemia (B-ALL), who were treated with CD19-targeted CAR T-cell therapy. We prospectively collected blood samples on CRS patients in Tokyo medical and dental university hospital. Samples were collected from 22 patients between July 2020 and April 2023 and 3 patients were excluded from the analysis of this study according to the reasons shown in Supplementary Figure 4E . Nineteen patients were treated with a lymphocyte-depletion regimen (Fludarabine 25 mg/m² + Cyclophosphamide 250 mg/m²: Flu+CY) from Day -5 to Day -3 and CD19-specific CAR-T cells (tisagenlecleucel, tisa-cel) (Novartis, Basel, Switzerland) were infused at Day 0. Blood samples were collected before CAR-T cell administration until Day 14. All patients were enrolled in this study with protocols approved by the Institutional Review Board of the Tokyo Medical and Dental University, and written informed consent was obtained from a legally authorized representative as per the Declaration of Helsinki. CRS grading was defined by clinicians blinded to the results of the analysis based on Penn CRS grading ( Supplementary Table 1 ) (20). Blood samples were collected in vacutainer tubes containing heparin. Sample tubes were delivered to the laboratory within 4 hours of the sample draw. Peripheral blood mononuclear cells (PBMCs) were purified via density gradient centrifugation and were processed for flow cytometry analysis. Lymphocyte separation solution (d = 1.077, d = 1.119) (Nacalai Tesque Inc., Kyoto, Japan) was used. The blood samples were placed at room temperature for 30–60 min after the blood collection and then kept on ice for all experimental steps until FACS analysis, except for density gradient centrifugation. Plasma samples were stored at -80°C.

Patients’ medical records were reviewed for laboratory data and outcomes. All data were collected from documentation in the medical records. Clinical laboratory studies, including complete blood count (CBC) with differential, C-reactive protein (CRP), lactate dehydrogenase (LDH) and fibrinogen (Fbg) were performed in the Tokyo Medical and Dental University Hospital.

After staining with specific antibodies, cells were analyzed using a FACS AriaTM III (BD Bioscience, Franklin Lakes, NJ, USA) and then by FlowJo software (Treestar Inc., Ashland, OR, USA). For sorting of monocytes, BM cells were stained with PE/Cy5-conjugated antibodies against TER119, B220, Ly6G, CD4, CD3e, CD8a (Thermo Fisher Scientific), NK1.1 and CD19 (Biolegend). After washing, cells were incubated with anti-Cy5 microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) and monocytes were briefly isolated by negative isolation with Auto MACS (Miltenyi Biotec). Monocytes were then stained with a specific antibody and the Ly6C^hi monocyte fraction was isolated using a FACS AriaTM III. For counting cell numbers, CountBright™ absolute counting beads (Thermo Fisher Scientific) were used.

Total mRNAs were reverse-transcribed to cDNAs and gene expression levels were determined using a Light Cycler 480 and SYBR Green I Master (04707516001, Roche Diagnostics, Basel, Switzerland). The values were normalized according to the expression of β-actin. Specific primers used for real-time PCR are shown in Supplementary Table 4 .

To measure cytokine production from monocytes, BM monocytes obtained from wild type (WT) mice before and 12 hours after lipopolysaccharide (LPS, L2880–100mg, Sigma-Aldrich, Saint Louis, MO, USA) treatment (5 mg/kg) were cultured overnight with RPMI-1640 containing 10% FBS in the absence or presence of LPS (100 ng/ml). Supernatants were collected and analyzed using a LEGENDplex™ Mouse Macrophage/Microglia Panel kit (BioLegend, San Diego, CA, USA). Casp3/7^Δmono and their littermate control mice were treated with 20 mg/kg LPS and plasma was collected 2 hours after the LPS injection to evaluate the levels of TNF-α, IL-6, IL-1β and IL-10 using ELISA kits (R&D Systems, Minneapolis, MN, USA and Biolegend). For blood samples of CRS patients, a LEGENDplex™ Human Inflammation Panel 1 kit (Biolegend) was used to measure the levels of cytokines.

CD11b⁺Ly6C^hiLy6G^- monocytes were obtained from the BM 12 hours after treatment (5 mg/kg). Cells were stained with Diff Quik (Sysmex, Kobe, Japan) and the diameters of nuclei in each population were evaluated using Image J.

Cells were obtained from the BM, peripheral blood, and spleen 2, 4, 6 and 12 hours after LPS injection (5 mg/kg). After staining cell surface markers for neutrophils and monocytes, cells were stained with a PE-conjugated anti-Annexin V antibody (Thermo Fisher Scientific, Waltham, MA, USA) and propidium iodide (PI) and were analyzed by flow cytometry. For the detection of cells expressing active caspase-3/7, cells were stained with CellEvent™ Caspase-3/7 Detection Reagents (Thermo Fisher Scientific) after the cell surface marker staining. For detection of cells expressing active caspase-1, cells were stained with a FAM-FLICA^(R) Caspase 1 Assay Kit (Immunochemistry Technologies, Davis, CA, USA) after the cell surface marker staining.

Monocytes, neutrophils, and B cells isolated from naïve WT mice were cultured with LPS (10–1000 ng/ml) and/or cytokines (50 ng/ml, Biolegend) at 37°C for 3 hours. For caspase-blockade, cells were pre-cultured with inhibitors for Caspase-1 (Z-YVAD-FMK), Caspase-3 (Z-DEVD-FMK), Caspase-8 (Z-IETD-FMK) or Caspase-9 (Z-LEHD-FMK) (1–100 μM) (R&D Systems) before the stimulation. After stimulation, live cell numbers were counted by flow cytometry using CountBright™ absolute counting beads (Thermo Fisher Scientific).

Mice were anesthetized by intraperitoneal injection of an anesthetic mixture of medetomidine, midazolam and butorphanol. The abdomen of each mouse was then shaved and a laparotomy was performed. The cecum was exposed and tightly ligated 1.0 cm from the distal end. The ligated cecum was then perforated twice with a 26G or 19G needle. The cecum was returned to the peritoneal cavity after gentle squeezing to extrude a small amount of feces from the perforated sites. The peritoneum was sutured, and the skin was closed using a clip. Twelve hours after CLP induction, the mice were sacrificed and the frequencies and absolute numbers of neutrophils and monocytes in the BM and the peritoneal cavity were determined using flow cytometry.

Carrier-free recombinant mouse TNF-α and IFN-γ (10 μg/mouse each) (Biolegend, endotoxin level: <0.1 EU/µg protein) were mixed and administered intraperitoneally to WT C57BL/6J mice after which the numbers, PI⁺ dead cell ratio, and active caspase3/7-expressing apoptotic cell ratio in monocytes and neutrophils obtained from the BM and peripheral blood were evaluated 6 hours after the cytokine treatment.

Mice were anesthetized by an intraperitoneal injection of an anesthetic mixture of medetomidine, midazolam and butorphanol, after which monocytes obtained from the BM of CAG-EGFP mice were transplanted into the tibias of recipient B6.SJL mice. After the intra-BM transplantation, mice were quickly recovered by an injection of atipamezole. The mice were then treated with LPS (5 mg/kg) 1 hour after the transplantation. Eighteen hours after treatment, the distribution, expression of monocyte markers and cell death of donor cells were analyzed using FACS. To examine the apoptosis in monocytes, monocytes sorted from the BM of naïve TLR4-deficient or sufficient mice were transplanted into the BM of B6.SJL mice. Then, LPS (5 mg/kg) was intraperitoneally injected into the recipient mice and the apoptotic cell ratio was determined in the BM donor monocytes 6 hours after the LPS-treatment.

To evaluate the migration capacity of monocytes toward CCL2, Transwell migration assays (3-μm pores, Corning, Pittsburgh, PA, USA, #3402) were performed. Monocytes (3×10⁵ cells) obtained from the BM before and after LPS treatment (5 mg/kg) were placed in the upper chamber and different doses of recombinant CCL2 (0, 2.5, 10, 50 ng/ml) (Biolegend) were added in the lower chamber. The cells were then cultured 12 hours at 37˚C with 5% CO₂. Cells that had migrated from the lower chamber were counted by FACS with cell counting beads (Thermo Fisher Scientific).

Western blotting was performed as previously described (21). Casp3/7^Δmono mice were treated with 100 mg/kg tamoxifen intraperitoneally for 5 days. Two days after the last injection of tamoxifen, monocytes (2–3×10⁵ cells) were sorted from the BM of WT mice and of Casp3/7^Δmono mice and were lysed in lysis buffer containing 1% Triton X-100, 20 mM Tris–HCl (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 10 μg/ml aprotinin and 10 μg/ml leupeptin. The lysates were incubated on ice for 15 minutes and the supernatants were collected after centrifugation at 15,000 rpm, 4 ˚C for 20 min. The supernatants were mixed with 1 volume of 2x Laemmli sample buffer and were separated by electrophoresis on SDS-PAGE gels (Biorad, Hercules, CA, USA). The proteins were transferred to Immobilon P membranes (Millipore, Darmstadt, Germany) using a wet transfer method. Western blotting was performed using specific antibodies against β-actin (Sigma-Aldrich), Caspase3 (Cell Signaling Technology, Danvers, MA, USA, #14220T) and Caspase7 (Cell Signaling #12827T) and the proteins detected by those antibodies were visualized using a chemiluminescence kit (Western Lightning Plus-ECL, PerkinElmer, Waltham, MA, USA).

Statistical analyses were performed using Microsoft Excel or Prism software version 3 (GraphPad, San Diego, CA, USA). A two-tailed Student’s t-test was used for statistical analyses of two-group comparisons. Multigroup comparisons were performed using one-way analysis of variance (ANOVA) followed by the Tukey–Kramer multiple comparisons test. The criterion of significance was set at p < 0.05. Results are expressed as means ± standard error of the mean (SEM). For analysis of survival of the high dose LPS challenge, each group was subjected to Kaplan-Meier analysis and comparisons were performed using the log-rank test. CRS grading was defined by clinicians blinded to the results of analysis based on Penn CRS grading ( Supplementary Table 1 ). Samples from 3 patients were excluded from the analysis as shown in Supplementary Figure 4E . Randomization of the groups was not performed. No statistical methods were used to estimate sample size.

To examine the kinetics of monocytes in acute systemic inflammation, we intraperitoneally administered a sublethal dosage of LPS (5 mg/kg) to WT C57BL/6 mice to induce sepsis. After LPS treatment, the upregulation of chemoattractants for monocytes (Ccl2, Ccl3, Ccl4, Ccl6, Ccl7 and Ccl8) and neutrophils (Cxcl1 and Cxcl2) was examined in various peripheral organs ( Supplementary Figure 1A ). Consistent with previous reports, major monocyte and neutrophil chemoattractants such as CCL2, CCL7, and CXCL1/2 were substantially elevated ( Supplementary Figure 1A ), which are driven by common transcription factors NF-κB and/or AP-1 both in mice and in humans (22–24). We then examined the kinetics of monocytes in the BM, peripheral blood, spleen, and peritoneal cavity of mice after LPS treatment and found that about half of classical monocytes (hereafter referred to as “monocyte”) and neutrophils were lost in the BM within 12 hours after sepsis induction ( Figures 1A–C ; gating strategy is shown in Supplementary Figure 1B ). Interestingly, the increase of monocytes was not observed in the blood and spleen until 72 hours after LPS treatment ( Figures 1A, B ), although neutrophils were greatly increased in the peripheral blood 12 hours after sepsis induction ( Figure 1C ). Thus, we investigated where the monocytes that were lost in the BM went. However, LPS injection did not increase the number of monocytes in most peripheral organs at 12 hours after treatment ( Figure 1D ). In some organs, such as the peritoneal cavity and kidney, an increase of monocytes was observed within 12 or 24 hours after LPS treatment ( Figures 1A, D ). Consistent with previous reports (25, 26), monocytes accumulated transiently in the lungs 2 hours after the injection of LPS. However, those effects were too small to explain the decrease of monocytes in the BM ( Figures 1A, D , Supplementary Figure 1C ). We transferred Ly6C^hi monocytes from CAG-EGFP mice (CD45.2) into the BM of B6.SJL mice (CD45.1) and examined the donor-derived monocytes in the BM and spleen 12 hours after LPS treatment ( Figure 1E ). Compared to naïve conditions, the LPS treatment significantly decreased the ratio of splenic monocytes/BM monocytes ( Figure 1F ), suggesting that BM monocytes do not distribute to the periphery at the early stage of LPS-induced sepsis. To further demonstrate that the LPS-induced monocyte reduction in the BM was unlikely due to the egress of monocytes from the BM, we induced LPS-induced sepsis in CCR2-deficient mice ( Figure 1G ). As the decrease of monocytes in the BM was observed even in CCR2-deficient mice 12 hours after LPS treatment ( Figure 1G ), the egress of monocytes from the BM seems not to be a cause of the monocyte reduction in the BM at the early stage of sepsis. To further examine whether LPS treatment altered the expression of monocyte markers, which would make monocyte identification difficult, we transferred BM-derived monocytes from CAG-EGFP mice into the BM of recipient mice and examined the expression of monocyte markers 12 hours after LPS treatment ( Figures 1E, H ). The expression of CD11b and Ly6C was maintained and Ly6G was not expressed in the donor-derived monocytes regardless of LPS-injection ( Figure 1H ), ruling out the possible alteration of monocyte markers.

To examine the reason for the disappearance of monocytes in peripheral organs after LPS treatment, we evaluated cell death in monocytes and neutrophils 12 hours after LPS treatment. Staining with propidium iodide (PI) and Annexin V revealed that the frequency of dead cells (PI⁺ Annexin V⁺) in monocytes was increased immediately after LPS treatment ( Figure 2A ), and the increase of cell death in monocytes was significantly higher than that in neutrophils ( Figure 2B ). When monocytes isolated from the BM of CAG-EGFP mice were transferred into the BM of recipient mice (as shown in Figure 1E ), the frequency of PI⁺ dead cells was increased in donor monocytes after LPS treatment ( Figure 2C ). Staining for active caspase-3/7 demonstrated that apoptosis is selectively induced in monocytes, but not in neutrophils or other leukocytes in the BM, spleen and peripheral blood after LPS treatment ( Figures 2D, E ), which suggests that monocyte apoptosis was induced in both the BM and the periphery. In contrast, the frequencies of active caspase-1-expressing cells were comparable between monocytes and neutrophils after LPS treatment ( Figure 2F ), suggesting that pyroptosis could not explain the selective disappearance of monocytes in peripheral organs after sepsis induction.

We next examined whether inflammatory stimuli cause the cell death of monocytes in vitro. We isolated Ly6C^hi monocytes from the BM of naïve WT mice and cultured them with LPS or with inflammatory cytokines for 3 hours ( Figure 2G ) and found that LPS and inflammatory cytokines such as IFN-γ and TNF-α but not IL-6, quickly decreased the number of live monocytes ( Figure 2H ). In contrast, the impact of the stimulation with LPS or with cytokines in the cell death of neutrophils and B cells was much lower than that of monocytes ( Figure 2I ), which suggests that the monocyte-specific cell death program is driven by infection/inflammation. Consistent with the low induction of apoptosis in B cells, we recently showed that the number of B cells was not decreased in the BM within 12 hours after a high dose LPS injection (27). To demonstrate how monocytes were killed by inflammatory stimuli, we pre-cultured monocytes in the presence of inhibitors for caspases-3, 8, 9 or 1 for 30 minutes and then stimulated them with LPS. Live cell counts revealed that only the caspase-3 inhibitor suppressed the cell death ( Figure 2J ), which indicates that monocytes are eliminated by apoptosis. Taken together, microbial and/or inflammatory signals eliminate monocytes by inducing apoptosis early after the onset of sepsis.

Half of the monocytes were eliminated by apoptosis early after sepsis, meaning that the other half of the monocytes survived ( Figures 1A, B ). We characterized the surviving monocytes in the BM. The monocytes from LPS-treated mice and from naïve monocytes showed kidney-shaped nuclei, a representative monocyte morphology ( Figure 3A ). However, the LPS-treated monocytes were significantly larger in size than the naïve monocytes ( Figure 3B ). Monocyte egress from the BM is dependent on CCR2, a receptor for CCL2, during infection (28). The BM monocytes from LPS-treated mice maintained their CCR2 expression even after LPS treatment ( Figures 3C–E ). However, they showed an impaired migration ability toward CCL2 compared to BM monocytes from naïve mice ( Figures 3F, G ), which suggested that surviving monocytes in the BM after sepsis had lost their capacity to migrate out of the BM.

M-CSF-mediated signaling promotes monocyte differentiation into macrophages, and LPS treatment induces ADAM17-dependent shedding of CD115, a receptor for M-CSF (29). In this context, CD115 expression on the surface of BM monocytes was largely abolished within 3 hours after LPS treatment ( Supplementary Figure 2A ). To examine the differentiation capacity of monocytes in vivo, we obtained BM monocytes before and after LPS treatment and transferred them into the peritoneal cavity of naïve B6.SJL mice, after which flow cytometry was used to examine the generation of F4/80⁺Ly6C^- macrophages from F4/80^-Ly6C⁺ monocytes ( Supplementary Figure 2B ). As expected, the monocytes obtained from LPS-treated mice showed impaired differentiation into macrophages ( Supplementary Figures 2C, D ).

Monocytes act as a source of cytokines during infections (17, 28). Thus, we evaluated the cytokine production ability of BM monocytes from naïve and from LPS-treated mice ( Figure 3H , Supplementary Figure 2E ) and found that BM monocytes from LPS-treated mice showed significantly lower production of TNF-α, a representative cytokine involved in sepsis induction, and CXCL1 ( Figure 3H ). In contrast, monocytes from LPS-treated mice, but not from naïve mice, were able to produce G-CSF ( Figure 3H ).

Together with the defects in migration, differentiation, and cytokine production, the surviving monocytes in the BM after LPS treatment are dysfunctional in terms of sepsis induction.

A previous report suggested that administration of high-dose LPS does not increase circulating monocytes, although low-dose LPS does that (26). However, the mechanism and physiological meaning of that are unknown. In this context, our results from monocyte cultures revealed that the stronger the inflammatory stimulus, the greater the apoptosis of monocytes ( Figure 2H ), which implies that severe inflammation more efficiently suppresses monocyte appearance in vivo. Using a cecal ligation and puncture (CLP) infection model, we induced mild and severe sepsis with 26G and 19G needles, respectively. Severe sepsis-induced mice showed significantly higher plasma levels of TNF-α compared to mild sepsis-induced mice ( Figure 4A ). In both settings, the numbers of monocytes and neutrophils were similarly decreased in the BM within 12 hours after CLP-induction ( Figure 4B ). However, the frequency and number of monocytes that migrated into the peritoneal cavity were significantly higher in mice with mild sepsis compared to mice with severe sepsis ( Figures 4C, D ). In contrast, the number of neutrophils was comparable between the severe and mild CLP settings ( Figures 4C, D ). As we observed in the LPS injection model ( Figures 2D, E ), the apoptotic cell ratio in BM monocytes of severe sepsis-induced mice was significantly higher than that of mild sepsis-induced mice ( Figures 4E, F ). Since it has been reported that the injection of low-dose LPS (20 ng/body) increases monocytes in the periphery (26), we also tested the impact of treatment with low-dose (20 ng/body) and high-dose (100 μg/body) LPS in the apoptosis of monocytes ( Supplementary Figure 2F ). Consistent with our results with the CLP model, low-dose LPS injection did not induce apoptosis, although high-dose LPS injection did induce apoptosis. Thus, monocytes are eliminated by apoptosis only when the inflammation is excessive.

A CCR2-deficiency improved the severity of sepsis and a supply of peripheral monocytes restores sepsis in CCR2-deficient mice (17), suggesting that the existence of monocytes in the periphery enhances systemic inflammation. To further examine if the apoptosis-mediated reduction of monocytes alters the magnitude of inflammation, we generated Casp7^-/-;Casp3^flox/flox;CCR2-CreERT2 (Casp3/7^Δmono ) mice ( Supplementary Figures 3A–C ). We injected tamoxifen (2 mg/mouse/day) into Casp3/7 ^Δmono mice and their littermate control mice (Casp3^fl/fl;Casp7^-/- mice) for 5 consecutive days ( Figure 5A ) and confirmed that monocytes from Casp3/7 ^Δmono mice after that treatment lacked Caspase-3 and 7 by western blotting ( Supplementary Figure 3D ). The frequency of Annexin⁺ PI^- early apoptotic cells was also successfully decreased in monocytes by the genetic ablation of Casp3/7 after LPS treatment ( Figures 5B, C ). On the other hand, the frequency of the Annexin-V⁺ PI⁺ fraction, which is a mixture of necrotic cells and late apoptotic cells, was not significantly changed ( Figures 5B, C ), which suggests that the monocyte death was induced by non-apoptotic cell death such as necrosis in Casp3/7 ^Δmono mice. Consistently, the blockade of apoptosis results in the induction of necroptosis (30–32). Indeed, the number of monocytes was equally reduced in control and in Casp3/7 ^Δmono mice ( Figure 5D ). In this context, there was no significant difference in the frequency of monocytes expressing active caspase-1, an indicator of pyroptosis, between control and Casp3/7 ^Δmono mice ( Supplementary Figure 3E ). Collectively, these results imply the necrosis-mediated clearance of monocytes in Casp3/7 ^Δmono mice. To evaluate the importance of monocyte death by apoptosis in fine-tuning the magnitude of inflammation, we examined blood cytokine levels 2 hours after LPS treatment and found that the levels of TNF-α and IL-6 were increased in Casp3/7 ^Δmono mice compared to control mice ( Figure 5E ). In contrast, the levels of IL-1β and IL-10 were comparable between the two groups ( Supplementary Figure 3F ). Reflecting the severe inflammation, the weight and size of the spleens in Casp3/7 ^Δmono mice were slightly increased ( Supplementary Figure 3G ). In addition, the mortality of Casp3/7 ^Δmono mice was significantly increased compared to control mice during sepsis induced by treatment with a lethal dosage of LPS (20 mg/kg) ( Figures 5F, G ). These results suggested that the apoptosis-dependent reduction of monocytes is crucial for fine-tuning excessive inflammation during sepsis.

In contrast to immune paralysis, which is induced at a later phase of sepsis (33), we showed that monocyte apoptosis at an early stage of sepsis protects the host from excessive inflammation, implying that this mechanism might work independently of microbial stimuli. To examine that hypothesis, TLR4-deficient or TLR4-sufficient monocytes were mixed with monocytes from TLR4-sufficient CAG-EGFP mice at a 1:1 ratio and were directly transplanted into the BM of B6.SJL mice. When we examined the ratio between EGFP⁺ and EGFP^- donor (TLR4-deficient or TLR4-sufficient) monocytes 12 hours after LPS treatment ( Figure 6A ), the TLR4-deficiency did not alter that ratio ( Figure 6B ). Next, we administrated LPS (5 mg/kg) into SJL mice that received an intra-BM transplantation of WT or TLR4-deficient monocytes ( Figure 6C ). Importantly, the TLR4-deficiency did not alter the ratio of Annexin V⁺PI⁺ dead cells, Annexin V⁺ PI^- early apoptotic cells ( Figures 6D, E ) or active caspase3/7-expressing cells in donor monocytes ( Figure 6F ). These results suggested that TLR4-mediated signaling is not required to trigger monocyte apoptosis in vivo and implied the importance of pro-inflammatory cytokine-mediated signaling. To directly test this in vivo, we treated WT mice with recombinant TNF-α and IFN-γ as reported previously (34) ( Figure 6G ), which caused monocyte cell death in vitro ( Figure 2H ), and found that the numbers of monocytes and neutrophils were significantly reduced in the BM 6 hours after the treatment ( Figure 6H ). Importantly, monocytes simultaneously disappeared from the peripheral blood, while the number of neutrophils was greatly increased ( Figures 6H, I ). At this time, the increases in dead cells, which were evaluated by PI staining, and caspase3/7-expressing cells were pronounced in monocytes compared with neutrophils and CD11b^- cells in the BM ( Figures 6J–L ). Collectively, these results suggested that a microbial stimulus is not required while a pro-inflammatory cytokine stimulation is sufficient to induce monocyte apoptosis, which resulted in the selective disappearance of monocytes in the periphery.

As pro-inflammatory cytokines are sufficient to induce apoptosis-mediated monocyte disappearance in mice ( Figure 6 ), we aimed to determine whether it is also observed in human patients during pathogen-unrelated systemic inflammation. To this end, we focused on cytokine-release syndrome (CRS), which is caused by the release of large amounts of cytokines accompanying an excessive immune response not only during systemic infection but also during CAR-T cell immunotherapy. It should be emphasized that CAR-T cell administration has the advantage of monitoring CRS from the starting point of inflammation, which is not feasible in patients with infectious diseases or other inflammatory diseases because the symptoms are already present at the time the patient is seen in the hospital. In the CAR-T cell-induced CRS model with humanized mice and human CAR-T cells, human monocytes act as a major source of cytokines (35). We monitored the kinetics of CD14⁺CD16^low classical monocytes, neutrophils, c-reactive protein (CRP), and fibrinogen (FBG) in the blood of 19 patients until day 14 after CAR-T cell administration ( Figure 7A , the gating strategy of classical monocytes and neutrophils is shown in Supplementary Figure 4A ). The severity of CRS was graded based on the Penn grading scale ( Supplementary Table 1 ). All patients who received CAR-T cell therapy developed CRS (n=19), but only four patients developed severe CRS (grade 3 and 4, 15.8%). Patient characteristics with each CRS grade were shown in Supplementary Table 2 . The peak of blood neutrophil count almost coincided with the peaks of the inflammation markers CRP and FBG ( Figure 7B ). In contrast, the peak of classical monocyte count was later than the peaks of CRP and FBG, implying the suppression of monocyte appearance in the periphery during systemic inflammation in humans. In this context, the duration between the peak of monocytes, but not neutrophils, and the peak of CRP or FBG became larger in patients with severe CRS (grades 3 and 4) compared to those with relatively mild CRS (grades 1 and 2) ( Figures 7C, D ). Thus, the more severe the inflammation, the more delayed the appearance of classical monocytes in the blood during CRS. Related to this, patients with higher counts of classical monocytes before CAR-T therapy had a more severe CRS after receiving CAR-T therapy (day 0) ( Figure 7E ). On the other hand, the blood levels of cytokines, neutrophils and clinical indicators of inflammation at day 0 did not affect the severity of CRS ( Supplementary Figures 4B–D ; Supplementary Table 3 ), implying that the counts of classical monocytes in the blood before starting CAR-T therapy may serve as a biomarker to predict the severity of subsequent CRS after the treatment. Importantly, compared to day 0, the counts of monocytes in the blood were significantly increased at the peak of CRS in grade 1/2 patients, whereas grade 3/4 patients showed a decrease of monocytes in their blood at that time ( Figures 7F, G ). Those results obtained from patients are consistent with results obtained from mice ( Figure 4 ), and the magnitude of inflammation is critical for changing monocyte behavior both in mice and in humans.