Abstract

   The blood‒brain barrier (BBB) is a critical selective interface between
   the central nervous system (CNS) and the blood circulation. BBB
   dysfunction plays an important role in the neurological damage caused
   by sepsis. However, the mechanisms underlying the disruption of the BBB
   during sepsis remain unclear. We established a human induced
   pluripotent stem cell (iPSC)-derived BBB model and reported that
   treating with sepsis patient serum leads to structural and functional
   disruption of the BBB. In a cecal ligation and puncture (CLP)-induced
   mouse model of sepsis, we also observed disruption of the BBB,
   inflammation in the brain, and impairments in cognition. In both
   models, we found that the expression of TREM-1 was significantly
   increased in endothelial cells. TREM-1 knockout specifically in
   endothelial cells alleviated BBB dysfunction and cognitive impairments.
   Further study revealed that TREM-1 affects the expression of genes
   involved in the PI3K/Akt signaling pathway. The protective effects of
   TREM-1 inhibition on the BBB and cognition were abrogated by PI3K
   inhibitors. Our findings suggest that endothelial TREM-1 induces
   sepsis-induced BBB disruption and cognitive impairments via the
   PI3K/Akt signaling pathway. Targeting endothelial TREM-1 or the
   PI3K/Akt signaling pathway may be a promising strategy to maintain BBB
   integrity and improve cognitive function in sepsis patients.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12974-025-03469-5.

   Keywords: Sepsis, Blood‒brain barrier, Cognitive impairment, TREM-1,
   PI3K/Akt

Introduction

   Sepsis is a life-threatening multiorgan dysfunction syndrome caused by
   the host’s overactive immune response to pathogenic infections [[40]1].
   More than half of all deaths in intensive care units are caused by
   sepsis, and sepsis is the leading cause of death in noncardiac patients
   in intensive care units. Although progress has been made in the
   diagnosis and management of clinical sepsis, morbidity and mortality
   remain high. Preliminary extrapolations based on data from high-income
   countries suggest that, globally, there are an estimated 31.5 million
   cases of sepsis and 19.4 million cases of severe sepsis, with
   approximately 5.3 million deaths per year [[41]2]. Sometimes, sepsis
   first affects the brain [[42]3]. It usually induces diffuse cerebral
   dysfunction in the absence of direct central nervous system (CNS)
   infection. Sepsis-associated encephalopathy (SAE) clinically manifests
   as acute brain damage and is characterized by delirium, cognitive
   impairment and even coma [[43]4]. At present, sepsis symptoms are
   treated with anti-infective medications and supportive care; however,
   there are currently no treatments available that can effectively treat
   septic brain injury [[44]5]. Additionally, despite the multiple serious
   consequences of sepsis, the underlying mechanism remains elusive, and
   there is no effective clinical treatment. The pathophysiology may
   involve neurotransmitter dysfunction, brain inflammation, ischemic
   lesions, microglial activation, and blood‒brain barrier dysfunction
   [[45]6].

   Despite the incomplete elucidation of the pathological mechanisms
   underlying sepsis, several studies have demonstrated that blood‒brain
   barrier dysfunction significantly contributes to sepsis-induced
   neurological impairment [[46]7, [47]8]. The intact blood‒brain barrier
   is the critical selective interface between the CNS and blood
   circulation [[48]9]. The main component of the blood‒brain barrier is
   cerebral microvascular endothelial cells, which together with
   astrocytes, pericytes, and other supporting cells, form neurovascular
   units [[49]10–[50]12]. Changes in the structure and function of the
   blood‒brain barrier may be the key pathological mechanism of SAE.
   Maintaining blood‒brain barrier integrity is an important strategy for
   the treatment of brain dysfunction in sepsis patients [[51]13, [52]14].
   The autopsy results of adult patients who died from sepsis revealed
   that brain barrier damage was associated with severe organ dysfunction
   and systemic inflammation [[53]15]. Additionally, in animal models of
   sepsis, BBB disruption is always accompanied by neuroinflammation,
   which is manifested by the activation and proliferation of microglia in
   brain tissue and the elevation of proinflammatory cytokine levels
   [[54]16].

   Triggering receptor expressed on myeloid cell-1 (TREM-1) is an immune
   receptor that plays a key role in amplifying the inflammatory response
   [[55]17, [56]18]. Trem1 is expressed in different cell types, such as
   neutrophils, natural killer cells, macrophages, endothelial cells, and
   platelets [[57]18–[58]20]. Amplification of the inflammatory response
   by TREM-1 is also currently recognized as a key factor in the
   dysregulation of the immune response in sepsis. Studies have shown that
   TREM-1 expression on immune cells is elevated in patients with sepsis
   and is associated with increased mortality [[59]21]. Modulation of
   TREM-1 expression has proven to be a promising therapeutic strategy
   that could be used to develop new drugs and treatments for sepsis
   [[60]21]. The PI3K/AKT pathway, according to previous research, is a
   conserved family of signal transduction enzymes that are able to
   coordinate a variety of intracellular signals, control cellular
   responses to external stimuli, and regulate cell survival and
   proliferation [[61]22]. Previous studies have shown that PI3K/Akt
   signaling has anti-neuroinflammatory, antioxidative stress, and
   anti-apoptosis properties in neurons [[62]23]. Studies have shown that
   the PI3K/Akt axis is downstream of TREM-1 and is involved in
   TREM-1-mediated inflammation [[63]24]. However, the role of TREM-1 and
   its downstream PI3K/Akt pathway in the neuroinflammatory response to
   sepsis is still not fully understood.

   In this study, we constructed a human-induced pluripotent stem cell
   (hiPSC)-induced BBB model and simulated BBB disruption by treating the
   cells with plasma from septic patients. In addition, cecal ligation and
   puncture (CLP) was performed in a mouse model to mimic the development
   of sepsis. We found that sepsis led to BBB disruption and cognitive
   impairment in mice and that TREM-1 was involved in this process. A
   protective effect of TREM-1 inhibition on the BBB and brain function
   was found via the knockdown or knockout of TREM-1 in endothelial cells.
   We then investigated the role of the PI3K/Akt pathway in the cerebral
   protective effect of TREM-1 inhibition using a PI3K inhibitor. Our
   study suggests that TREM-1 can affect BBB disruption and cognitive
   impairment in sepsis and that the PI3K/Akt pathway is involved in the
   neuroprotective effect of TREM-1 inhibition. These results suggest that
   TREM-1 inhibition may be a potential therapeutic approach for improving
   BBB integrity and promoting cognitive recovery in sepsis patients.

Materials and methods

Animals

   Male mice were used in this study unless otherwise noted. Adult
   C57BL/6J mice aged 8 weeks were purchased from SPF (Beijing)
   Biotechnology Co., Ltd. Mice were housed in groups under a 12-h
   light/dark cycle. The mice were provided unrestricted access to water
   and food unless specific experimental requirements were met. All
   animals were raised and handled in accordance with the protocol
   approved by the institutional review board of the IBMS & Peking Union
   Medical College and were conducted according to the Beijing
   Administration Office of Laboratory Animals guidelines for the care and
   use of laboratory animals.

Dermal fibroblast reprogramming and cell culture

   Skin tissue was obtained from normal people and cultured with
   fibroblast medium consisting of high-glucose DMEM (Gibco), 20% fetal
   bovine serum (Gibco), 1% penicillin‒streptomycin (MACGEN), 1% GlutaMax™
   (Gibco), 1% nonessential amino acids (Gibco), and 0.1% β-ME (Gibco).
   After isolation from the tissue, fibroblasts at passage 2 were
   transduced with Sendai virus using the CytoTune™-ips2.0 Sendai
   Reprogramming Kit (Invitrogen, Cat. #A16517) according to the
   manufacturer’s instructions. The collected iPSCs were cultured in a
   6-well plate precoated with vitronectin (Gibco, Cat. #A31804). The
   cells were cultured with mTeSRTM Plus Medium (STEMCELL Technologies,
   Cat. #100–0276) at 37 °C and 5% CO2. The cells were passaged at a ratio
   of 1:6 every 3–4 days with a GCDR (STEMCELL Technologies, Cat.
   #100–0485) survival promoter.

Differentiation of iPSCs into endothelial cells

   iPSC differentiation into endothelial cells was performed using a
   STEMdiff Endothelial Differentiation Kit (STEMCELL Technologies, Cat.
   #08005) according to the manufacturer’s instructions. Specifically,
   iPSCs were seeded at a density of 7500 cells/cm2 in mTeSR™ Plus
   Medium + 10 µM Y-27,632 (STEMCELL Technologies, Cat. #72302) on a
   Matrigel^®-coated 6-well plate. Twenty-four hours after seeding, the
   cells were cultured in STEMdiff™ Mesoderm Induction Medium (STEMCELL
   Technologies, Cat. #05002) for 48 h and then in STEMdiff™ Endothelial
   Induction Medium (STEMCELL Technologies, Cat. #08006). Four days later,
   the cells were harvested using Accutane (STEMCELL Technologies, Cat.
   #07920) and seeded in a STEMdiff™ Endothelial Expansion.

Differentiation of iPSCs into neural progenitor cells (NPCs) and induction of
astrocyte differentiation

   iPSCs were seeded at a density of 2 × 10^5 cells/cm^2 in SMADi Neural
   Induction Medium + 10 µM Y-27,632 on a Matrigel^®-coated 6-well plate.
   After 24 h, the medium was changed to the STEMdiffTM SMADi Neuronal
   Induction Kit (STEMCELL Technologies), and the medium was changed
   daily. On day 8, the cells were resuspended at a density of 2 × 10^5
   cells/cm^2 in Matrigel-coated 6-well plates. The procedure was repeated
   twice. After the third passage, the medium was changed to STEMdiffTM
   Neural Progenitor medium, and the medium was changed daily.

   NPCs were collected using Accutase and seeded in neural induction
   medium on a Matrigel-coated 6-well plate. After 24 h, the medium was
   changed to astrocyte differentiation medium, and the medium was changed
   daily. The procedure was repeated twice. After the third passage, the
   medium was changed to astrocyte maturation medium. After 2–3 passages
   of cell culture, mature astrocytes were observed.

Differentiation of iPSCs into pericytes

   iPSCs were differentiated into pericytes by adapting previously
   published protocols [[64]25, [65]26]. iPSCs were seeded at a density of
   5 × 10^4 cells/cm^2 in mTeSR™ Plus Medium + 10 µM Y-27,632 on a
   Matrigel^®-coated 6-well plate. After 24 h, the medium was changed to
   STEMdiff™ Mesoderm Induction Medium. The medium was changed daily for 5
   days for mesoderm induction. The cells were seeded at a density of
   2.5 × 10^4 cells/cm^2 on a Matrigel-coated 6-well plate and maintained
   in Pericyte Medium (ScienCell, Cat. #1201).

Construction of the BBB cell model

   A triple culture model of the BBB was constructed. First, pericytes
   were cultured on the underside of the inserts of a Matrigel-precoated
   Transwell system. Twelve hours later, astrocytes were cultured in the
   same manner on the underside of the inserts. After 12 h, the Transwell
   apparatus was turned, and endothelial cells were seeded on the
   semipermeable membrane of the insert.

Transwell TEER measurement

   The TEER value of the BBB model was measured using EVOM2 (World
   Precision Instruments, Sarasota, FL, United States) every 24 h before
   the media were changed. Each Transwell insert was measured three times,
   and care was taken to measure quickly at room temperature to minimize
   temperature fluctuations. The actual resistance value was calculated as
   the resistance measurement minus the resistance value of the blank
   insert. Finally, this value was multiplied by the surface area of the
   Transwell membrane as the final resistance value, and the unit was
   Ω·cm^2.

FITC-dextran permeability detection

   To test the permeability of the model, fluorescein isothiocyanate
   (FITC)-dextran was added to the upper chamber of the Transwell system
   for one hour. Samples were collected from the upper and lower chambers
   and analyzed for fluorescence. The fluorescence intensity was measured
   at 490/520 nm. The permeability was denoted by Pe and was calculated
   using the following equation: Pe (cm/s)=([A]/t)× (I/A)×(V/[L]). where
   [A] is the concentration of FITC-dextran in the lower chamber, t(s) is
   the time interval, A(cm2) is the Transwell membrane surface area, V(m3)
   is the volume of solution in the lower chamber, and [L] is the
   concentration of FITC-dextran in the upper chamber.

Establishment of the CLP model

   The mice were deeply anesthetized, and the limbs were fixed. An
   incision of approximately 1 cm was made in the middle of the lower
   abdomen. The cecum was exposed, ligated at 50% using 3.0 silk, and
   punctured twice using a 21-gauge needle. The fecal contents were then
   extruded by squeezing the cecum. The intestinal tract was relocated to
   the abdominal cavity. The abdomen was closed. Sterile saline solution
   (37 °C) was subcutaneously injected for fluid resuscitation. The mice
   in the sham group underwent the same operation procedure but without
   ligation or puncture.

Lentiviral vector construction and transduction

   Lentiviral Vector Design: TREM-1 Knockdown (KD): Lentiviral vectors
   were constructed using the carrier element sequence
   *hU6-MCS-CBh-gcGFP-IRES-puromycin*. A scrambled shRNA vector
   (*hU6-MCS-CBh-gcGFP-IRES-puromycin*) served as the negative control.
   TREM-1 Overexpression (OE): The overexpression vector utilized the
   sequence *Ubi-MCS-3FLAG-CBh-gcGFP-IRES-puromycin*, with an empty vector
   (*Ubi-MCS-3FLAG-CBh-gcGFP-IRES-puromycin*) as the control.

   Viral Transduction: Cells were transduced with lentivirus at a
   multiplicity of infection of 50 for 72 h. Transduction efficiency was
   confirmed via GFP fluorescence.

rAAV packaging and tail vein injection

   rAAVs were packaged by Brain Case, China, and include:
   rAAV-nEF1a-spCas9-SV40 NLS-3xHA-polyA, rAAV-CMV-EGFP-bGH
   polyA-U6-SpsgRNA1(mTrem1). Mice were injected with virus by tail vein
   injection. rAAV was injected in 10^11 vector genome (v. g.).

Assessment of BBB integrity by Evans blue extravasation

   The mice were injected with 2% Evans blue (4 mL/kg body weight) via the
   tail vein 2 h prior to sacrifice. The mice were anesthetized and then
   perfused with ice-cold normal saline to remove dye from the blood
   vessels. Brain samples were collected, weighed, and then homogenized in
   formamide at 60 °C for 18 h. The concentration of infiltrated Evans
   blue was quantified at 620 nm using a multimode microplate reader
   (BioTek Synergy H1). The data are presented as nanograms per gram of
   brain weight.

Isolation of mouse brain endothelial cells

   The procedure begins with anesthesia and PBS perfusion of the mouse,
   followed by brain extraction on ice and coronal sectioning. Tissue
   sections are digested using a dissociation buffer and processed in a
   tissue dissociator for 15 min (mouse brain program). The resulting
   suspension is filtered through a 70 μm strainer, washed with DPBS, and
   centrifuged (300 g, 4 °C, 10 min). After removing debris via gradient
   centrifugation with a debris removal reagent (layered with DPBS,
   3000 g, 4 °C, 10 min), cells undergo sequential magnetic sorting:
   first, CD45-negative selection to deplete immune cells using magnetic
   beads (15 min incubation, 4 °C, protected from light), followed by
   CD31-positive selection to enrich endothelial cells under identical
   conditions. Purified cells are washed with DPBS/BSA buffer, and
   viability is assessed by mixing the suspension 1:1 with AO/PI stain for
   automated cell counter analysis. Critical steps include maintaining
   low-temperature operations, dual magnetic bead-based purification
   (CD45−/CD31+), and gradient centrifugation for debris removal to ensure
   high-purity endothelial cell isolation.

Isolation of RNA and real-time quantitative PCR (RT‒qPCR)

   Total RNA was extracted from cells or tissues using TRIzol reagent
   (Invitrogen, California, USA). The concentration and purity of the RNA
   were determined by spectrophotometry at a ratio of 260:280. Total RNA
   was reverse transcribed using a Transcriptor First Strand cDNA
   Synthesis Kit (04897030001, Roche). The expression levels of mRNA were
   assessed with real-time PCR by using FastStart Essential DNA Green
   Master Mix (06924204001; Roche). A fragment of GAPDH was amplified as
   the internal control. All reactions were run on a Roche LightCycler 96
   real-time PCR instrument. The data were analyzed using the comparative
   Ct method. All primer sequences are listed in Table [66]S1.

Protein electrophoresis and Western blotting

   The cells or tissues were ultrasonically lysed in CelLytic M
   (Sigma‒Aldrich, Cat. # C2978) supplemented with a protease inhibitor
   cocktail (Bimake, Cat. # [67]B14002). A Pierce BCA protein assay
   (Thermo Fisher, Cat. #23225) was used to determine the protein
   concentration. Protein samples were loaded onto SDS‒polyacrylamide gels
   and transferred to nitrocellulose membranes. The membranes were blocked
   with nonfat milk solution (5%) at room temperature for 1 h and
   incubated overnight at 4 °C with the primary antibody probe (listed in
   Table [68]S2). The membranes were incubated with the corresponding
   HRP-conjugated secondary antibodies (listed in Table [69]S2) at room
   temperature for 1 h. Target protein levels were analyzed via ImageJ and
   normalized to those of GAPDH or β-actin.

Enzyme-linked immunosorbent assay (ELISA)

   To measure the levels of chemokines released from the cell cultures,
   the supernatants were collected and centrifuged at 5,000 × g for
   10 min. IL-6 and IL-1β levels were measured with an ELISA kit according
   to the manufacturer’s protocols.

Neurobehavioral scores

   After CLP, the neurobehavioral scores were used to evaluate the
   neurobehavioral changes in the mice according to the corneal reflex,
   auricular reflex, tail flick reflex, righting reflex, and avoidance
   reflex. Corneal reflex: the cornea was gently touched with a cotton
   swab to observe whether the mouse blinked or shook its head. Auricular
   reflex: the mouse auricle was touched to observe whether the mouse
   turned its head. Tail flick reflex: the mouse’s tail was stimulated to
   see if the mouse would run away. Righting reflex and avoidance reflex:
   the mice were placed in the supine position, and whether the mice
   switched to the prone position with their feet on the ground was
   observed. Two points were given if the mouse responded normally. If the
   mouse was hyperreflexia (lacking a reflex for 10 s), 1 point was given.
   A score of 0 was given if the mouse had no reflex. The final
   neurobehavioral score was the total score of the above five items.

Morris water maze (MWM) test

   Spatial memory was evaluated using the MWM test [[70]27], which began
   on the sixth postoperative day. During the training phase, the mice
   located the hidden underwater platform using four visual cues on the
   pool wall for 4 consecutive days. For each trial, the mice had a
   maximum of 60 s of swimming time. The experiment ended when the mice
   swam on the platform or when the time reached 60 s. After resting on
   the platform for 10 s, the mice were returned to their cages and
   allowed to dry. The escape latency was the time taken for the mice to
   reach the platform. On the fifth day of the MWM test, the platform was
   removed, and a trial lasting 60 s was conducted. The movement
   trajectories of the mice were recorded during the experiment. The
   swimming time of the mice in the target quadrant and the number of
   times the mice crossed the original platform location were recorded.

Novel object recognition test

   The novel object recognition test is a relatively fast and efficient
   way to test learning and memory in mice. This experimental procedure
   was performed according to a previously reported protocol with some
   modifications [[71]28]. The experimental procedure consisted of three
   phases: habituation, training, and testing. The mice were placed in the
   experimental apparatus (no object) to explore freely for 10 min. The
   next day, two identical objects, free of any animal scents and
   difficult to move, were placed into the device, and the mice were
   allowed to explore freely for 10 min. On the third day, an old object
   was replaced with a new object, and the mice were allowed to explore
   the object for 5 min. The exploration of the mice in each experiment
   was recorded. The recognition index was calculated for each
   animal.Recognition index= number of explorations of novel object /
   (number of explorations of novel object + number of explorations of
   familiar object)

LY294002 administration method

   LY294002 is a broad-spectrum PI3K inhibitor. For cell experiments,
   cells were treated with LY294002 at a concentration of 50µM for 24 h.
   LY294002 was administered at a dose of 50 mg/kg via the tail vein to
   mice with treatment duration (3 days prior to sepsis induction).

Statistical analysis

   All statistical analyses were conducted using GraphPad Prism software
   (Version 8.3.0), with quantitative data expressed as mean ± SEM, where
   *n* represents the number of biologically independent replicates.
   Normality of distribution was first assessed via the Shapiro-Wilk test
   (α = 0.05). For datasets violating normality assumptions,
   non-parametric analyses were applied: Mann-Whitney U tests for
   two-group comparisons and Kruskal-Wallis tests with Dunn’s post hoc
   correction for three or more groups. Normally distributed data
   underwent parametric testing, utilizing unpaired two-tailed t-tests
   (with Welch’s correction for unequal variances) for two-group
   comparisons, while one-way ANOVA followed by Tukey’s honestly
   significant difference (HSD) post hoc test was employed for multi-group
   analyses to rigorously control type I error rates. Statistical
   significance thresholds were defined as *p < 0.05, **p < 0.01,
   ***p < 0.001, ****p < 0.0001.

Results

Establishment of a human iPSC-derived BBB model

   To investigate the molecular mechanisms by which sepsis affects the
   BBB, we established an in vitro BBB model employing human iPSCs. The
   formation of the human BBB involves interactions among three distinct
   cell types: brain endothelial cells, pericytes, and astrocytes [[72]29,
   [73]30]. We first defined a protocol for constructing an in vitro BBB
   cell model (Fig. [74]1a). We used human iPSCs to induce the
   differentiation of endothelial cells, astrocytes, and pericytes, and
   established the model using a Transwell system. Endothelial cells were
   seeded in the chamber above the permeable Transwell membrane, and
   astrocytes and pericytes were placed on the back side of the membrane.
   Immunofluorescence staining was performed to validate the successful
   differentiation of iPSCs into endothelial cells, pericytes, and
   astrocytes. Endothelial cells expressed the endothelial marker CD31
   (Fig. [75]1b), pericytes expressed α-SMA (Fig. [76]1c), and astrocytes
   expressed GFAP (Fig. [77]1d). Transepithelial/transendothelial
   electrical resistance (TEER) is a powerful indicator of cell barrier
   integrity and can be performed without causing cell damage [[78]31].
   The monolayer of endothelial cells in our Transwell system had a TEER
   value of approximately 550 Ω·cm^2, and the TEER value increased when
   the cells were cocultured with astrocytes or pericytes. When
   endothelial cells, astrocytes, and pericytes were cultured together,
   the TEER value was the highest, nearly 650 Ω·cm^2, indicating that the
   barrier integrity was greater in the coculture model than the other
   models (Fig. [79]1e). Next, we compared the molecular permeability of
   the BBB model. Compared with that in endothelial monolayers, the
   permeability of 4 kDa molecules in the BBB coculture model was reduced
   by approximately 70% (Fig. [80]1f). The permeability of Dextrans with
   larger molecular weights (10 kDa and 70 kDa) was also reduced in the
   BBB model than the endothelial monocultures, especially that of 70 kDa
   molecules, which was reduced by 80% (Fig. [81]1g-h), indicating that
   the BBB model exhibited lower permeability.

Fig. 1.

   [82]Fig. 1
   [83]Open in a new tab

   Reconstruction of the structure and physiological properties of the BBB
   in vitro. a, Schematic diagram showing the establishment of the
   iPSC-derived BBB in vitro model. b, iPSC-derived endothelial cells
   stained with CD31. c, iPSC-derived pericytes stained for α-SMA. d,
   iPSC-derived astrocytes stained with GFAP. e, TEER measurements of
   iPSC-derived endothelial cells, iPSC-derived endothelial cells plus
   astrocytes, iPSC-derived endothelial cells plus pericytes, and the BBB
   model in vitro. f-h, Permeability of endothelial cells alone,
   endothelial cells plus astrocytes, endothelial cells plus pericytes, or
   the BBB model to fluorescently-labeled dextran (4kD, 10kD, 70kD) in
   vitro. i, BBB model endothelial cells stained with ZO-1, PECAM-1,
   LRP-1, and GLUT1. j-m, RT–qPCR analysis of the expression levels of
   ZO-1, PECAM-1, LRP-1, and GLUT1. n-q, Western blot analysis of ZO-1,
   PECAM-1, LRP-1, and GLUT1 protein expression. n = 6 per group for e-h.
   n = 4 per group for j-q. White scale bar: 100 μm. The data are
   presented as the means ± SEMs. ns, not significant, *P < 0.05,
   **P < 0.01, ***P < 0.001, Student’s t test

   The low paracellular permeability of the BBB is due mainly to the tight
   junctions between endothelial cells, which are composed of the proteins
   zonula occludens-1 (zo-1) and platelet endothelial cell adhesion
   molecule-1 (PECAM-1) [[84]32]. Moreover, the expression of receptors
   such as the glucose transporter type 1 (GLUT-1) and the low-density
   lipoprotein receptor–related protein-1 (LRP-1) is an important feature
   of the BBB and helps mediate the crossing of substances across the
   barrier [[85]33]. Immunofluorescence staining confirmed the expression
   of BBB-associated proteins ZO-1, PECAM-1, LRP-1, and GLUT1 in our in
   vitro cell-based BBB model (Fig. [86]1i). Finally, we examined the
   expression levels of genes involved in maintaining the structure and
   function of the BBB by a quantitative real-time PCR (qRT‒PCR) assay in
   endothelial cells cultured alone or in coculture with astrocytes and
   pericytes. Compared with that in endothelial cells cultured alone, the
   expression of the BBB-related genes Zo-1, Pecam-1, Lrp-1, and Glut1 was
   significantly increased in endothelial cells in the BBB model
   (Fig. [87]1j‒m). Moreover, the Western blot results confirmed that the
   expression of the BBB-related proteins ZO-1, PECAM1, LRP-1, and GLUT1
   was increased in cocultured endothelial cells (Fig. [88]1n-q).

Serum from sepsis patients disrupts the structure and function of the human
iPSC-derived BBB model

   Next, we investigated the impact of serum from sepsis patients on the
   structure and function of the BBB model. First, we examined the serum
   levels of IL-1β and IL-6, which are often elevated in sepsis [[89]34,
   [90]35]. Our results revealed that the serum levels of these two
   inflammatory factors increased significantly, indicating that
   peripheral inflammation occurred in patients with sepsis (Fig. [91]2a,
   b). Next, we determined the optimal timing and concentration of serum
   applied to the BBB model by examining the barrier integrity of the
   model. When the serum concentration was 20%, the model had the lowest
   TEER values and the highest permeability at 48 h, indicating that the
   serum at high concentrations had the most destructive effect on the
   model (Fig. [92]S1). When the BBB model was treated with serum from
   sepsis patients, the TEER value was significantly lower, indicating
   that the barrier integrity of the cell layer was reduced (Fig. [93]2c).
   Moreover, the permeability of the model was examined. The permeability
   of the BBB model to 4 kDa, 10 kDa, and 70 kDa dextran increased,
   especially that of macromolecular dextran (Fig. [94]2d), which further
   suggested that the integrity of the BBB model was disrupted. We
   subsequently performed qRT‒PCR analysis of endothelial cells in the
   model. We found that the expression of the BBB-related genes Zo-1
   (Fig. [95]2e) and Pecam1 (Fig. [96]2f) in endothelial cells decreased
   after the addition of serum from sepsis patients, indicating that the
   structure of the BBB was destroyed. In addition, the expression of
   Glut1 and Lrp-1 decreased (Fig. [97]2g, h), indicating that the
   function of the blood-brain barrier may be disrupted. Western blot
   analysis revealed that the expression of the BBB-related proteins ZO-1,
   PECAM-1, and GLUT1 was decreased, whereas LRP-1 expression was not
   significantly changed (Fig. [98]2i‒l). Our study demonstrated that the
   addition of serum from patients with sepsis to the BBB cell model
   resulted in decreased integrity and increased permeability of the BBB,
   as well as reduced expression levels of several BBB-related genes and
   proteins, indicating that the structure and function of the BBB are
   disrupted during sepsis. Next, we performed transcriptional profiling
   by RNA sequencing of endothelial cells isolated from BBB models treated
   with normal human or septic serum. A total of 1719 upregulated genes
   and 1658 downregulated genes were identified (Fig. [99]2m). Further
   pathway enrichment analysis of the differentially expressed genes
   revealed that calcium signaling pathway, PI3K/Akt pathway, etc. were
   enriched in the downregulated genes in the endothelial cells treated
   with sepsis patient serum (Fig. [100]2n).

Fig. 2.

   [101]Fig. 2
   [102]Open in a new tab

   Sepsis disrupts the structure and function of the BBB in an in vitro
   BBB model. a, ELISA of the expression level of the proinflammatory
   factor IL-1β in the serum of normal individuals or patients with
   sepsis. b, ELISA of the expression level of the proinflammatory factor
   IL-6 in the serum of normal individuals or patients with sepsis. c,
   TEER values of the BBB model treated with normal or septic serum. d,
   Permeability of endothelial cells alone, endothelial cells plus
   astrocytes, endothelial cells plus pericytes, or the blood-brain
   barrier to fluorescently labeled dextran in vitro. e-h, RT‒qPCR
   analysis of the expression levels of Zo-1, Pecam-1, Lrp-1, and Glut1.
   i-l, Western blot analysis of ZO-1, PECAM-1, LRP-1, and GLUT1 protein
   expression. m, Volcano plot for differential expression analysis
   between endothelial cells after adding serum from patients with sepsis
   or normal human serum. n, Kyoto Encyclopedia of Genes and Genomes
   (KEGG) enrichment analysis of the differentially expressed genes. n = 7
   per group for a. n = 8 per group for b. n = 6 per group for c-h. n = 3
   per group for i-l. The data are presented as the means ± SEM. ns, not
   significant, *P < 0.05, **P < 0.01, ***P < 0.001, Student’s t test

Sepsis induced by CLP surgery leads to disruption of the BBB and cognitive
impairment

   Previous studies have revealed that sepsis model animals present brain
   inflammation and that the BBB is compromised in the CLP animal model
   [[103]36, [104]37]. Our study confirmed the occurrence of brain
   inflammation and disruption of the BBB in adult male C57BL/6J wild-type
   (WT) mice subjected to CLP surgery. In this study, the CLP procedure
   was used to induce sepsis in model mice (Fig. [105]3a). All of the sham
   mice appeared to be normal, with 100% survival, whereas the septic mice
   died after CLP, with a final survival rate of 60% (12 of 20 mice
   survived) (Fig. [106]3b). Compared with the relatively stable weight of
   the sham mice, the surviving mice after CLP surgery presented
   significant weight loss (Fig. [107]3c). In addition, diarrhea, pyuria,
   closed eyes or eye secretions, reduced movement and other phenomena
   were observed in the CLP-induced sepsis model mice.

Fig. 3.

   [108]Fig. 3
   [109]Open in a new tab

   Sepsis induced by CLP surgery leads to BBB barrier disruption and brain
   function impairment. a, Schematic illustration of the experimental
   design. b-c, Mortality and changes in body weight of the mice in the
   sham (n = 12 mice) and CLP (n = 20 mice) groups within 10 days after
   the operation. d, Neurobehavioral scores of the mice in the sham and
   CLP groups within 10 days after surgery. e, Representative brain images
   showing CLP surgery-induced Evans blue leakage and its quantification,
   expressed as nanograms of dye per gram of brain weight (ng/mg), n = 5
   per group. f-g, ELISA of the expression level of the proinflammatory
   factor IL-1β and IL-6 in the hippocampi, n = 8 per group. h-k, Serum
   lactate concentration (mmol/L), ALT (U/L), AST (U/L), and Serum
   creatinine (µmol/L) levels measured 24 h after sham or CLP surgery,
   n = 6 per group. l, Recognition indices of the mice in the sham and CLP
   groups, n = 12 per group. m, Latency time of mice in the sham and CLP
   groups. n, Representative images of the trajectories of the mice in the
   MWM test. o-q, Latency, duration of stay in the target quadrant, and
   number of crossing platform locations in sham and CLP groups of mice.
   n = 15 per group for m, o-q. The data are presented as the means ± SEM.
   ns, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, Student’s t
   test. r, Volcano plot of differentially expressed genes between the
   CLP-1 day group and the sham group. s, KEGG enrichment analysis of the
   differentially expressed genes between the CLP-1 day group and the sham
   group. t, Volcano plot of differentially expressed genes between the
   CLP-10 day group and the sham group. u, Venn diagram showing the number
   of changes in the transcriptional profile of mouse brain endothelial
   cells at 1 and 10 d after CLP. v, KEGG enrichment analysis of the genes
   with consistent expression in mouse endothelial cells on day 1 and day
   10 after CLP

   To assess the behaviors of the mice, we used neurobehavioral scores to
   evaluate the neurobehavioral changes in the mice, including the corneal
   reflex, righting reflex, auricle reflex, tail flick reflex, and
   avoidance reflex [[110]38]. As shown in Fig. [111]3d, during our
   observation period, the mice in the CLP group performed poorly. In the
   first three days after surgery, the neurobehavioral scores of the CLP
   group were significantly lower than those of the sham group. One day
   after CLP surgery, we observed that sepsis triggered the loss of BBB
   integrity in the mice, as demonstrated by the results of the Evans blue
   extravasation experiment (Fig. [112]3e).

   Considering that sepsis triggers a systemic inflammatory response that
   affects the brain [[113]36], we then evaluated inflammation within
   brain tissue. We found that, 24 h after the onset of sepsis, IL-1β
   (Fig. [114]3f) and IL-6 (Fig. [115]3g) levels were increased in the
   hippocampi of the mice in the CLP group compared with the sham group.
   In the CLP-induced sepsis mouse model, significant elevations in four
   critical biomarkers were observed 24 h post-surgery (Fig. [116]3h-k).
   Lactate, a marker of anaerobic metabolism and tissue hypoxia, showed
   increased serum levels, indicative of systemic hypoperfusion and
   metabolic acidosis. Both alanine aminotransferase (ALT) and aspartate
   aminotransferase (AST), enzymes released during hepatocyte injury,
   demonstrated heightened activity, reflecting sepsis-associated liver
   dysfunction. Similarly, serum creatinine, a key indicator of renal
   filtration capacity, exhibited a marked rise, suggesting impaired
   glomerular filtration and acute kidney injury. These findings
   collectively highlight the development of multiorgan dysfunction,
   encompassing metabolic derangements, hepatic damage, and renal failure,
   in response to CLP-induced systemic inflammation.

   Sepsis can lead to associated cognitive dysfunction [[117]37], so we
   then performed behavioral tests on the mice one week after CLP surgery
   to observe their cognitive function (Fig. [118]3a). The results of the
   novel object recognition experiment revealed that the mice in the CLP
   group exhibited fewer explorations of the novel object, and the
   recognition index was significantly lower than that in the sham group
   (Fig. [119]3l). In the Morris water maze test, we measured the escape
   latency (Fig. [120]3m) and observed that mice in the CLP group traveled
   a significantly longer distance before locating the platform
   (Fig. [121]3n). Assessment of motor activity in the MWM test revealed a
   significant increase in the time to reach the platform in septic mice
   compared with the sham mice (Fig. [122]3o). We also found that
   CLP-treated mice spent less time in the target quadrant (Fig. [123]3p)
   and crossed the target quadrant fewer times (Fig. [124]3q), suggesting
   that sepsis caused spatial memory dysfunction in these mice. In
   addition, we collected cerebral endothelial cells from mice at 24 h and
   10 days after CLP and performed transcriptome sequencing. Compared with
   those in the sham group, 369 genes were upregulated and 156 genes were
   downregulated in the cerebral endothelial cells of mice at 1 day after
   CLP (Fig. [125]3r), and further pathway enrichment analysis of the
   differentially expressed genes suggesting several possible related
   pathways (Fig. [126]3s). Compared with those in the sham group, 102
   genes were upregulated and 26 genes were downregulated in the cerebral
   endothelial cells of mice at 10 day after CLP (Fig. [127]3t). A total
   of 61 genes with consistent expression were identified by
   transcriptional profiling of mouse brain endothelial cells on days 1
   and 10 after CLP (Fig. [128]3u). Further KEGG pathway analysis revealed
   enrichment of several pathways for consistently expressed genes
   (Fig. [129]3v).

Endothelial TREM1 is involved in BBB disruption in both the human
iPSC-derived BBB model and the mouse CLP model

   Next, we analyzed the results of previous experiments and evaluated the
   changes in the transcriptional profiles of endothelial cells in cell
   models treated with sepsis patient serum or normal human serum and the
   changes in the transcriptional profiles of mouse brain endothelial
   cells 1 day and 10 days after CLP surgery. Eleven genes, Bst1, Prss22,
   Hcar2, Cd14, Marco, Trem1, Itih4, Mybpc2, Vwf, Plac8 and Csf2rb, were
   found to have consistent expression changes in the three groups
   (Fig. [130]4a). We subsequently used qPCR to examine gene expression in
   cell and animal samples and detected consistent expression of 5 genes,
   namely, Marco, Trem1, Cd14, Hcar2 and Mybpc2 (Fig. [131]S2-4). Previous
   studies have shown that TREM-1 can amplify inflammation and cause
   functional damage to organs [[132]39, [133]40]. Here, we hypothesize
   that TREM-1 is involved in BBB disruption in sepsis. First, we detected
   increased TREM-1 gene expression in cell models supplemented with
   septic patients and in brain endothelial cells from mice 1 or 10 days
   after CLP surgery (Fig. [134]4b-d). The same phenomenon was also
   detected via Western blotting, and the protein expression level of
   TREM-1 was significantly increased (Fig. [135]4e-g). Next, we
   constructed TREM-1-knockdown or TREM-1-overexpressing lentiviruses
   (Fig. [136]4h), added them to the cell culture medium, and harvested
   endothelial cells after three days of treatment (Fig. [137]4i). Next,
   endothelial cells with TREM-1 knockdown or overexpression were
   generated via the addition of virus (Fig. [138]4j-k), and the TEER
   values and permeability were examined. In analyses of tightness and
   permeability, compared with normal endothelial cells, TREM-1-knockdown
   endothelial cells were found to be less responsive to septic serum, as
   indicated by a higher TEER value in the BBB model (Fig. [139]4l) and a
   lower increase in permeability to FITC‒dextran (Fig. [140]4m). These
   findings suggest that TREM-1 knockdown, to a certain extent, hinders
   the destructive effect of serum from septic patients on BBB function.

Fig. 4.

   [141]Fig. 4
   [142]Open in a new tab

   Endothelial TREM1 is involved in BBB disruption. a, Venn diagram of
   changes in the transcriptional profiles of endothelial cells in cell
   models treated with serum from patients with sepsis and normal human
   serum in cell experiments and changes in the transcriptional profiles
   of mouse brain endothelial cells at 1 and 10 days after CLP. b, RT‒qPCR
   analysis of TREM-1 expression in endothelial cells isolated from BBB
   models supplemented with normal human or septic serum. c-d, RT‒qPCR
   analysis of the expression level of TREM-1 on day 1 and day10 after the
   operation in the sham operation group and the CLP group. e, Western
   blot analysis of TREM-1 expression in endothelial cells isolated from
   BBB models treated with normal human or septic serum. f-g, Western blot
   analysis of the expression level of TREM-1 on day 1 and day 10 after
   the operation in the sham operation group and the CLP group. h,
   Schematic representation of the lentiviral vector. i, Schematic of
   protein extraction from endothelial cells after virus injection. j-k,
   RT‒qPCR analysis and western blot analysis of TREM-1 expression in
   endothelial cells with TREM-1 knockdown or overexpression. l, TEER of
   the BBB models constructed with normal, TREM-1-knockdown, or
   TREM-1-overexpressing endothelial cells after the addition of septic
   patient serum. m, Magnitude of increase in permeability of
   fluorescently labeled dextran (10 kDa) in the BBB models with normal,
   TREM-1-knockdown, or TREM-1-overexpressing endothelial cells after the
   addition of septic patient serum.n = 4 per group for b-e, j-k. n = 3
   per group for f-g. n = 6 per group for l-m. The data are presented as
   the means ± SEM. ns, not significant, *P < 0.05, **P < 0.01,
   ***P < 0.001, b-g, Student’s t test. j-m, One-way ANOVA

TREM-1 knockout in endothelial cells alleviates cognitive impairment in a CLP
mouse model

   On the basis of the increased expression of TREM-1 in the brain
   endothelial cells of septic mice, we used AAV-BI30 as a gene transfer
   vector for endothelial expression of a single guide (sgRNA) against
   TREM-1 (Fig. [143]5a) to investigate the function of TREM-1 in sepsis.
   AAV-BI30 can achieve efficient and endothelial cell-specific
   Cre-mediated genetic modulation in the CNS [[144]41]. Moreover,
   recombinant adeno-associated virus (rAAV), which we constructed, can
   target brain endothelial cells. One month after rAAV injection via the
   tail vein, TREM-1 was not expressed in the brain endothelial cells of
   the mice injected with TREM-1 sgRNA compared with those of the control
   mice (Fig. [145]5b). After rAAV injection, the mice were subjected to
   CLP surgery, and the antiseptic effect of TREM-1 knockout on the brain
   endothelial cells of CLP mice was evaluated by body weight assessment,
   neurobehavioral scoring, novel object recognition tests and MWM tests
   (Fig. [146]5c). The number of mouse deaths in each group was counted,
   and the mice in the sham operation group behaved normally, with a 100%
   survival rate. In the CLP group, the survival rate of the mice injected
   with the TREM-1 sgRNA was 63.6%, which was greater than that of the
   control mice (55%) (Fig. [147]5d). In the body weight assessment,
   compared with the sham operation, the CLP surgery significantly reduced
   the body weights of the mice.While the difference was not statistically
   significant, TREM-1 knockout mice tended to recover body weight more
   rapidly than control mice (Fig. [148]5e). The neurobehavioral scores of
   the sham group returned to baseline immediately after a slight decrease
   on the first day after surgery. In the CLP group, the amplitude of the
   scores decreased on the first day after the operation and slowly
   returned to baseline at 1 week after the operation. Moreover, the
   neurobehavioral scores of the TREM-1 knockout mice returned to normal
   more rapidly than did those of the control mice at one week, although
   they also underwent CLP surgery (Fig. [149]5f). One day after CLP, we
   performed an Evans blue test to evaluate the permeability of the BBB in
   mice and found that TREM-1 knockout alleviated the BBB damage caused by
   CLP (Fig. [150]5g). Next, we examined whether TREM-1 knockout could
   alleviate cognitive impairment in sepsis. Behavioral tests were
   performed starting on day 6 after CLP. The recognition index, which is
   the frequency of exploration of the novel object by the mouse, was
   recorded in the novel object recognition test. Compared with control
   mice, TREM-1 knockout mice presented a higher recognition index after
   CLP (Fig. [151]5h). In the MWM experiment, the movement trajectory of
   the mice was recorded, along with the latency to reached the platform,
   the movement time in the target quadrant, and the number of times the
   mice crossed the target quadrant.The water maze test revealed that
   TREM1 knockout mice, although not different in terms of the latency to
   reach the platform on the last day, reached the platform in a shorter
   time during the first few days of training than did control mice, which
   also underwent CLP surgery (Fig. [152]5i, j). In addition, we found
   that TREM-1 knockout mice had longer movement times and more crossing
   times in the target quadrant than control mice did, indicating that
   TREM-1 knockout alleviated cognitive and learning impairments in these
   mice (Fig. [153]5k, l). These results suggest that TREM-1 knockout in
   endothelial cells can alleviate BBB disruption and brain function
   impairment in sepsis. The inhibition of TREM-1 remains a promising
   target for the development of therapies for sepsis.

Fig. 5.

   [154]Fig. 5
   [155]Open in a new tab

   TREM-1 knockout in endothelial cells alleviates cognitive impairment in
   a CLP mouse model. a, Schematic representation of rAAV vectors
   expressing control (CTL) or TREM-1 sgRNA. b, Western blot analysis of
   TREM-1 expression in mouse brain endothelial cells after injection of
   the TREM-1 sgRNA. c, Schematic illustration of the experimental design.
   d, Mortality of mice in the CON + sham (n = 20 mice), CON + CLP (n = 22
   mice), TREM-1 sgRNA + sham (n = 20 mice), and TREM-1 sgRNA + CLP
   (n = 22 mice) groups within 10 days after the operation. e, Changes in
   the body weights of mice in the CON + sham (n = 20 mice), CON + CLP
   (n = 12 mice), TREM-1 sgRNA + sham (n = 20 mice), and TREM-1
   sgRNA + CLP (n = 14 mice) groups within 10 days after the operation. f,
   Neurological scores of mice in the CON + sham (n = 20 mice), CON + CLP
   (n = 12 mice), TREM-1 sg RNA + sham (n = 20 mice), and TREM-1
   sgRNA + CLP (n = 14 mice) groups within 10 days after surgery. g, EB
   leakage in the CON + sham (n = 20 mice), CON + CLP (n = 12 mice),
   TREM-1 sg RNA + sham (n = 20 mice), and TREM-1 sgRNA + CLP (n = 14
   mice) groups 24 h after surgery. h, Recognition index of mice in the
   CON + sham, CON + CLP, TREM-1 sg RNA + sham, and TREM-1 sgRNA + CLP,
   n = 12 mice per group. i-j, Latency time of mice in the CON + sham,
   CON + CLP, TREM-1 sg RNA + sham, and TREM-1 sgRNA + CLP, n = 15 mice
   per group. k, Time spent in the target quadrant by mice in the
   CON + sham, CON + CLP, TREM-1 sg RNA + sham, and TREM-1 sgRNA + CLP,
   n = 15 mice per group. l, Number of crossing over the platform in the
   CON + sham, CON + CLP, TREM-1 sg RNA + sham, and TREM-1 sgRNA + CLP,
   n = 15 mice per group. The data are presented as the means ± SEMs. ns,
   not significant, *P < 0.05, **P < 0.01, ***P < 0.001, Student’s t test

Inhibition of TREM-1 prevents degradation of the BBB through the PI3K/Akt
signaling pathway, thereby alleviating brain injury in sepsis

   Previous studies have shown that PI3K/Akt signaling plays an
   antineuroinflammatory role in neurological diseases [[156]42]. Our
   RNA-sequencing results also revealed that PI3K/Akt signaling may be
   involved in the pathogenesis of sepsis-induced disruption of the BBB
   (Figs. [157]2n and [158]3s). We further investigated whether the
   inhibition of TREM-1 alleviates disruption of the BBB through the
   PI3K/Akt signaling pathway, thereby ameliorating sepsis-induced brain
   injury. After the behavioral tests, Western blot analysis of PI3K/Akt
   pathway-related proteins in brain endothelial cells was performed to
   determine whether TREM-1 knockout in mouse brain endothelial cells
   triggers changes in the PI3K/Akt signaling pathway. Compared with the
   sham operation, the CLP operation significantly reduced the levels of
   PI3K and p-Akt, while the protein expression of PI3K and p-Akt
   increased after TREM-1 knockout in cerebral endothelial cells
   (Fig. [159]6a, b). The results of qRT‒PCR revealed that the mRNA
   expression level of PI3K in the CLP group was significantly lower than
   that in the control group and that the mRNA expression level of PI3K in
   the TREM-1 knockout group was increased (Fig. [160]6c). Furthermore, in
   the human iPSC-derived BBB model, we used LY294002, a PI3K antagonist,
   to determine whether the PI3K/Akt signaling pathway is involved in the
   ability of TREM-1 inhibition to slow BBB disruption and brain injury in
   sepsis. After adding LY294002, we observed decreased expression of PI3K
   and p-Akt in endothelial cells within the cellular model
   (Fig. [161]7a), confirming the inhibitory effect of LY294002 on the
   PI3K pathway. TREM-1 knockdown has a protective effect on the integrity
   of the BBB. However, after the addition of LY294002, which inhibits
   PI3K, the TEER value of the model was reduced to approximately 430
   Ω·cm^2, and the barrier integrity was reduced (Fig. [162]7b). The
   permeability experiments also revealed that TREM-1 knockout reduced the
   permeability of the BBB model, and the addition of LY294002 to inhibit
   PI3K increased the permeability of the model (Fig. [163]7c). These
   results indicated that the original protective effect of TREM-1
   inhibition on the BBB was abolished by the addition of the PI3K
   inhibitor LY294002. These findings suggest that the PI3K/Akt pathway is
   involved in the protective effect of TREM-1 inhibition on the BBB.
   Furthermore, we conducted relevant animal experiments. After
   administering LY294002 to mice, we collected mouse brain endothelial
   cells and performed western blotting to assess the expression of key
   molecules in the PI3K pathway. The results demonstrated that LY294002
   indeed reduced the expression of PI3K and phosphorylated Akt (p-Akt) in
   brain endothelial cells (Fig. [164]7d, e). In TREM-1 knockout mice,
   disruption of the BBB was alleviated after CLP, whereas LY294002
   aggravated disruption of the BBB. These findings indicated that
   inhibition of the PI3K pathway could significantly eliminate the
   protective effect of TREM-1 knockout on the BBB in mice (Fig. [165]7f).
   Behavioral experiments in mice also demonstrated that PI3K inhibition
   partially counteracted the protective effect of TREM-1 knockout on
   brain function. In the novel object recognition test, compared with
   control mice, TREM-1 knockout mice, which also underwent CLP surgery,
   presented an increased recognition index, demonstrating the protective
   effect of TREM-1 knockout on cognitive and memory abilities. However,
   after the injection of LY294002, the recognition index of the mice
   decreased (Fig. [166]7g). In the water maze test, we found that TREM-1
   knockout mice took a shorter amount of time to find the platform, moved
   in the target quadrant for a longer period of time, and crossed the
   target quadrant more times, indicating that the mice had improved
   cognitive function. However, inhibition of PI3K by LY294002 resulted in
   poor performance in the water maze (Fig. [167]7h-j). These findings
   suggest that the PI3K/Akt pathway is involved in the protective effect
   of TREM-1 inhibition on brain function and that the inhibition of PI3K
   can partially eliminate the protective effect of TREM-1 knockout on
   cognition and memory ability.

Fig. 6.

   [168]Fig. 6
   [169]Open in a new tab

   Changes in the protein expression in the PI3K/AKT signaling pathway in
   mouse brain endothelial cells. a, The protein levels of PI3K, AKT, and
   p-AKT were measured by Western blotting. β-actin served as a reference
   protein. b, The expression ratios of the PI3K/β-actin and p-Akt/Akt
   proteins were quantified via ImageJ software. c, The expression levels
   of PI3K and AKT were measured via qRT‒PCR. n = 3 per group for b. n = 6
   per group for c. The data are presented as the means ± SEMs. ns, not
   significant, *P < 0.05, **P < 0.01, ***P < 0.001, One-way ANOVA

Fig. 7.

   [170]Fig. 7
   [171]Open in a new tab

   Inhibition of PI3K partially abolished the protective effect of TREM-1
   knockout on the BBB and cognitive function. a, The protein levels of
   PI3K, AKT, and p-AKT were measured by Western blotting. β-actin served
   as a reference protein. b, TEER values of the BBB models in the
   CON + septic serum, TREM-1-KD + septic serum, and TREM-1-KD + septic
   serum + LY294002 groups. c, Permeability of fluorescently labeled
   dextran (10 kDa) in the CON + septic serum, TREM-1-KD + septic serum,
   and TREM-1-KD + septic serum + LY294002 groups. d, Representative
   western blot analysis of PI3K, Akt, and p-Akt expression in brain
   endothelial cells isolated from control mice and LY294002-treated mice.
   e, Western blot analysis of PI3K/Akt pathway components in brain
   endothelial cells from CON + CLP, TREM-1-KO + CLP and
   TREM-1-KO + CLP + LY294002 groups. f, EB leakage in the CON + septic
   serum, TREM-1-KO + septic serum, and TREM-1-KO + septic
   serum + LY294002 groups. g, Recognition indices of mice in the
   CON + septic serum, TREM-1-KO + septic serum, and TREM-1-KO + septic
   serum + LY294002 groups. h, Latency time of mice in the CON + septic
   serum, TREM-1-KD + septic serum, and TREM-1-KD + septic
   serum + LY294002 groups. i, Time spent in the target quadrant by mice
   in the CON + septic serum, TREM-1-KD + septic serum, and
   TREM-1-KD + septic serum + LY294002 groups. j, Times at which the mice
   crossed the platform in the CON + septic serum, TREM-1-KD + septic
   serum, and TREM-1-KD + septic serum + LY294002 groups. n = 4 per group
   for a, d-f. n = 6 per group for b-c. n = 12 per group for g, n = 15 per
   group for h-j. The data are presented as the means ± SEMs. ns, not
   significant, *P < 0.05, **P < 0.01, ***P < 0.001

Discussion

   Brain dysfunction occurs during the onset of sepsis. However, many
   patients with sepsis still experience long-term cognitive impairment
   after discharge, which affects memory, attention, verbal fluency, and
   executive ability [[172]43]. Autopsy studies have shown that the damage
   to the nervous system caused by sepsis is profound, and ischemia can be
   observed in several brain regions, including the hippocampus, amygdala,
   and hypothalamus [[173]13]. Owing to the complex pathophysiological
   features of sepsis-induced brain injury, there are currently no
   specific medical treatments for this disease [[174]44]. BBB dysfunction
   plays an important role in the pathophysiology of sepsis and SAE. The
   disruption of the BBB enables toxic substances to enter the brain,
   which leads to changes in the structure and function of the brain and
   induces brain inflammation [[175]9]. After the BBB is disrupted, the
   CNS becomes very vulnerable to neurotoxic factors such as free
   radicals, inflammatory mediators, and plasma and intravascular proteins
   [[176]13, [177]45]. In the present study, we found that endothelial
   cells expressing TREM-1 mediated sepsis-induced BBB disruption and
   cognitive impairment via the PI3K/Akt pathway. The addition of serum
   from sepsis patients to the BBB cell model reduced the integrity of the
   cell monolayer. The CLP model septic mice also showed disruption of the
   hemorrhagic brain barrier and impaired brain function.

   There are many obstacles to developing effective treatments for sepsis,
   such as poor translation between human disease and animal models,
   heterogeneity of patient samples selected in clinical trials, and
   variation in disease etiology. The establishment of normal
   human-sourced iPSC-derived endothelial cells, pericytes, and astrocytes
   as in vitro cell models of the BBB paves the way for the study of BBB
   dysfunction. In vitro human iPSC-derived BBB models have been
   established using different differentiation protocols [[178]25,
   [179]46–[180]48] and can be used to study BBB disorders in a variety of
   diseases, such as Alzheimer’s disease and multiple sclerosis [[181]49,
   [182]50]. This encouraged us to generate human iPSCs using skin samples
   collected from healthy people to construct BBB models and to simulate
   sepsis by adding serum from septic patients to the model cells to
   explore sepsis-induced BBB dysfunction, which may reveal a pathological
   process that is not fully understood.

   To investigate the underlying mechanisms of sepsis and associated
   validation responses, several experimental animal models have been
   developed, all of which attempt to mimic the pathophysiological changes
   typical of septic patients [[183]51]. CLP is widely used to establish
   an experimental sepsis model, and the CLP sepsis model is currently
   considered the gold standard for sepsis research [[184]52–[185]54]. The
   establishment of a sepsis model via CLP leads to an increase in the
   brain water content and disruption of the BBB. This model can be used
   to simulate the pathology of clinical septic patients [[186]55]. In our
   study, a CLP sepsis model was established, and the levels of
   inflammatory factors in the serum of model mice were measured. The
   levels of IL-1β and IL-6 were increased, which was consistent with
   previous studies [[187]56]. In CLP, the cecal ligation distance is a
   major determinant of surgical severity and mortality. In addition, the
   size of the puncture needle, the number of punctures, and fluid
   resuscitation all have an impact on the surgical effect [[188]52].
   Therefore, performing the CLP procedure with high consistency and
   reproducibility is particularly important. Our experiment strictly
   controlled the above operations to ensure the consistency of the CLP
   surgery. In addition, we also utilized an in vitro human iPSC-derived
   BBB model, which to some extent complemented the animal model.

   The TREM family includes at least 5 distinct receptors that share
   sequence homology. Among them, Trem-1 plays an important role in the
   process of inflammation-related diseases [[189]57]. LPS can activate
   TLR4 and further amplify the inflammatory response. LR12, a clinically
   available TREM-1 inhibitory peptide, can prevent the binding of TREM-1
   ligands, specifically inhibiting TREM-1 and blocking the amplification
   of inflammation [[190]58]. However, the use of TREM-1 inhibitory
   peptides such as LR12 has limitations, and the “endothelial protection”
   observed in vivo may simply be the result of systemic inflammation and
   reduced activation of inflammatory cells [[191]58]. To eliminate this
   concern, we used rAAV vectors to target brain endothelial cells and
   deliver sgRNA for TREM-1 knockdown. This method ensured the normal
   expression of TREM-1 in other non-endothelial cells. Using animal
   experiments, we found that depletion of TREM-1 in brain endothelial
   cells protected against the disruption of the BBB and cognitive
   impairment caused by CLP in mice, suggesting that endothelial TREM-1
   mediates sepsis-associated brain damage.

   The PI3K-AKT signaling pathway regulates signal transduction; cell
   proliferation, apoptosis, and metabolism; and other biological
   processes. Accumulating evidence suggests that the PI3K/Akt pathway
   plays a neuroprotective role in a variety of diseases, which provides
   an effective resource for the discovery of potential therapeutic drugs
   [[192]59]. The PI3K/Akt signaling pathway is thought to play an
   important role in sepsis [[193]60]. Studies have shown that activation
   of the PI3K/Akt pathway negatively regulates the nuclear factor-κB
   (NF-κB) activation pathway, suppresses proinflammatory responses, and
   improves cardiac function and prolongs survival time in mice with
   sepsis [[194]22, [195]61, [196]62]. Our results indicate that sepsis
   leads to decreased levels of PI3K and phosphorylated AKT. However,
   TREM-1 knockdown or knockout significantly increased the levels of
   phosphorylated AKT, indicating that TREM1 is involved in the disruption
   of the BBB during sepsis and may exert its effects through PI3K/AKT
   pathway. In this study, LY294002, a PI3K antagonist, was used to
   examine the effect of PI3K inhibition on cognitive behavior in septic
   mice with cerebral endothelial cell-specific TREM-1 knockout. PI3K
   inhibition abolished the protective effect of TREM-1 knockout on the
   BBB and brain function. The regulatory interplay between TREM-1 and the
   PI3K/Akt pathway involves multi-step cascades. Existing studies confirm
   that PI3K/Akt pathway activation can suppress pro-inflammatory
   responses by negatively regulating the nuclear factor-κB (NF-κB)
   activation pathway, improving cardiac function and extending survival
   in septic mice [[197]22, [198]61, [199]62]. Preclinical evidence
   indicates that PI3Kα-specific inhibitors (e.g., Alpelisib) targeting
   the p110α mutant significantly enhance therapeutic specificity in
   breast cancer [[200]63], suggesting that endothelium-restricted PI3Kα
   inhibitors may represent optimized therapeutic strategies for sepsis
   management [[201]64]. Although the PI3K/Akt pathway is one of the
   downstream pathways of TREM-1, the exact mechanism by which this
   signaling pathway contributes to the pathophysiological process of
   TREM-1 needs to be further investigated.

   There are several limitations in our study. First, the cells used to
   construct the BBB model were derived from healthy individuals, and the
   serum of patients or healthy individuals was added to simulate the
   sepsis group or control group. Considering that there may be a genetic
   component in the development of sepsis, iPSC-derived BBB models
   established using cells derived from sepsis patients should also be
   established for comparison with normal controls and for further
   studies. Second, only male mice were used in the animal experiments in
   this study, whereas epidemiological studies have shown that young women
   have better prognoses and clinical outcomes than men do in with the
   context of sepsis [[202]65]. Additional studies on the role of TREM-1
   in sepsis-induced BBB disruption in female mice may provide further
   evidence of its function in sepsis. Finally, the molecular and cellular
   mechanisms through which TREM-1 induces BBB disruption and cognitive
   impairment during sepsis need to be further investigated, especially
   its role in regulating the PI3K/Akt pathway.

   In summary, our findings suggest that endothelial TREM-1 mediates BBB
   disruption and cognitive impairment in sepsis. TREM-1 inhibition can
   alleviate BBB disruption and cognitive dysfunction through the PI3K/Akt
   pathway. TREM-1 inhibition can be used as a treatment option for
   sepsis. However, the pathogenesis of sepsis is complex, and future
   studies should further explore the signaling mechanisms of TREM-1 in
   endothelial cells and the interactions between TREM-1 and the PI3K/Akt
   pathways and other signaling pathways.

Conclusions

   Our study revealed the mechanism of BBB disruption during sepsis. In
   both the human iPSC-derived BBB model and the mouse model of sepsis
   induced by CLP surgery, endothelial TREM-1 was involved in the
   disruption of the BBB. CLP treatment disrupted the BBB and impaired
   cognitive ability in mice. We demonstrated that the inhibition of
   TREM-1 improves the function of the BBB and alleviates cognitive
   impairment through the PI3K/Akt pathway. Our findings suggest that
   TREM-1 and PI3K/Akt pathway members may be promising therapeutic
   targets for the treatment of sepsis.

Electronic supplementary material

   Below is the link to the electronic supplementary material.
   [203]Supplementary Material 1^ (781.7KB, docx)
   [204]Supplementary Material 2^ (6.5MB, docx)

Abbreviations

   BBB
          Blood-brain barrier

   CLP
          Cecal ligation and puncture

   CNS
          Central nervous system

   FITC
          Fluorescein isothiocyanate

   GFAP
          Glial fibrillary acidic protein

   GLUT1
          glucose transporter type 1

   IL-1β
          Interleukin-1 beta

   IL-6
          Interleukin-6

   KD
          Knock down

   KO
          Knock out

   LPS
          Lipopolysaccharide

   LRP-1
          Low-density lipoprotein receptor-related protein-1

   MWM
          Morris water maze

   NPC
          Neural progenitor cell

   PECAM-1/CD31
          platelet endothelial cell adhesion molecule-1

   PI3K
          Phosphatidylin-ositol-3-kinase

   SAE
          Sepsis-associated encephalopathy

   TEER
          Transepithelial/transendothelial electrical resistance

   TREM-1
          Triggering receptor expressed on myeloid cell-1

   ZO-1
          zonula occludens-1

   α-SMA
          Alpha smooth muscle actin

Author contributions

   Y.S., W.Z., T.S., L.H., M.Q., and Q.W. performed the research; Y.S. and
   L.H. analyzed the experimental data; Y.S. wrote the main manuscript
   text. Q.X., Y.L., and J.X. supervised the laboratory experiments and
   substantively revised the manuscript; and all other authors reviewed
   and commented on the manuscript.

Funding

   This work was supported by research grants from STI2030-Major Projects
   (2022ZD0211701, 2021ZD0203001), the CAMS Innovation Fund for Medical
   Sciences (2021-I2M-1-020), the National High Level Hospital Clinical
   Research Funding (2022-PUMCH-B-109), Special Research Fund for Central
   Universities, Peking Union Medical College (3332024120), and the
   Fundamental Research Funds for the Central Universities, Peking Union
   Medical College.

Data availability

   Raw data are available upon reasonable request from the corresponding
   authors.

Declarations

Ethics approval and consent to participate

   All animals were raised and handled in accordance with the protocol
   approved by the institutional review board of the IBMS & Peking Union
   Medical College and were conducted according to the Beijing
   Administration Office of Laboratory Animals guidelines for the care and
   use of laboratory animals. The work involving human serum samples in
   this article was approved by the Ethics Institutional Review Board of
   Peking Union Medical College Hospital (PUMCH) (serial number: ZS-2165).

Consent for publication

   All authors agreed to the content of this manuscript prior to
   submission.

Competing interests

   The authors declare no competing interests.

Footnotes

   Publisher’s note

   Springer Nature remains neutral with regard to jurisdictional claims in
   published maps and institutional affiliations.

   Yuwen Su and Wanwan Zhu have contributed equally to this work.

Contributor Information

   Qi Xu, Email: xuqi@pumc.edu.cn.

   Yi Li, Email: billliyi@126.com.

   Jianbo Xiu, Email: xiujianbo@ibms.cams.cn.

References