This study was approved by the Ethics Committee of Dongfang Hospital of Beijing University of Chinese Medicine (JDF-IRB-2022000101), and each patient signed the relevant informed consent form.

The patients in this study were all from the wards and Emergency department of Dongfang Hospital, Beijing University of Chinese Medicine. Patients with K. pneumoniae bloodstream infection sepsis in the acute stage (KPN_ACU) met the following criteria: 1. Blood culture positive for K. pneumoniae; 2. Third International Consensus Definitions of Sepsis and Septic Shock (Sepsis 3.0) (10); 3. Body temperature > 38°C; 4. Sequential Organ Failure Assessment (SOFA) score ≥ 2. Patients with K. pneumoniae bloodstream infection sepsis in the recovery stage (KPN_REC) met the following criteria: 1. Body temperature < 37°C for 24 consecutive hours; 2. SOFA score < 2. The healthy control (HC) group was comprised of patients from the Physical Examination Department of Dongfang Hospital of Beijing University of Chinese Medicine, that met the following criteria: 1. No infectious diseases; 2. Normal liver and kidney function; 3. Blood counts and inflammatory markers are normal.

Three different types of vacuum blood collection tubes (Greiner VACUETTE) were used for whole blood collection: 2ml EDTA (ethylenediaminetetraacetic acid) anticoagulant tube, 3ml sodium citrate anticoagulant tube and 3ml serum tube. All venous blood was collected on an empty stomach. In the HC group, blood was retrieved from fasting patients in the morning. Whole blood collection in the KPN_ACU group was performed within 24 hours after diagnosis of K. pneumoniae bloodstream infection sepsis, while in the KPN_REC group blood was collected within the first 24 hours during the recovery stage. The EDTA anticoagulant whole blood was immediately subjected to complete blood count and PBMCs isolation. The sodium citrate anticoagulant whole blood was centrifuged for 1500g for 10 mins and plasma was separated for D-dimer detection. The whole blood collected from the serum tube was centrifuged for 2000g for 10 mins and serum was separated and stored in the refrigerator (Haier Biomedical) at -80°C for later use.

Whole blood in the 2ml EDTA anticoagulant vacuum blood collection tube was inverted five times, and a complete blood count was performed by automatic hematology analyzer XN-2000 (Sysmex, Japan). D-dimer was detected in the plasma isolated from 3ml sodium citrate anticoagulant whole blood using the automatic coagulation analyzer CS 5100 (Sysmex, Japan). Serum C-reactive protein (CRP) and procalcitonin (PCT) were measured using the automatic electrochemical luminescence immunoanalyzer Roche 801 (Roche, Switzerland).

PBMCs were isolated from fresh EDTA anticoagulant whole blood samples using density-gradient centrifugation, following a previously established protocol (11). Briefly, whole blood was transferred to 50 mL centrifuge tubes (Cornig, USA) and diluted approximately 1:1 with 1×PBS at room temperature. The mixture was then layered on top of Ficoll-Paque Plus (GE Healthcare, USA) and centrifuged at 500g for 30min. The PBMCs layer obtained was transferred to a fresh 50 mL centrifuge tube and washed three times with 10ml RPMI-1640 (Gibco, USA). Cell viability and count were confirmed using an Automated Cell Counter TC20 (BIO-RAD, USA). The criteria for sample inclusion were a minimum of 1×10⁶ per sample and a cell viability exceeding 90%. Only preparations with high cell viability were subjected to sequencing library preparation.

Single PBMC cells were captured using a 10x Chromium Controller (10x Genomics, USA) and libraries were prepared following the guidelines provided in the Single Cell 3’ Next GEM V3.1 Reagent Kits User Guide (10x Genomics, USA). The captured PBMCs were loaded onto a Chromium Single Cell Chip (10x Genomics) as per the manufacturer instructions facilitating co-encapsulation with barcoded Gel Beads at a target capture rate of 8000 individual cells per sample. Once into the chip, captured cells were lysed and the released RNA was barcoded through reverse-transcription in individual single-cell gel beads within the emulsion (GEMS). Within each droplet, complementary DNA (cDNA) was generated and amplified through reverse-transcription on a T100 PCR Thermal Cycler (Bio-Rad, USA). Subsequently, cDNA concentration and quality were assessed using Qubit Fluorometer (Thermo Scientific, USA) and bioanalyzer 2100 (Agilent, USA), respectively. Following the manufacturer’s recommendations, scRNA-seq libraries were constructed and sequenced to a depth of 30,000-60,000 reads per cell on a NovaSeq 6000 platform (Illumina, USA).

The raw data was initially processed using Cell Ranger version 3.0.2, converting the BCL file into a fastq file. The feature barcode matrix was generated using the human genome reference version GRCh38. Then, we conducted data quality control with R software DoubletFinder (v.3.5.1) considering indicators such as Total Unique Molecular Identifier (UMI) counts, number of detected genes, expression rate of hemoglobin genes, expression rate of mitochondrial genes, and the minimum cells per gene. Cells meeting the following criteria were included in the analysis: 1. UMI counts less than 25,000; 2. detected genes between 500 and 5,000 (500 < nFeature_RNA < 5000) ; 3. less than 20% of mitochondrial genes; 4. less than 1% of hemoglobin genes; 5. minimum cells per gene>10. After completing the quality control process, low quality cells were removed, resulting in a data set consisting of higher-quality single-cell data (12, 13).

‘IntegrateData’ function from the Seurat3 package in R (v.3.5.1) was used to eliminate batch effects between samples (14). The primary approach involved dimensionality reduction through Canonical Correlation Analysis (CCA), followed by subspace cell computation based on the Mutual Nearest Neighbors (MNN) algorithm, and finally, anchoring. Highly variable genes were identified using the ‘FindVariablesFeatures’ function also in the Seurat3 package. A principal component analysis (PCA) was performed using the ‘RunPCA’function to reduce the dimensions that would be used in subsequent analysis, and Elbow Point was used for selection (9). The Elbow Point diagram was generated using a heuristic evaluation method, and the 9th principal component was determined as the inflection point. Finally, the Seurat3 package’s ‘RunUMAP’ function was used to further reduce and visualize the PCA results. The number of cell clusters in each case was calculated at a resolution of 0.2, resulting in a total of 14 clusters.

The Seurat3 package’s ‘FindAllMarkers’ function was employed to identify markers for each cell cluster (15). Cluster annotation was performed using typical markers associated with specific cell types. To validate cluster annotations, we utilized the R package SingleR and the human BlueprintEncodeData dataset. This involved comparing the transcriptome of each single cell with the reference dataset to determine cell identity. In a two-step process, we first used the R package SingleR for machine annotation. Subsequently, manual annotation was performed based on to the Cell Maker website. The final classification resulted in five main cell clusters: T cells, B cells, NK cells, monocytes, and dendritic cells (DCs). Each cell type was further divided into multiple subtypes.

The ‘FindMarkers’ function in R package Seurat3 was employed to analyze DEGs between distinct cell groups, and the p-value were adjusted using the Bonferroni correction method. Significantly DEGs were identified based on an adjusted p-value < 0.05. The statistical method used for analyzing DEGs in different groups of immune cells was Wilcoxon sign rank test.

DEGs derived from experimental data were mapped genes to Gene Ontology (GO) (16). The enrichment of each GO entry in the gene list was calculated by comparing the ratio of the observed number of genes to the expected number based on the random distribution of the background genome, using Fisher precision tests in SingleR. To control for multiple comparisons and reduce false positive results, Bonferroni correction was applied in SingleR.

We conducted a comprehensive pathway analysis for all major immune cell types employing Ingenuity Pathway Analysis (IPA). DEGs (False Discovery Rate [FDR] < 0.01) for each cell type were enriched. Using multiple manually curated pathways and FDR values, we performed a pathway activity analyses to assess whether significantly enriched pathways (FDR < 0.01) were activated or inhibited. Pathway module scores were determined for each cell type by aggregating DEGs across all groups (17). We utilized gene formation modules within DEGs, classified by biological processes or pathways of interest in GO databases, and select module genes based on actual enriched GO processes. The pathway module scores were calculated, and the significance of the scores for different modules was tested through the ‘AddModuleScore’ function of the Seurat3 package (18).

Stored serum samples were retrieved from the -80°C refrigerator and allowed to thaw at room temperature. Human inflammatory cytokines and chemokines were determined using a Cytokine/Chemokine/Growth Factor Convenience 45-Plex Human Panel 1 (Thermo Fisher, USA) with Luminex xMAP platform (Thermo Fisher, USA). Complement components C1q, C3 and B Factor were detected using the automated biochemical analyzer Cobas 701 (Roche, Switzerland). The assessment of SOD levels was performed using a colorimetric detection kit (Ruiyuan, China).

Continuous variables that fit the normal distribution are represented by mean and standard deviation (SD). Continuous variables that do not fit the normal distribution are represented by median and interquartile spacing. Categorical variables are represented by frequency and proportion. Continuous variables between the two groups were compared using either the Student’s t test or the non-parametric Mann-Whitney test. 2-sided p-value < 0.05 were considered statistically significant using Graphpad prism software, version 9.0.

To investigate the dynamic immune response of PBMCs in patients with sepsis caused by K. pneumoniae bloodstream infection, we assembled three distinct groups for the study. The K. pneumoniae acute stage group (KPN_ACU) comprised seven patients, aged 69-91 years, with positive blood cultures of K. pneumoniae. Among them, four women, and unfortunately, four patients in this group did not survive. For the K. pneumoniae recovery group (KPN_REC), which involved a follow-up of the acute stage group, 3 patients, aged 69-80 years (2 women), successfully recovered(8 to 15 days after determining KPN_ACU). Additionally, we included a healthy control group (HC) consisting of 3 individuals, aged 67-88 years (2 women), selected from the physical examination department. Peripheral blood and serum samples were collected from patients in the KPN_ACU and HC groups, with one sample per patient. For patients in the KPN_REC group, two sets of samples were collected: one during the acute infection stage and another during the recovery stage. Basic clinical data of the patients are provided in Supplementary Table 1. The collected samples underwent follow-up scRNA-seq, analysis of inflammatory indicators, and assessment of immune indicators to comprehensively study the immune response dynamics. First, we evaluated eight sepsis-associated indicators in the serum of patients and compared the results across different groups. The findings revealed significant differences among the groups. In comparison to the HC group, the KPN_ACU group, exhibit significantly elevated levels of PCT (p < 0.05) (Figure 1A), C-reactive protein ([CRP], p < 0.05) (Figure 1B), D-dimer (p < 0.05) (Figure 1C), white blood cell count ([WBC], p < 0.05) (Figure 1D), polymorphonuclear leukocyte percentage ([PMN%], p < 0.05) (Figure 1E), monocyte count ([Mono], p < 0.05) (Figure 1G), tumor necrosis factor alpha ([TNF-α], p < 0.05) (Figure 1H), and interferon-gama ([IFN-γ], p < 0.05) (Figure 1I), along with a significant decrease in lymphocyte proportion ([Lym%], p < 0.05) (Figure 1F). When comparing KPN_REC with KPN_ACU, a notable reduction was observed in the levels of CRP (p < 0.05) and PCT (p < 0.05) and Mono (p < 0.05), indicating a positive response to treatment. Other indicators such as D-dimer (p = 0.12), WBC (p = 0.27), PMN% (p = 0.18), TNF-α (p = 0.12) and IFN-γ (p = 0.21) showed a decreasing trend, while Lym% exhibited an increasing trend, though statistical significance varied. Importantly, when comparing KPN_REC with HC, there were no statistically significant differences in inflammation markers (CRP, PCT, IFN-γ), except for TNF-α, coagulation marker (D-dimer), and blood routine markers (WBC, PMN%, Lym%, Mono) in KPN_REC (p > 0.05) (Figure 1).

The above results indicate that the patients with K. pneumoniae bloodstream infection sepsis are predominantly elderly, facing high mortality rates, and experiencing significant alterations in immune cell proportions and inflammatory responses. The elevated D-dimer indicates abnormal blood clotting function, potentially leading to thrombosis. During the recovery stage, there is a significant decrease in inflammatory factors, thrombus levels, SOFA score, and body temperature, accompanied by a tendency for the immune cell proportion to return to normal. However, it is noteworthy that the recovery of inflammatory factors is notably faster than the restoration of the immune cell proportion. This suggests that there may still be some imbalance or disorder in the composition of immune cells among patients during the recovery stage.

Next, the PBMCs isolated from the blood of patients in the three study groups (HC, n = 3; KPN_ACU, n = 7; KPN_REC, n = 3) were subjected to scRNA-seq. We collected 122,374 high-quality cells, with an average of 9,413 cells per sample. After strict quality control, all samples’ cells were integrated following batch effect correction. They were then analyzed using unbiased clustering and uniform manifold approximation and projection (UMAP). All clusters were composed of cells from each sample, indicating a successful batch effect correction. We identified 14 clusters with distinct transcriptomic signatures and annotated cells based on canonical annotation marker genes of immune cells. These were defined as five typical immune cell types: T cells (CD3E⁺CD8A⁺IL7R⁺), NK cells (GNLY⁺NKG7⁺KLRB1⁺), monocytes (CD14⁺CD68⁺S100A2⁺), B cells (CD79A⁺MS4A1⁺), and DCs (CLEC4C⁺IL3RA⁺). The violin plots show good differentiation of different cells (Figures 2A, B, D). We observed changes in the proportion of various cell types among the three study groups, revealing dynamic changes during disease progression. Compared with HC group, the proportion of T cells significantly decreased in KPN_ACU group and slightly decreased in KPN_REC group. However, the proportion of T cells in KPN_REC group was slightly higher than that in KPN_ACU group. The proportion of NK cells was similar among the three groups. Monocytes in KPN_ACU group increased significantly and showed a trend of recovery in KPN_REC group. The proportion of B cells in KPN_ACU group significantly decreased, and remained not at a lower level in KPN_REC group. The proportion of DCs was also similar among the three groups (Figure 2C).

The results above evidence that K. pneumoniae bloodstream infection caused significant interference in the proportions of PBMCs. Different types of immune cell may be dysfunctional in both acute infection and recovery stages.

T cells, important immune cells in the body against bacterial infection, exhibited a decreased proportion during the acute stage of K. pneumoniae bloodstream infection sepsis and showed a recovery trend in recovery stage (Figure 2C). Therefore, the changes in T cell composition and immune function during the whole disease stage may reflect the severity of the disease and the progress of recovery. To further understand the biological functional changes involved in T cells response to K. pneumoniae sepsis, we performed a Gene Ontology (GO) enrichment analysis revealing differences in the T cells transcriptome between the three groups of study. Compared to the HC group, T cells from the KPN_ACU group exhibited upregulation in the antigen presentation and cytotoxicity pathways, indicating the activation of the bacterial defensive function of T cells (Figure 3A). On the other hand, the biological processes regarding response to cytokine, humoral immune response and inflammatory response are downregulated, suggesting that T cells are in immunosuppressant state at the same time (Figure 3A). Compared with the KPN_REC group, genes related to cytolytic granule production and T cell mediated cytotoxicity were upregulated in the T cells from KPN_ACU patients, while functions such as leukocyte migration and dendritic cell chemotaxis are downregulated (Figure 3B). This indicates that the cytotoxicity of T cells was enhanced during the acute infection stage, but their auxiliary immune function was inhibited to a certain extent. Nevertheless, with the recovery of patients, the auxiliary immune function also gradually recovered. Compared to the HC group, the cytolytic granule production, apoptotic process, MHC class II receptor activity, inflammatory response, and adaptive immune response of T cells from the KPN_REC group were upregulated, while the cytokine-mediated signaling pathway and C-C chemokine receptor activity were downregulated (Figure 3C), indicating that the patients were gradually recovering, but not fully recovered.

We also studied the expression of immune function genes in T cells and found that T cell exhaustion-related markers, including PDCD1, TIGIT, LAG3, and PRDM1, were significantly upregulated in the KPN_ ACU group compare to T cells from HC group. However, these markers were not downregulated in the T cells from the KPN_ REC group (Figure 3D), indicating that the T cell exhaustion that occurred during acute infection did not fully recover during the recovery stage. The apoptosis-related markers TNFRSF1B, CASP1, PRF1, GZMB, CTSW, CTSD and CTSC were significantly upregulated in the T cells from both KPN_ ACU and KPN_REC groups compared to HC group, indicating that T cell apoptosis, which occurred during acute infection, did not fully recover during the recovery stage (Figure 3E). These findings support our previous observation of a lower proportion of T cells in the KPN_REC group compared to the HC group. The immunosuppression of sepsis mainly manifests as lymphocyte apoptosis and exhaustion, especially T cells (19), as confirmed by the increased expression of apoptosis related genes. Similarly, T cell cytotoxicity markers CTSW, GZMA, GZMB, FCGR3A, NKG7, GNLY, CTSC, KLRK1, and GZMH were found significantly upregulated in both KPN_ ACU and KPN_REC groups (Figure 3F), indicating a significant increase in T cell killing ability during acute infection stage that persists into the recovery stage.

We performed a subpopulation cluster analysis of T cells to study their dynamics at a finer resolution. Furthermore, we clustered the T cells into three subpopulations: CD4+ T cells (CD4⁺CCR7⁺), CD8+ T cells (CD8A⁺GZMA⁺), and STMN1+TUBA1B T cells (STMN1⁺TUBA1B⁺) (Figures 3G, H), to study their dynamics at a finer resolution. We observed a significant decrease in the proportion of CD4+ T and a simultaneous increase in the proportion of CD8+ T cells in the KPN_ACU group compared to the HC group (Figure 3I). This suggests that the overall decrease in T cell proportion was primarily attributable to the decrease in CD4+ T cells. CD4+ T cells are helper T cells, whose main function is to secrete cytokines, chemokines, and recruit target cells (20). The decrease of CD4+ T cells proportion in KPN_ACU group indicates that the ability of the body to rapidly initiate immune response is reduced, and the immunosuppressive state of CD4+ T cell subtypes is the main cause of T cell immunosuppression. CD8+ T cell is a cytotoxic T cell, whose main function is to clear cells infected by pathogens (19). The proportion of CD8+ T cells in KPN_ACU group increased (Figure 3I), indicating that there were more infected cells. The decrease of CD4+/CD8+ T cells ratio indicates that the patient’s immune function is decreased. During the recovery stage, the killing ability was still high, and the killing ability was mainly derived from CD8+ T cells, which was consistent with the increase in the proportion of CD8+ T cells (Figures 3F, I). Genes related to MHC-II antigen presentation were significantly upregulated in T cells from KPN_ACU group. Although these genes were downregulated in T cells from individuals in the KPN_REC group, their expression levels remained higher than those in T cells recovered from the HC group. MHC-II antigen presentation genes were upregulated mainly in CD8+ T cells (Supplementary Figures S1A–C). These results indicate that CD8+ T cells play a more important role in antigen recognition of K. pneumoniae infection defense. STMN1+TUBA1B T lymphocytes also highly expressed CD8A and GZMB markers, indicating that STMN1+TUBA1B T cells were cytotoxic. STMN1+TUBA1B T cells accounted for a low proportion of T cells, that was found significantly higher in the KPN_ACU group, decreasing to the level of the HC group in the KPN_REC group (Figure 3I). Detection of STMN1+TUBA1B T cells may be able to predict recovery from sepsis. STMN1+TUBA1B T cells have downregulated biological processes of antigen presentation and B cells activation during acute stage compared to HC group (Figure S1D; Supplementary Table 2), indicating that STMN1+TUBA1B T cells have different functions to CD8+ T cells and CD4+ T cells, but their specific functions need further study.

In summary, T cells were characterized by apoptosis and exhaustion, and the helper function of CD4+ T cells was decreased, and the cytotoxicity and antigen presentation function of CD8+ T cells were enhanced.

Monocytes are the largest immune cells in peripheral blood, which have important immune functions such as phagocytosis and antigen presentation (17). Compared with their counterparts in the HC group, the proportion of monocytes in KPN_ACU group was significantly higher. Although the proportion of monocytes in KPN_REC group was lower, it also remained slightly higher than that in HC group (Figure 2C). Additionally, the proportion of monocytes lowered with the decrease of inflammatory indicators (Figures 1, 2C), suggesting that monocytes are an important part of the body’s defense system against K. pneumoniae. DGEs analysis showed that compared with HC group, the expression of genes related to phagocytosis, such as CLEC7A, NCF2, CD14, PECAM1, FGR, CORO1C, PYCARD, MYD88, ITGB1, CALR, FCGR1A and SYK, was increased in the monocytes from the KPN_ACU group (Figure 4A), indicating that the monocytes phagocytosis function of the patient is activated after infection with K. pneumoniae. These results were further supported by the GO enrichment analysis that showed an upregulation of the phagocytosis pathway in monocytes from the KPN_ACU group (Figure 4B). On the other hand, the expression of the phagocytosis genes was decreased in KPN_REC group compared to KPN_ACU group but still higher than that in the HC group (Figure 4A). Furthermore, the phagocytosis pathway (positive regulation of phagocytosis) was also observed to be downregulated in the KPN_REC group compared to the KPN_ACU group, but not as much as in the HC group (Figure 4A; Supplementary Figures S2A, B).

The levels of interleukin-1β (IL-1β) interleukin-18 (IL-18) and interleukin-1 receptor antagonist (IL-1RA) in serum were significantly increased in KPN_ACU group compared to the HC group and, in the case of IL-1RA, also compared to the KPN_REC group (Figures 4C, D, E). The PBMCs scRNA-seq data showed that IL-1β, IL-18 and IL-1RA genes (IL1B, IL18, IL1RN) were mainly expressed in monocytes (Supplementary Figures S2C, D, E), indicating that among the PBMCs, IL-1β, IL-18 and IL-1RA in infected patients were mainly produced by monocytes in PBMCs. Similarly, compared to the HC group, IL-1β, IL-18 and IL-1RA genes in KPN_ACU and KPN_REC were significantly upregulated (Figure 4D; Supplementary Figures S2D, E), indicating that patients in acute stage had obvious inflammatory response, while patients in recovery stage had not fully recovered. IL-1RA is an immunosuppressive cytokine that inhibits IL-1β activity by binding to IL-1R1 (21), indicating that the anti-inflammatory genes of monocytes are also activated, and there is an interaction of pro-inflammatory and anti-inflammatory functions in the patient. In KPN_ACU group, IL-1β content in serum was significantly increased (Figure 4C), which could promote the antigen presentation ability of APC such as monocytes (22).The expression of monocytes antigen presentation genes was downregulated in the monocytes of the KPN_ACU group (Supplementary Figure S2F), and the GO enrichment analysis showed that multiple biological processes related to antigen presentation (peptide antigen assembly with MHC class II protein complex, MHC class I protein binding) were also downregulated (Figure 4B). This suggests that the ability of IL-1β to promote monocytes antigen presentation maybe inhibited by IL-1RA.The high expression level of TNF-α receptor TNFRSF1B in monocytes (Supplementary Figure S2H) suggests that TNF-α could induce the production of IL-1β in this cell type. In addition, TNF-α activates phospholipase A2 and releases superoxide, causing DNA breakage (23). Our study also found a downregulation of biological process that would inhibit the production of reactive oxygen species in patients from the KPN_ACU group compared to HC group (Figure 4B). Additionally, the serum levels of superoxide dismutase (SOD) decreased in individuals from both KPN_ACU and KPN_REC groups (Figure 4F), suggesting that the ability to remove oxygen free radicals is reduced in patients with sepsis. At present, it is believed that diffuse intravascular coagulation and toxic shock caused by Gram-negative bacteria are caused by excessive TNF-α production stimulated by bacterial endotoxin (24). Therefore, the increase of D-dimer content maybe caused by excessive TNF-α in KPN_ACU (Figures 1C, H).

To further investigate the immune function of monocytes during K. pneumoniae bloodstream infection, we split the monocytes population into four subtypes that represent different functions: CD14+ monocytes (CD14⁺), HLA-DR+ monocytes (HLA-DRA⁺HLA-DQB1⁺), PF4+PPBP+ monocytes (PF4⁺PPBP⁺), FCGR3A+ monocytes (FCGR3A⁺CDKN1C⁺) (Figure 4G). They were no significant differences in the proportion of these cells between the different study groups, except for a significant increase in the numbers of FCGR3A+ monocytes in the KPN_ACU group compared to HC group (Figure 4H). Compared with HC group, CD14+ monocytes from the KPN_ACU group were observed to highly express the inflammatory and cell activation markers S100A9 and S100A8, indicating that CD14+ monocytes maybe play a major inflammatory function during acute infection stage (Supplementary Figure S2G). Similarly, CD14+ monocytes from KPN_ACU group were found to have significantly upregulated pathways related to bacterial defense and cytokines (Supplementary Table 3). HLA-DR+ monocytes from the KPN_ACU group highly expressed antigen presentation-related markers MHC-II and CD74 (Supplementary Figure S2G), and the biological processes involved in multiple antigen processing and presentation monocytes were also significantly upregulated, evidencing their role as antigen presenting cells (Supplementary Table 3). As expected, the PF4+PPBP+ monocytes from individuals in the KPN_ACU group expressed the platelet-related markers PF4, PPBP (Supplementary Figure S2G), and the clotting pathway (hsa04610: Complement and coagulation cascades) and platelet activation were upregulated (Supplementary Table 3), evidencing that these monocytes are involved in coagulation or platelet immunity after K. pneumoniae infection. FCGR3A+ monocytes from the KPN_ACU group highly upregulated multiple lysosome biological processes, indicating that these cells mainly played a killing function (Supplementary Table 3).

In summary, monocytes are activated after K. pneumoniae infection, with increased cell proportion, enhanced phagocytosis function, and are able to interact with a variety of cytokines, and different subtypes play different functions. However, antigen presentation was downregulated. Therefore, monocytes play an important role in the defense of K. pneumoniae.

DCs is the main antigen presentation cell type in vivo, and changes in their function affect the efficiency of pathogen recognition by immune cells (25). Our study did not find any significant difference in the proportion of DCs between the different study groups (Figure 2C). We suspected DCs may have changed primarily in function rather than proportions. So, we detected the expression of DCs exhaustion markers (ENTPD1, CTLA4, TIGIT). These genes were significantly upregulated in the KPN_ACU group (Figure 5A), which could explain a slight downward trend in cell proportion. In the KPN_REC group, the expression levels of exhaustion markers were lower than those found in the KPN_ACU group but still higher than the levels in HC group (Figure 5A). This indicates that DCs exhaustion may not fully recover during the recovery phase. Furthermore, DEGs analyses showed a downregulation of DCs genes (HLA-DQB1, HLA-DPB1, HLA-DPA1, HLA-DQA1, HLA-DMA, HLA-DRB5) related to antigen presentation in KPN_ACU group. This suggests that the antigen presentation capabilities of DCs from these individuals is probably decreased (Figure 5B), impacting the recognition of K. pneumoniae. The average expression of antigen presentation-related genes in the KPN_REC group was higher than that in HC and KPN_ACU groups (Figure 5B), indicating that the antigen presentation capacity of DCs recovered rapidly during the recovery stage, possibly overcompensating the low DCs proportion during infection. This was also supported by the GO enrichment analysis (Figures 5C, D). So overall, DCs antigen presentation is reduced and there is some immunosuppression in patients during the acute infection stage, but rapid recovery during the recovery stage.

We divided the DCs population into 2 subtypes, plasmacytoid DCs ([pDCs], CST3⁺MSRB1⁺) and conventional DCs ([cDCs], TNFRSF17⁺ IGHA2⁺) (Figure 5E; Supplementary Figure S3A). Compared with HC group, the proportions of pDCs and cDCs in KPN_ACU group were similar, but the proportions of pDCs in KPN_REC showed an increasing trend, while the proportions of cDCs showed a decreasing trend (Figure 5F; Supplementary Figure S3B). The average expression of antigen presentation-related genes is downregulated in both pDCs and cDCs from KPN_ACU group. The recovery rate of antigen presentation-related genes in cDCs was faster than that in pDCs (Supplementary Figures S3C, D).

Cell surface chemokine receptor 2 (CCR2) is highly expressed in DCs (26). CCR2 expression was downregulated in DCs during acute stage but upregulated during recovery stage (Figure 5G). The serum level of monocyte chemoattractant protein-1 ([MCP-1], a chemokine recognized by CCR2) was increased in KPN_ACU group but decreased to normal levels in KPN_REC group (Figure 5H). These results suggest that overproduction of MCP-1 during acute stage maybe compensate for the lack of CCR2. Further DEGs analysis specifically looking at each of the DCs subtypes revealed that the CCR2 gene is mainly expressed on pDCs (Figure 5I), and gets downregulated during acute stage (Figure 5G; Supplementary Figure S3E).

In conclusion, DCs has weakened antigen presentation in patients with K. pneumoniae bloodstream infection sepsis. pDCs response to MCP-1 chemokine is weakened in the acute stage. In the recovery stage, the recovery rate of antigen presentation is faster than the proportion of DCs (Figures 2C, 5B).

B cells in PBMCs can be transformed into plasma cells secreting antibodies by pathogen stimulation, and participate in humoral immune response. B cells can also process and present antigens, and secrete cytokines such as interleukin, TNF, IFN, etc (27). B cell apoptosis was suspected based on the significantly low proportion of these cells during acute and recovery stages (Figure 2C). Correspondingly, the expression of apoptosis-related genes (BCL2A1, ZBTB16, PRF1, GZMB, CD70 and HRK) was increased during acute stage, while the expression levels of apoptosis inhibitors-related genes (BCL2, MCL1, NFKBIA) was observed to be decreased (Figures 6A, B). Furthermore, KEGG analysis showed that the apoptosis related pathway (hsa04210: apoptosis) was activated in the B cells during acute stage (Supplementary Table 4), and the GO enrichment analysis showed that the biological processes related to apoptosis were also upregulated in these cells (Figure 6C). This supports our suspicions of B cell apoptosis during K. pneumoniae-promoted sepsis, which may persist during the recovery stage. Furthermore, genes related to the production of interferon in B cells (MIX1, IFI6, IFITM3) were found to be downregulated in the KPN_ACU and KPN_REC groups compared the HC group (Supplementary Figure S4A). Compared with KPN_ACU group, the average expression of apoptosis genes was decreased and apoptosis inhibitor gene was increased (Figures 6A, B) and the apoptosis related pathway (hsa04210: apoptosis) was downregulated in KPN_REC group (Supplementary Table 4), but the cell proportion was similar to that of KPN_ACU (Supplementary Figure 2C), indicating that the recovery rate of apoptosis gene was faster than that of cell proportion. Antigen presentation is also an important immune function of B cells (28). On the contrary, the average expression of genes related to antigen presentation (HLA-DQB1, HLA-DQA1, HLA-DPA1, HLA-DMB, CIITA and CTSS) was increased in the B cells from KPN_ACU group (Figure 6D), indicating that the antigen presentation function of B cells was enhanced during acute stage. It is an important antigen presentation cell in the defense of K. pneumoniae.

CR1 is widely distributed on the surface of B cells, and its ligand is C3b, which is a regulator of B cell activation (29). On the other hand, the expression of complement receptor 1 (CR1) and complement receptor 5a (C5AR1) was found to be downregulated in the B cells during sepsis (Figure 6F), while the levels of complement 3 ([C3], Figure 6E), B factor and complement 1 q ([C1q], Supplementary Figure S4B) appeared decrease in the in serum of patients suffering acute stage too. Furthermore, this gene expression pattern and complement protein levels seems to remain low during recovery stage This suggests that complement activation is impaired in B cells during acute stage and continue to be damaged during recovery stage (Supplementary Figure S4B).

We divided B cells into three biologically relevant subtypes: B naive cells (TCL1A⁺IL4R⁺), B memory cells (AIM2⁺CD27⁺TNFRSF13B⁺), and CD2+IL32 B cells (CD2⁺IL32⁺) (Figure 6G; Supplementary Figure S4C). The proportion of the three cell subtypes did not change significantly between the three study groups. However, the CD2+IL32 B cells numbers showed a decreasing trend during acute stage and recovery stage (Supplementary Figure S4E). CD2+IL32 B cells is maybe a new type of B cell, which has not been reported yet. Even in depleting numbers, these cells displayed an upregulated expression of genes related to cell activation (FCRL1, TNFRSF13C, BANK1, SWAP70, MEF2C, CD22, CD79A, POU2F2) during acute stage (Figure 6H). The expression of some of those genes (FCRL1, TNFRSF13C, CD22, CD79A, POU2F2) continue to be upregulated during sepsis recovery stage too (Figure 6H). Similarly, the expression of antigen presentation-related genes (HLA-DQB1, HLA-DPB1, CTSS, HLA-DPA1, HLA-DMB, CIITA, HLA-DQA1, HLA-DRB5) was found to be upregulated in both KPN_ACU and KPN_REC groups (Figure 6I), with a higher antigen presentation module score in the recovery group (Supplementary Figure S4D). GO enrichment analysis showed that multiple biological processes of antigen presentation, such as antigen processing and presentation of exogenous peptide antigen via MHC class II, were upregulated in KPN_ACU group compared with HC group, and continue to be upregulated in the KPN_REC group (Supplementary Figures S4F, G), indicating that the antigen presentation function of CD2+IL32 B cells was activated, which maybe play an important role in the clearance of K. pneumoniae. These results demonstrate that CD2+IL32 B cells are important effector B cells. Unfortunately, we did not note plasma cells to analyze the antibody defense effect of K. pneumoniae.

In general, B cells were continuously decreased in KPN_ACU and KPN_REC groups, and interferon gene expression level and complement response were decreased, so defense function was decreased. However, a new B cell subtype has emerged, which has enhanced activation and antigen presentation function, and may play an important role in the recovery of K. pneumoniae.

NK cells are important cytotoxic cells in peripheral blood, which can kill bacteria and bacteria infected cells (30). Our study found that the proportion of NK cells did not change significantly among HC, KPN_ACU and KPN_REC groups (Figure 2C). NK cells were divided in two biologically relevant subtypes: FCER1G+CD160 NK cells (FCER1G⁺CD160⁺), TRGC1+TRDV2 NK cells (TRGC1⁺TRDV2⁺) (Figures 7A, B). Although no significant differences were observed in the proportion of the NK subtypes between the different study groups, we observed that the proportion of FCER1G+CD160 NK cells tended to decrease during acute stage and partially recover during the recovery stage. Conversely, TRGC1+TRDV2 NK cells showed an increase during acute stage, and a decreasing trend during recovery stage (Figure 7C).

The function of the NK cells seems to change following disease progression. During acute stage, the expression of cytotoxicity- related genes decreased (Figure 7D), while the expression of complement receptor- and immunoglobulin- related genes was increased and their pathways upregulated (Figures 7E, F, G). This suggests that the killing ability of NK cells during acute stage may be reduced, and their role may be more focused on complement activation and immunoglobulin secretion. During recovery stage, the expression of complement receptor genes and immunoglobulin related genes in decreased to levels similar to the HC group (Figure 7E), indicating that the related function of complement and immunoglobulin in NK cells downregulated in the recovery stage of the disease. We detected the complement content among the three groups, and C1q, C3 and B factor were all decreased in KPN_ACU and KPN_REC, with statistical significance compared with HC (Figure 6E; Supplementary Figure S4B). It may be that with the development of the disease, a large amount of complement is consumed, which causes the compensatory expression of the complement receptor gene to increase. During the recovery stage, the complement content did not return to normal, and the complement receptor gene C3AR1, C5AR1 in KPN_REC group was significantly downregulated compared with KPN_ACU group (Figures 6E, 7E; Supplementary Figure S4B).

In this study, the serum IFN-γ content in the KPN_ACU group was significantly higher than that in the HC group (Figure 1I). By analyzing the expression of the IFN-γ gene (IFNG), we found that IFNG was mainly expressed in NK cells during acute and recovery stages (Figure 7H), and that the expression levels of IFNG were similar between the two NK subtypes (Figure 7I). Furthermore, GO analysis found upregulation of a positive regulation of interferon-gamma production in NK cells (Figure 7F). Still, the IFN receptors IFNGR1 and IFNGR2 were not significantly upregulated in NK cells but significantly upregulated in monocytes during acute stage (Figure 7J), indicating that secreted IFN might have an effect on monocytes but not in NK cells. IFNG gene expression in NK cells during recovery stage was downregulated compared to the KPN_ACU group, but it was still higher than that in HC group (Figures 7H, I). IFN-γ content in serum was also decreased in KPN_REC group, which showed no statistical difference compared with HC group (Figure 1H), indicating that IFN-γ content in serum decreased faster than IFNG gene recovery. Some patients in the three groups had low serum levels of IFN-α (concentration less than 1pg/ml), while some patients had no serum levels of IFN-α (concentration 0pg/ml) (Supplementary Table 5). IFN-α is mainly produced by mononuclear-macrophages. Although the proportion of monocytes in patients with acute infection and recovery increased significantly, K. pneumoniae, as a Gram-negative bacterium, did not induce the production of a large amount of IFN-α in patients during the infection and recovery stage. Therefore, the immune defense function of K. pneumoniae bloodstream infection sepsis is mainly IFN-γ rather than IFN-α.

In conclusion, the cytotoxic capacity of NK cells is decreased in patients with bloodstream infection of K. pneumoniae with sepsis. NK cells are the main IFN-γ secreting cells in PBMCs, and IFN-γ, rather than IFN-α, plays an important defense role in K. pneumoniae.