Intravenous Lipopolysaccharide Challenge in Healthy Participants Reveals Pharmacodynamic Markers in Blood and Bone Marrow for Early‐Phase Oncology Drug Development

Abstract

Demonstrating pharmacological effects in early‐phase oncology clinical trials remains challenging, largely due to the lack of robust pharmacodynamic markers. Lipopolysaccharide (LPS) is used as an immune challenge agent in healthy participants to study drugs for autoimmune conditions. We hypothesized that the intravenous LPS challenge model could be implemented in oncology drug studies in healthy participants to evaluate in vivo pharmacodynamics of new drugs. This open‐label study included 13 healthy males who received a single intravenous dose of 1 or 2 ng/kg LPS. Blood and bone marrow samples were collected pre‐ and post‐LPS for bulk RNA sequencing, multiplex cytokine measurements, immunophenotyping, and immunohistochemistry of bone marrow biopsies. Differentially expressed genes were integrated with the OpenTargets Platform and the COSMIC database to assess their tractability and relation to cancer hallmarks. The LPS challenge and bone marrow sampling were safe and well tolerated. In peripheral blood, LPS induced a rapid, transient inflammatory response, with myeloid cell activation and enrichment of inflammation‐related pathways. Conversely, bone marrow showed an attenuated cellular response and activation of hematopoietic and cell cycle‐related transcriptional programs. We identified 91 LPS‐induced oncology‐relevant targets in blood and 24 targets in bone marrow. The oncogenic transcription factor MYC was upregulated in the bone marrow, which was confirmed at the protein level by a dose‐dependent increase in c‐Myc expression. Our findings highlight the potential of immune challenges in healthy participants to pharmacodynamically profile novel oncology agents, with the aim of generating early pharmacological data ahead of pivotal clinical trials in cancer patients.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Dose selection in early‐phase oncology studies is still often guided by safety and tolerability outcomes, despite the shift in the oncology landscape toward more targeted agents and immunotherapies. Moreover, early clinical studies with oncology drugs are typically performed in patients, where comorbidities and pretreatment hinder reliable pharmacodynamic (PD) evaluation. Regulatory initiatives such as Project Optimus emphasize the need for PD‐guided dosing, yet suitable methods to efficiently profile novel compounds early in development remain limited. Intravenous lipopolysaccharide (LPS) challenge is a well‐established tool to characterize anti‐inflammatory drugs in healthy participants, but its translational utility for oncology drug development has not been defined.

• WHAT QUESTION DID THIS STUDY ADDRESS?

This study addresses whether an intravenous LPS challenge in healthy participants can safely induce oncology‐relevant transcriptional and protein responses in peripheral blood and bone marrow, and whether the identified tractable targets and their corresponding PD markers could be applicable in early‐phase oncology drug development.

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

This study demonstrates that intravenous LPS challenge can induce oncology‐relevant tractable targets in a controlled human setting. Blood and bone marrow showed distinct, compartment‐specific cellular and transcriptional responses. Oncogenic drivers such as MYC were inducible and quantifiable in the bone marrow compartment, with dose‐dependent mRNA and protein upregulation. Lastly, bone marrow sampling was also well tolerated when combined with the LPS challenge.

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

The LPS challenge model can be used for proof‐of‐mechanism and in vivo target engagement evaluation of novel oncological compounds early in drug development, thereby supporting more informed decision‐making and aiding dose selection for subsequent patient studies.

Current oncology drug development focuses mainly on novel molecular targeted agents (MTAs) and immuno‐oncology drugs, shifting the paradigm from traditional cytotoxic chemotherapies to therapies that either directly target pathways within tumor cells or indirectly kill malignant cells through immune mechanisms. This transition to novel types of compounds raises the importance of early‐phase dose‐finding studies in oncology. Contrary to the historical toxicity‐guided approach, dose‐finding should be conducted with the drug's pharmacodynamic (PD) activity in mind, as also outlined by the Project Optimus initiative. ¹ Although PD markers, defined as biomarkers of pharmacologic response, ² are often used in oncology, challenges remain due to tumor heterogeneity, extensive patient pretreatment, and insufficiently informative PD analyses. ³ Moreover, insufficient use of PD markers and the historical assumption that higher doses always yield greater therapeutic activity have led to the necessity for post‐approval dose‐finding studies for some MTAs, which were required to optimize drug tolerability by reducing the doses. ⁴ Naturally, this negatively impacts patients and incurs additional costs for drug development. Moreover, because the plateau of clinical response is often achieved below the maximum tolerated dose (MTD), and because many new drugs are intended for chronic use in already vulnerable patient populations, accurately characterizing PD responses in early‐phase clinical development is crucial. ⁵ Altogether, this translates into the need for more informative PD markers that can be used to demonstrate target engagement and aid in early dose selection.

In clinical immunology and rheumatology, dose‐finding first‐in‐human studies frequently utilize PD markers to establish the pharmacologically active dose range. Various immune challenge models are used in healthy participants to mimic disease‐relevant immune processes and to induce measurable responses that serve as PD markers for novel drugs. One such model uses lipopolysaccharide (LPS), a component of the outer membrane of Gram‐negative bacteria, which binds to Toll‐like receptor 4 (TLR4) and induces an acute and transient inflammatory response. LPS has previously been administered both locally (intradermally) ⁶ , ⁷ and systemically ⁸ , ⁹ and is a valuable tool to explore the PD effects of novel anti‐inflammatory drugs. We hypothesize that such immune challenge models, specifically the LPS challenge model, could be implemented in early‐phase oncology studies to provide in vivo PD readouts and demonstrate proof‐of‐concept for novel drugs. Inflammation is a well‐established hallmark of cancer, ¹⁰ and many oncogenic pathways, such as PI3K/AKT, MAPK, and NF‐κB signaling, are activated during LPS‐induced inflammation. ¹¹ As they often encompass therapeutic targets of novel oncology drugs, LPS‐induced upregulation of these targets in healthy participants could provide measurable PD markers for evaluating drug effects in a controlled in vivo setting. Additionally, the human bone marrow niche is crucial in hematologic malignancies, and due to its unique precursor and differentiating cell populations, could yield novel PD markers for oncology‐relevant targets that are not readily detectable in peripheral blood. Previously, animal studies were performed to evaluate gene expression profile related to LPS ¹² , ¹³ and in humans, post‐LPS transcriptomic profiles of neutrophils and eosinophils after LPS administration were characterized in blood and bone marrow. ¹⁴ , ¹⁵ , ¹⁶ However, a comprehensive transcriptomic analysis of the effects of LPS in human bone marrow is lacking and potential oncology‐relevant PD markers in this context remain underexplored.

The primary aim of this study was to determine whether intravenous LPS challenge in healthy participants can induce expression of oncology‐relevant druggable targets, thereby providing PD markers that could be applied in future early‐phase studies with novel oncology drugs. Additionally, we sought to comprehensively characterize the peak inflammatory response to LPS in peripheral blood and bone marrow at multiple levels, including cellular composition and activation, secreted inflammatory cytokines, and genome‐wide transcriptional changes, to provide the context for identifying oncology‐relevant PD markers. Selected targets were subsequently evaluated at the protein level in bone marrow biopsies. Our secondary aim was to establish the feasibility of utilizing the human bone marrow environment for future early‐phase clinical pharmacology studies.

MATERIALS AND METHODS

The study was executed in accordance with the Declaration of Helsinki, at the Center for Human Drug Research (CHDR), Leiden, The Netherlands, between September 20, 2022 (first screening) and December 30, 2022 (last participant visit). The study protocol was approved by the Medical Ethics Committee “Medisch Ethische Toetsingscommissie van de Stichting Beoordeling Ethiek Biomedisch Onderzoek” (Assen, The Netherlands), and was registered in the ISRCTN Registry under number ISRCTN16913007. All participants provided written informed consent prior to any study activity. The study is reported in accordance with the CONSORT‐DEFINE guidelines. ¹⁷

Study participants

Only male participants were included in this study as our previous experience with systemic administration of 2 ng/kg LPS found this dose less tolerable for females. Healthy male participants were recruited using online advertisements and CHDR's database. Participants were included if they were between 18 and 35 years old at the time of signing the informed consent, if they were overtly healthy based on standard screening assessments, and with a body mass index (BMI) of 18.0–30.0 kg/m² and a minimum body weight of 50 kg. Main exclusion criteria were a history of hip fracture or surgery in the pelvic area, previous participation in a systemic LPS challenge trial within the previous year, and febrile illness within 30 days before the start of the first study day. No concomitant medications were allowed within 7 days before study start.

Study design and treatment

This was an open‐label, single‐center study consisting of a validation part and a study part. Participants in the validation part (n = 3) underwent only baseline blood and bone marrow sample collection. The study setup is shown in Figure 1 . An open‐label design was deemed appropriate because no bias was expected for the primary endpoints used in the study.

Figure 1.

Study design overview. Peripheral blood and bone marrow of healthy participants were sampled on Day 1 and Day 7. LPS was administered intravenously on day 7; afterwards, blood and bone marrow were collected at 4 h post‐LPS. Immunophenotyping, cytokine analysis, bulk RNA sequencing, and immunohistochemistry were performed. Figure created with BioRender.com.

Participants were institutionalized on days 1 and 7 (validation part included Day 1 only). On Day 7 (± 1 day), all participants received intravenous (i.v.) prehydration (1.5 L 2.5% glucose/0.45% sodium chloride). Two hours afterward, a 2‐min bolus injection of either 1 or 2 ng/kg LPS (E. coli O113:H10:K‐negative; US Standard Reference Endotoxin, manufactured by List Biological Laboratories) was administered i.v. and hydration was continued (at an infusion rate of 150 mL/h for 6 h). Blood was collected at baseline, 1, 2, 4, 6, and 24 h post‐dose. Bone marrow was collected at baseline on Day 1, and at 4 h post‐dose on Day 7. This timepoint was chosen to capture the peak of the inflammatory response. ¹⁴ , ¹⁵ Bone marrow aspiration was performed by a trained physician and collected using T‐Lok™ Bone Marrow Biopsy Needle (8G × 4 in) in 10 mL dipotassium ethylenediaminetetraacetic acid (K2EDTA) tubes. Blood was collected in K2EDTA tubes for immunophenotyping and cytomorphology, in sodium heparin tubes for cytokine analyses and in PAXgene™ Blood RNA tubes for bulk RNAseq. The follow‐up visit was scheduled 7–11 days after Day 7 assessments.

Safety assessments

Safety assessments included systolic and diastolic blood pressure, pulse rate, respiratory rate, temperature, and physical examination. Pulse, blood pressure, and ECGs were taken after 5 min in the supine position.

Pain intensity after bone marrow puncture was measured by using a visual analog scale (VAS), a horizontal scale of 100 mm in length, anchored by “no pain” (score of 0) and “worst imaginable pain” (score of 100). Participants assessed pain during both study days: on Day 1 prior to and immediately after bone marrow sampling, at 1 and 6 h; on Day 7 prior to LPS administration, at 1 h, 2 h, 4 h (immediately after bone marrow sampling), 6 h, 24 h, and during the follow‐up visit.

Blood was collected for hematology, chemistry, glucose, coagulation, and serology parameters; urine was collected at screening for dipstick analysis and drug screening.

Bone marrow aspirate and whole blood immunophenotyping

Within 1 h after collection, K2EDTA whole blood and K2EDTA bone marrow were incubated with red blood cell lysis buffer (BioLegend, catalog number 420301) for 5–10 min. Cells were washed, counted, and stained using standard procedures. Two immunophenotyping panels, myeloid and lymphoid, were assessed in both blood and bone marrow (full antibody list in Table S1 ). After a final washing step, the cells were analyzed using the MACSQuant 16 analyzer (Miltenyi Biotec, Bergisch Gladbach, Germany).

Plasma cytokine measurements

Venous blood for cytokine analysis was collected in sodium heparin tubes at pre‐dose, and 1 h, 2 h, 4 h (prior to bone marrow sampling), 6 h, and 24 h post‐dose. Plasma was collected after centrifugation at 2000 g for 20 min at 2–8°C. Samples were analyzed by Ardena Laboratories Assen. Concentrations of cytokines in plasma were measured using MSD multiplex assays (full panel and limits of quantification in Table S2 ).

Bulk RNA sequencing and data analysis

RNA isolation was performed on whole blood and bone marrow specimens for whole transcriptome sequencing. For RNA isolation, PAXgene® tubes were properly mixed and incubated for 2 h at room temperature before starting the RNA isolation procedure. Isolations were performed using the PAXgene® Blood RNA kit and PAXgene® Bone Marrow RNA Kit, respectively, according to manufacturer's protocol. Bulk RNAseq was performed at GenomeScan BV, Leiden, using the Illumina NovaSeq 6000. Data visualization and statistical analyses of the RNAseq dataset were performed using R statistical software (v4.4.1). ¹⁸ An overview of samples and associated information is provided in Table S3 , and details of RNAseq analysis are in Supplementary Materials and Methods . Briefly, the molecular response to LPS was investigated using differential expression analysis (DEA) and subsequent pathway analyses to compare the following groups of samples: baseline whole blood, baseline bone marrow, 4 h post‐dose whole blood (1 and 2 ng/kg), and 4 h post‐dose bone marrow (1 and 2 ng/kg). The abundances of 22 immune cell type populations across all samples in the RNAseq dataset were quantified using CIBERSORTx. ¹⁹

Identification of tractable drug targets and integration with cancer hallmarks

Differentially expressed genes (DEGs) were integrated with the OpenTargets platform ²⁰ to identify tractable targets and cross‐referenced with cancer hallmarks using COSMIC database annotations. ²¹ Tractable targets were defined as DEGs with existing OpenTargets annotations, and target tractability was assessed for small molecules and antibodies. Additionally, protein–protein interaction analyses were performed using STRINGdb. ²² Full methodological details and definition of tractable targets are provided in Supplementary Materials and Methods .

Immunohistochemistry (IHC)

Bone marrow punch biopsies (1.5 mm diameter) were collected from the iliac crest of healthy male participants at baseline and 4 h after i.v. LPS. Specimens were decalcified, fixed in 4% formaldehyde, and paraffin embedded. Biopsies were stained for c‐Myc and E‐cadherin using automated immunohistochemistry, scanned digitally, and quantified using validated image analysis workflows. Full methodological details are provided in Supplementary Materials and Methods .

Statistics

No formal power calculation was performed for this exploratory study. The sample size of three participants in the validation part and 10 participants in the study part was based on previous experience with systemic LPS challenge studies and typical cohort sizes in early‐phase clinical studies.

Immunophenotyping data and plasma cytokine concentrations were analyzed separately for bone marrow and whole blood. For endpoints with a single post‐baseline measurement, an analysis of covariance (ANCOVA) model was applied with treatment as a fixed factor and baseline value as a covariate. For endpoints with repeated post‐baseline measurements, a mixed‐effects model was used with treatment, time, and treatment‐by‐time interaction as fixed effects, subject as a random effect, and baseline value as a covariate. Pre‐dose values below the lower limit of quantification (LLOQ) were imputed as 50% of the LLOQ, and values above the upper limit of quantification (ULOQ) were set to the ULOQ. Cytokines for which all 24 h post‐dose values were below the LLOQ (IL‐10, IL‐6, and MIP‐1α) were analyzed without the 24 h timepoint. For all analyses, contrasts between 2 ng/kg and 1 ng/kg LPS dose levels were estimated within the respective models and contrasts relative to baseline were evaluated. Details are provided in the Supplementary Materials and Methods .

RESULTS

Systemic inflammatory response induced by LPS coupled with bone marrow sampling was well tolerated

A total of three participants were enrolled in the validation part and 10 participants were administered i.v. LPS (5 participants in the 1 ng/kg group and 5 participants in the 2 ng/kg group). All participants completed the study. Baseline characteristics are provided in Table S4 .

LPS administration induced expected transient flu‐like symptoms and low‐grade fever in most participants. The combination of systemic LPS administration and bone marrow sampling was safe and well tolerated. Pain associated with bone marrow sampling, assessed using VAS, was higher after LPS administration compared with baseline but remained within a tolerable range and showed no clear dose‐dependency. A summary of adverse events and pain assessments is provided in Table S5 and Figure S1 .

LPS triggered myeloid cell activation in blood and less pronounced myeloid responses in bone marrow

To establish the cellular context for transcriptomic analyses and understand compartment‐specific differences, immunophenotyping was performed on whole blood and bone marrow samples obtained at baseline and 4 h post‐dose. This timepoint was selected based on prior studies demonstrating pronounced neutrophil responses following systemic LPS administration. ¹⁴ , ¹⁶ Following LPS administration, the total number of granulocytes in peripheral blood increased, most prominently after 2 ng/kg. In bone marrow, granulocyte numbers increased after 1 ng/kg but not after 2 ng/kg (Figure 2 a ). In blood, neutrophils showed evidence of activation, including increased CD11b + and a dose‐dependent decrease in CD62L+ neutrophils. In contrast, bone marrow neutrophils showed increased proportions of both CD11b+ and CD62L+ neutrophils, although the latter was observed only after 1 ng/kg (Figure 2 b ).

Figure 2.

Intravenous LPS‐induced myeloid cell activation in peripheral blood and less pronounced myeloid cell responses in the bone marrow. Granulocyte (a) and monocyte (e) cell count in blood and bone marrow. Cell activation markers CD11b and CD62L (L‐selectin) in neutrophils (b) and monocytes (f) in blood and bone marrow. (c) Distribution of neutrophils based on the neutrophil maturation marker CD16, presented as mean percentage of CD16^hi, CD16^int, and CD16^lo neutrophils. Neutrophil (d) and monocyte (h) abundance in blood and bone marrow, derived using CIBERSORTx. Post‐LPS samples were pooled for analysis. (g) Mean percentage of classical, intermediate and nonclassical monocytes, as percentage of total monocytes. Data are represented as mean ± SD (a, b, e, f) and median ± IQR (d, h). P‐values were calculated from a mixed‐model ANOVA; all baseline and post‐dose values were pooled, and comparison between two different dose levels was performed (a, b, e, f). Wilcoxon signed rank test with a Benjamini–Hochberg correction was used to calculate the P‐values for CIBERSORTx analysis (d, h). If the P‐value is not shown, the difference was not statistically significant. IQR, interquartile range; SD, standard deviation.

Subdivision of neutrophils based on CD16 expression revealed distinct compartment‐specific responses (Figure 2 c ; gating strategy provided in Figure S2 ). In peripheral blood, the proportion of CD16^hi mature neutrophils remained high post‐dose, accompanied by an increase in CD16^int intermediate neutrophils and a decrease in CD16^lo immature neutrophils. In bone marrow, CD16^hi neutrophils increased post‐dose, whereas CD16^int neutrophils decreased, and CD16^lo neutrophils showed a dose‐dependent response (increase after 1 ng/kg but decrease after 2 ng/kg). Consistent with immunophenotyping results, CIBERSORTx analysis of bulk RNAseq data indicated increased neutrophil abundance in both blood and bone marrow following LPS administration (Figure 2 d ; Figure S3 ).

In contrast to neutrophils, circulating monocyte numbers decreased after LPS administration, whereas bone marrow monocyte numbers remained stable after 1 ng/kg and decreased following 2 ng/kg LPS (Figure 2 e ). In both compartments, the proportion of CD11b + monocytes increased post‐dose, indicating monocyte activation. CD62L+ monocytes showed a distinct dose‐dependent response, with a significant decrease after 2 ng/kg LPS compared with 1 ng/kg, in both peripheral blood and bone marrow (Figure 2 f ). Analysis of monocyte subsets showed a shift toward classical monocytes, accompanied by a reduction in intermediate and nonclassical subsets, particularly in the bone marrow (Figure 2 g ). Consistent with these findings, CIBERSORTx analysis revealed a trend toward reduced monocyte abundance in peripheral blood and a significant reduction in monocyte abundance in bone marrow following LPS administration (Figure 2 h ). Changes in dendritic cells and lymphocyte populations, as well as detailed analyses of granulocyte and monocyte subsets are provided in Figures S3 and S4 and Table S6 .

Rapid, dose‐dependent inflammatory cytokine responses occurred following LPS administration

Next, we profiled the time course of the cytokine responses to LPS in peripheral blood, as shown in Figure 3 a . LPS administration induced a rapid, dose‐dependent increase in pro‐inflammatory cytokines, including TNF, IL‐2, IL‐8, IL‐6, and MIP‐1α, which peaked at 2 h post‐dose. Other cytokines, including IFN‐γ and IL‐10, peaked at 4 h post‐dose, while IL‐18 remained elevated at 24 h post‐dose. Consistent with circulating protein responses, expression of multiple inflammatory cytokine genes, including IL1A, IL1B, IL18, IL10, and CCL4, was significantly upregulated in both blood (Figure 3 b ) and bone marrow (Figure 3 c ) compared with baseline. In addition, IL6 and CXCL12 were significantly upregulated in bone marrow but not in blood, whereas IL16 was significantly downregulated in bone marrow. Cytokines time–concentration and statistical analyses are provided in Figure S5 and Table S7 .

Figure 3.

Intravenous LPS‐induced dose‐dependent secretion of pro‐inflammatory cytokines and upregulated cytokine‐related genes in peripheral blood and bone marrow. (a) Cytokine time–concentration heatmaps. Cytokines are categorized based on their primary functional effects. Values are the least‐squares means (LSM) from a mixed‐model ANCOVA, normalized per cytokine to obtain a value between −1 (minimum; white) and 1 (maximum; dark red), with baseline defined as 0. Heatmaps illustrate expression (z‐scores of variance‐stabilized [VST] counts) of differentially expressed cytokine‐related genes in peripheral blood (b) and bone marrow (c).

Transcriptional profiling revealed broad inflammatory gene upregulation in blood, while bone marrow activated unique cell cycle and hematopoiesis pathways

Having established the cellular and cytokine‐level responses to LPS, we next examined genome‐wide transcriptional changes to identify differentially expressed genes and pathway enrichment. At baseline, blood and bone marrow showed distinct transcriptional profiles. Following LPS administration, unsupervised clustering revealed the presence of four distinct clusters of samples, corresponding to blood and bone marrow at baseline and post‐dose, respectively (Figure 4 a ). Samples from the 1 and 2 ng/kg dose groups clustered together within each compartment, indicating largely similar transcriptional responses across doses.

Figure 4.

Distinct transcriptional changes were induced by LPS in peripheral blood and bone marrow. (a) Sample clustering visualized by t‐distributed Stochastic Neighbor Embedding (t‐SNE). (b) Volcano plots showing results of the differential expression analysis (DEA) in peripheral blood and bone marrow, with selected DEGs labeled. All post‐LPS samples were pooled for analysis. (c) Bar plots displaying gene set enrichment analysis (GSEA) normalized enrichment scores (NES) for a selection of significantly enriched pathways in LPS‐treated samples compared with baseline in blood and bone marrow.

In peripheral blood, LPS induced broad transcriptional activation characterized by upregulation of inflammatory and cytokine‐responsive genes (top upregulated DEGs included DHRS13, OSM, HLX, PIM3, SOCS3, IL1A, CAMK1G, GPR84, LINC01093, SLC9A4) (Figure 4 b ). Pathway analysis revealed significant enrichment of immune‐related pathways, including TNF/NF‐κB signaling, interferon responses, and cell apoptosis (Figure 4 c ). Additionally, pathways related to PI3K‐Akt signaling, atypical NF‐κB signaling, and IL8‐CXCR1 signaling were uniquely enriched in blood.

In bone marrow, LPS also induced inflammatory gene expression (top upregulated DEGs included HCAR3, HCAR2, SOCS3, ADM, CD55, CCL7, LINC01093, IL1A, ACOD1, and SLC9A4), although the transcriptional response was distinguished by significant activation of pathways related to cell cycle progression and hematopoiesis, including G2M checkpoint, E2F targets, M phase and mitotic spindle pathways (Figure 4 b , c ). In contrast to blood, the Fc gamma receptor signaling pathway was upregulated in bone marrow.

Consistent with immunophenotyping findings, genes encoding markers of myeloid activation and trafficking, including ITGAM, CD14, and SELL, were upregulated in blood, whereas transcripts associated with lymphocytic populations were downregulated. In bone marrow, genes involved in neutrophil retention and mobilization showed compartment‐specific regulation, including downregulation of CXCR4 and upregulation of CXCL12.

Overall, these results demonstrate that systemic LPS induces robust and measurable inflammatory transcriptional responses in both blood and bone marrow, while also activating hematopoietic and proliferative programs in the bone marrow compartment.

LPS stimulation upregulated tractable oncology drug targets in blood and bone marrow

To determine whether LPS induced the overexpression of genes that are also tractable drug targets, we cross‐referenced upregulated DEGs with the OpenTargets database ²³ (Figure 5 a ). In total, 1914 and 716 tractable targets were identified in blood and bone marrow, respectively, of which 1462 and 264 were unique to each compartment. Among the most prominent tractable targets in blood were MMP9, PFKFB3, BCL3, OSM, PIM3, ITGAM, and HCAR2/3 (Figure 5 b ). In bone marrow, highly upregulated tractable targets included NNMT, CD163, HSPA8, CD36, and MARCO.

Figure 5.

LPS upregulates tractable drug targets in peripheral blood and bone marrow of healthy participants. (a) Flowchart illustrating the identification of LPS‐induced oncology‐related tractable targets. Differentially expressed genes (DEGs) were filtered using OpenTargets annotation based on drug development stages. The resulting targets were cross‐referenced with COSMIC database hallmarks of cancer information to identify oncology‐related tractable targets. (b) Heatmap showing the expression of selected tractable targets upregulated in peripheral blood and bone marrow.

The identified genes were subsequently cross‐referenced with the hallmarks of cancer classification ¹⁰ in order to identify oncology‐relevant tractable targets. In total, 17 tractable targets common to blood and bone marrow were associated with cancer hallmarks (Figure S6B ), alongside 74 targets unique to blood (Figure S6C ) and 7 targets unique to bone marrow (Figure S6E ). Across both compartments, tractable targets were primarily associated with hallmarks such as resisting cell death, activating invasion and metastasis, and sustaining proliferative cell signaling. Tractable targets within cancer hallmarks unique to bone marrow were MYC, SRC, CALR, FAT1, CDH1, and GNA11. Subsequently, protein–protein interaction analysis was performed using STRINGdb ²² (Figure S6A,D ). In blood, highly connected nodes within the interaction network included BRCA1, PTEN, BRAF, MAPK1, KRAS, STAT3, HIF1A, and JAK1. In bone marrow, central network nodes included SRC, STAT3, MYC, and CDH1.

To assess whether LPS‐induced transcriptional changes translated to protein‐level effects in bone marrow, immunohistochemical staining was performed on bone marrow biopsies obtained at baseline and 4 h post‐dose for c‐Myc (encoded by MYC), due to its central role in cell proliferation, and E‐cadherin (encoded by CDH1), due to its importance in cell adhesion (Figure 6 a ). c‐Myc showed a dose‐dependent increase in percentage of positive cells following LPS administration, whereas E‐cadherin showed a nonsignificant trend toward increase at both LPS doses compared with baseline (Figure 6 b ). We present in Figure 7 an overview of tractable targets categorized by cancer hallmarks and compartment, with examples of approved or investigational drugs targeting these oncogenic drivers. A full list of tractable targets categorized by cancer hallmarks is provided in Table S8 .

Figure 6.

Analysis of c‐Myc and E‐cadherin protein expression in human bone marrow following LPS administration. (a) Representative images of bone marrow biopsies stained with H&E, c‐Myc, and E‐cadherin at baseline and at 4 h after 2 ng/kg LPS (original magnifications 5× and higher magnification inserts at 20×). (b) Quantification of the percentage of cells positive for c‐Myc and E‐cadherin, and corresponding MYC and CDH1 RNA log2 fold changes from differential expression analysis (DEA) of bone marrow aspirates, relative to baseline. For quantitative analysis of biopsies, percentage of positive cells represents the average of two biopsy sections per individual, and statistical analysis was performed using paired t‐test (n = 4 for 1 ng/kg LPS [one post‐dose biopsy not harvested]; n = 5 for 2 ng/kg LPS). Increase in percentage of c‐Myc‐positive cells was 6.77 ± 3.47 at 1 ng/kg LPS (P = 0.03, n = 4), and 13.77 ± 3.26 at 2 ng/kg (P < 0.001, n = 5), compared with baseline. E‐cadherin showed nonsignificant trend toward increased protein levels at both LPS doses compared with baseline. Data are shown as mean ± SD. DEA, differential expression analysis; H&E, hematoxylin and eosin; log₂FC, log₂ fold change; ***P < 0.001; *P < 0.05; ns, not significant.

Figure 7.

Distribution of oncology‐related tractable drug targets across the hallmarks of cancer. The outer circle (gray border) denotes targets commonly upregulated in peripheral blood and bone marrow, the middle circle (light red border) targets uniquely upregulated in blood, and the inner circle (dark red border) targets uniquely upregulated in bone marrow. The segments represent the percentage of tractable drug targets per cancer hallmark. Representative targets per hallmark are displayed alongside example drugs inhibiting these targets. Targets associated with multiple hallmarks are counted in each relevant category. SRC and MYC, uniquely upregulated in bone marrow, were associated with the majority of cancer hallmarks, and so are highlighted separately. The accompanying bar chart quantifies the number of tractable targets corresponding to each hallmark of cancer.

Lastly, we assessed the potential translational utility of LPS‐induced transcriptional changes by presenting in Table S9 a selection of PD markers which are associated with the identified tractable targets, together with relevant drug candidates or approved drugs. For several key pathways, validated phosphorylation‐based markers were already broadly used in clinical trials, such as pSTAT3 for JAK/STAT inhibitors, pERK1/2 for RAS/MAPK inhibitors, and pAKT for PI3K pathway inhibitors. For targets lacking canonical phospho‐markers, circulating protein levels or target mRNA quantification can serve as PD markers, such as MMP9 protein levels in plasma (for CXCR1/2 antagonists) or HEXIM1 mRNA levels in tissues (for BET inhibitors). Importantly, targets without established clinical assays, including CALR, CDH1, YAP1 and GNA11, which were identified in the bone marrow in our study, may be likewise assessed by direct quantification of mRNA. Altogether, these findings demonstrate that LPS challenge can provide a controlled in vivo context for evaluating target engagement across a range of oncology‐relevant pathways, by using either established phosphorylation readouts or protein‐ and transcript‐level endpoints.

DISCUSSION

We identified LPS‐induced tractable targets in blood and bone marrow that are relevant as therapeutic targets in oncology and could serve as PD markers in future early‐phase oncology trials. Furthermore, we have established a foundation for using immune challenge models in oncology drug development and demonstrated that integrating bone marrow analyses into healthy volunteer studies for PD assessments is feasible. While human in vivo responses to i.v. LPS have been studied in the context of hyperinflammatory and immunosuppressive conditions, ²⁴ our work specifically focuses on identifying targets for oncology drug development.

Consistent with previous studies, systemic administration of 1 or 2 ng/kg of LPS to healthy males was safe and induced a transient inflammatory response characterized by neutrophil activation and redistribution. ¹⁴ , ²⁵ Importantly, although bone marrow sampling following LPS administration was associated with somewhat higher pain scores compared with baseline, the procedure remained well tolerated. On a cellular level, the increase in CD11b expression and shedding of CD62L on circulating neutrophils has been described and reflects neutrophil activation and migration. ²⁶ In bone marrow, increased proportions of CD62L+ neutrophils were accompanied by transcriptional changes in genes involved in neutrophil retention and mobilization, including downregulation of CXCR4 and upregulation of CXCL12, suggesting that LPS induced a release of mature neutrophils from the bone marrow to the circulation. ²⁷ These compartment‐specific observations highlight the advantage of controlled in vivo challenge models, compared with ex vivo whole blood or PBMC stimulation or other in vitro systems, which are confined to circulating cells and cannot fully capture the complex physiological dynamics demonstrated here. Incorporating the in vivo LPS challenge model in early‐phase studies with novel compounds allows for the integration of drug pharmacokinetics with target engagement and provides proof of in vivo drug activity accounting for drug distribution, metabolism, and bioavailability.

Administration of LPS upregulated multiple tractable cancer‐related genes, which provides an opportunity to evaluate individual PD markers in a controlled in vivo setting. Notably, LPS was previously shown to have dual effects in tumor immunity, depending on the duration of exposure (acute vs. chronic), and the location of effect (healthy immune system vs. tumor microenvironment). Acute TLR4‐mediated inflammation has been investigated as an intra‐tumoral immunomodulatory therapy ²⁸ and vaccine adjuvant, ²⁹ and synthetic TLR4 agonists were investigated as cancer treatments. ³⁰ , ³¹ , ³² Conversely, chronic LPS‐driven inflammation can promote T‐cell exhaustion and resistance to anti‐PD‐1 therapy in murine models, ³³ as well as enhance metastasis through activation of inflammasome. ³⁴ In the context of early‐phase drug development, we focused on upregulated tractable targets due to their relevance for evaluating PD effects of antagonistic molecular targeted agents, where LPS serves as a controlled inflammatory challenge. Conversely, for therapies relying on pro‐inflammatory mechanisms, LPS may function as a prototype drug. Downregulated targets, while potentially informative, were therefore not the primary focus of this study.

Regardless of the role of TLR4‐mediated processes in oncological disease, LPS may serve as a valuable tool for evaluating the effects of novel cancer therapeutics. One example is the PI3K/Akt signaling pathway, which was induced following LPS administration and characterized by increased expression of PIK3CA, PIK3CB, PIK3CG, PIK3CD, and PTEN. This pathway is crucial in regulating cell growth and survival and is frequently overactive in cancer. PD effects within this pathway can be assessed using established readouts, including phosphorylation of AKT and mTOR/pS6K, alongside transcriptional and cellular readouts. An illustrative example is the phase I study of the selective PI3Kδ inhibitor ME‐401 investigated for B‐cell malignancies conducted in healthy participants, in which PD guided efficient dose selection prior to patient studies. ³⁵ This study also highlights efficient dose‐finding, as 60 mg ME‐401 (out of a maximum of 150 mg) was chosen for the subsequent patient trial due to the lack of additional inhibitory effects at higher doses and a superior toxicity profile. Since LPS transiently induced PIK3CD, it could provide a controlled setting to evaluate pathway‐specific drug effects, including PI3Kδ inhibition and downstream phosphorylation of AKT and mTOR/pS6K. Another example is IL‐8 (CXCL8), which was induced by LPS and is known to drive expression of MMP9, one of the most strongly upregulated genes in our study. The IL‐8/MMP9 axis has been implicated in tumor invasion and metastasis, and both components have been explored as therapeutic targets in oncology. In this context, strategies targeting CXCR1/CXCR2 signaling or MMP9, including CXCR1/CXCR2 antagonists and anti‐MMP9 antibodies, have entered clinical development. ³⁶ , ³⁷ The concurrent induction of IL‐8 at the protein level and MMP9 at the transcriptomic level following LPS challenge provides a controlled in vivo setting to evaluate PD effects of novel therapies targeting this pathway in healthy participants. For targets such as MMP9 and CXCL8 that lack canonical phospho‐markers, circulating protein levels and mRNA quantification offer potential PD markers, and both approaches have already been explored in clinical trials of MMP9 and CXCR1/2 inhibitors. ³⁸ , ³⁹

Incorporating bone marrow analyses in clinical studies provides unique opportunities to assess PD relevant to oncogenic drivers that are not readily captured in peripheral blood. In this study, LPS administration induced a significant upregulation of MYC expression in bone marrow at both transcriptomic and protein levels, which is consistent with its central role in regulating hematopoietic cell proliferation and survival. ⁴⁰ Dysregulation of MYC is common in cancer, and despite its long‐standing classification as undruggable, direct MYC inhibition has recently entered clinical development, exemplified by the successful completion of a phase I trial with the MYC inhibitor OMO‐103. ⁴¹ In addition, targeting transcriptional regulation of MYC through inhibition of bromodomain and extra‐terminal domain (BET) proteins represents an established therapeutic strategy, particularly in hematologic malignancies. ⁴² BET inhibitors suppress MYC‐driven transcriptional programs and have also demonstrated the ability to attenuate LPS‐induced inflammatory gene expression in preclinical models. ⁴³ , ⁴⁴ Induction of HEXIM1 mRNA was used as a transcript‐level endpoint for BET‐mediated MYC suppression in clinical studies of BET inhibitors. ⁴⁵ Current clinical development of BET inhibitors in immuno‐inflammatory indications, in phase I in healthy participants, ⁴⁶ shows potential of such PD‐guided trials in both immunology and oncology settings.

Our study also illustrates a broader principle, namely, that transcriptomics can be used to narrow the selection of potential PD markers for subsequent clinical studies with LPS challenge model, or other controlled in vivo models. Nomination of a candidate PD marker based on transcriptional induction is therefore a hypothesis‐generating step, and prospective validation would be required before broader use in a drug development program. Importantly, however, mRNA‐based PD markers are increasingly used in early‐phase oncology programs, such as AXIN2 mRNA for porcupine inhibitors ⁴⁷ or HES1 for γ‐secretase inhibitors, ⁴⁸ , ⁴⁹ either alone or in combination with phospho‐based markers.

Our study has some limitations. First, with a small cohort size, our findings should be interpreted as exploratory, although the consistency of LPS‐induced responses across individuals and two LPS dose levels strengthen the reliability of our findings. Second, the hallmarks of cancer framework may not fully capture immune‐related targets such as PI3Kδ, which is predominantly expressed in leukocytes, and future extensions may better integrate hallmarks of immune‐oncology. ⁵⁰ In addition, the COSMIC database does not yet fully incorporate recent updates to the hallmarks of cancer, including cancer‐enabling characteristics ⁵¹ and the most recent refinements. ⁵² We also acknowledge limitations of the OpenTargets platform, as tractability annotations may not be exhaustive and some targets may ultimately prove unfeasible to pursue. Furthermore, while we observed concordant LPS‐induced upregulation of MYC at both the RNA and protein levels in bone marrow, E‐cadherin showed only nonsignificant changes at the protein level, despite clear gene overexpression. This discordance may reflect limitations of our sampling timepoint (e.g., early mRNA upregulation preceding protein accumulation) or small sample size. Comprehensive validation in early‐phase (phase 0) clinical studies should include both RNA and protein endpoints to ensure the validity of candidate PD markers.

In conclusion, we identified tractable oncological targets in both blood and bone marrow of healthy participants following LPS‐induced inflammation, demonstrating feasibility of this model for early‐phase oncology drug development. Bone marrow provided complementary information about specific targets not detectable in peripheral blood at peak inflammation, and transcriptomic and protein‐level analyses enabled identification of candidate PD markers. Altogether, these findings establish a framework for integrating immune challenge models into PD‐guided early‐phase oncology clinical trials, which should ultimately lead to more efficient development of novel cancer therapies.

FUNDING

This study was entirely funded and conducted by the Center for Human Drug Research (CHDR). CHDR was involved in the study design, data collection, analysis, and interpretation, as well as in the writing of the report and the decision to submit the article for publication.

CONFLICT OF INTEREST

The authors declared no competing interests for this work.

AUTHOR CONTRIBUTIONS

IR and IT wrote the manuscript; JJB, MM, and IR designed the research; IR, JJB, HWG, DRP, MO, PvB, MAAJ, and LP performed the research; IT, MJJ, and MLdK analyzed the data.

Supporting information

Figure S1.

Figure S2.

Figure S3.

Figure S4.

Figure S5.

Figure S6.

Table S1.

Table S2.

Table S3.

Table S4.

Table S5.

Table S6.

Table S7.

Table S9.

Table S8.

Data S1.

DATA AVAILABILITY STATEMENT

The RNAseq datasets that support the findings of this study are not openly available due to reasons of confidentiality and are available from the corresponding author upon reasonable request.