Comparison of transcriptomic profiles between intracellular and extracellular Bartonella henselae

Abstract

The Bartonella genus of bacteria encompasses ubiquitous species, some of which are pathogenic in humans and animals. Bartonella henselae, the causative agent of Cat Scratch disease, is responsible for a large portion of human Bartonella infections. These bacteria can grow outside of cells, replicate in erythrocytes and invade endothelial and monocytic cells. We have previously reported reduced antibiotic susceptibility of intracellular Bartonella. In this study we performed comparative transcriptomic analyses between the extracellular and intracellular B. henselae phenotypes. Overall, specific genes involved in invasion, virulence, extracellular adhesion of type 4 secretion system were downregulated following intracellular invasion of B. henselae. Downregulation included BadA, a well-characterized adhesin molecule, of critical importance for cell invasion. These studies demonstrate the ability to purify Bartonella RNA from infected cells and offer a repository of gene expression data for future research. The development of novel therapeutics will benefit from the ability to determine target expression by Bartonella in relevant microenvironments.

Transcriptomic analysis of Bartonella henselae, differentiating between the extracellular (planktonic) and intracellular environment. Changes in gene expression reflect virulence, invasion and adhesion for extracellular matrices.

Introduction

Bartonella henselae is a gram-negative zoonotic and facultative intracellular bacterium, which infects numerous animal species, including humans and cats; 30–60% of domestic or feral cats in the U.S. are infected with B. henselae. Transmission among cats is mediated by the cat flea, Ctenocephalides felis, and to humans by cat scratches or cat bites^1,2. As the primary reservoir host, chronic, relapsing bacteremia occurs in cats, most often without clinical signs of disease. Infection of immunocompetent individuals with B. henselae results in cat scratch disease (CSD), which is a usually a self‐limiting infection³. In contrast, immunosuppressed and immunocompetent individuals infected with B. henselae can develop systemic infections accompanied by diverse pathologies involving the cardiovascular, musculoskeletal and nervous systems^4–14.

B. henselae has the capacity to survive in murine macrophages, and dog and human endothelial cells by inhibition of phagosome fusion with lysosomes, where the bacteria replicate within these cells^15–18. As facultative intracellular pathogens, the proposed pathogenesis involves initial infection of endothelial cells, after which Bartonella infects a subpopulation of erythrocytes^19,20. Adhesion to endothelial cells by B. henselae has mainly been studied in human umbilical vein endothelial cells, which has provided an in vitro system to study the basis of angioproliferation²¹. B. henselae expresses surface adhesins, including BadA²², which is essential for the attachment to host extracellular matrix (ECM) proteins including collagen, laminin, and fibronectin, with fibronectin of particular importance for bacterial adhesion^23,24. The interaction with fibronectin occurs at repetitive motifs in the BadA neck/stalk region²⁵ that helps the bacteria to adhere to host cells. A recent study has evaluated the role of fibronectin as a mediator in bacterial cell-host adhesion, where the fibronectin knockout diminished Bartonella adherence to the endothelial cells²⁶, documenting a pivotal role of this ECM protein for bacterial adhesion. Bartonella, once attached, is known to be internalized by conventional phagocytosis and by an invasome-mediated internalization mechanism in endothelial cells^17,27. Generally, bacterial internalization leads to apoptosis of the host cell. However, Bartonella evolved the anti-apoptotic effect²¹ that is mediated by the pathogen’s BepA protein that activates human adenylyl cyclase 7 to elevate cAMP levels^28,29, resulting in enhanced bacterial survival within host cells. Colonization of endothelial cells is important in both reservoir and incidental mammalian hosts and is considered essential for the establishment and maintenance of the angioproliferative lesions reported in cats, dogs, humans and other animals^1,21,30.

A case report from 1985 showed the presence of gram-negative bacilli, presumed to be B. henselae, as clumps in vessel walls, observed both intracellularly and free in necrotic debris in a patient with CSD³¹. These observations of bacteria located near or adjacent to the cell nuclei were made using staining techniques, such as Warthin-Starry silver stain, which are not species-specific. In another case report of five patients, histopathologic findings demonstrated the presence of bacteria extracellularly and inside macrophages of skin biopsy specimen³². Similarly, these bacteria were found both extracellularly and intracellular in areas of necrosis in lymph node. Later, another study identified the presence of extracellular bacteria only, forming small groups in necrotic and viable regions of lymph node biopsies from two different patients³³. An ultrastructural analysis of the lymph node of a CSD patient with AIDS demonstrated the presence of clusters of small, pleomorphic bacteria in degenerated extracellular collagenous tissue of the lymph node as well as in the walls of blood vessels³⁴. A few bacteria were also observed in phagosomes of macrophages. More recently, B. henselae was detected in a melanoma biopsy tissue specimen, where the bacteria appeared to be intracellular⁵. These observations collectively support B. henselae’s dual lifestyle in human infections. This adaptability likely enhances the pathogen’s survival by allowing it to employ different virulence strategies based on its location. For instance, intracellular B. henselae in melanoma cells increased expression of VEGF and interleukin-8, potentially promoting angiogenesis⁵. This ability to exist both intracellularly and extracellularly may contribute to the diverse clinical manifestations of Bartonella infections and its capacity to evade host immune responses³⁵.

Historically, to identify functional interactions of the transcriptional regulatory systems and to perform quantitative gene expression analyses, quantitative PCR (qPCR) was utilized. Using qPCR, researchers determined that the expression of the rpoE gene was upregulated when B. quintana was exposed to the body louse (28 °C) versus the human host (37 °C) temperature³⁶. Subsequently, a custom microarray platform was utilized to study and analyze temperature-specific and growth phase-specific transcriptomics of B. quintana at host and vector body temperatures³⁷. A cluster of growth-specific genes were found, and it was determined that B. quintana has unique transcriptional profiles when grown in vitro on agar at human host (37 °C) temperature compared with arthropod vector (28 °C) temperature³⁷. Microarray has also been utilized to study the gene expression of B. henselae during infection of human endothelial cells³⁸. Microarray technology was the first to document that BatR is a protein directly involved in the regulation of genes encoding the VirB Type IV secretion systems (T4SS)and the Bartonella effector proteins (Beps) that are associated with Bartonella pathogenicity³⁸. As RNA sequencing (RNA-seq) was developed and became more readily available, it was used to further study the importance of gene expression and regulation in Bartonella. RNA-seq has revealed nine novel, highly-expressed intergenic transcripts of B. henselae that belong to the Brt family of RNAs³. The identification of these genes has further enhanced our knowledge of their role in expression of a major virulence factor gene (badA) and the pathogenesis of B. henselae infections³.

Intracellular bacterial infections are often studied using cell lines as host models^39,40. DH82 cells are a canine macrophage cell line that was isolated from a 10-year old dog with histiocytic sarcoma⁴¹. This cell line is been applied as a useful in vitro tool for the investigation of canine macrophages in a wide range of scientific fields, including veterinary oncology and infectious diseases⁴². In the field of infection research, DH82 cells are capable of being infected with several bacterial species, including Rickettsia⁴³, Ehrlichia (sp. HF, a model bacterium to study fatal human Ehrlichiosis)⁴⁴, Brucella canis⁴⁵, and Leishmania infantum⁴⁶. Investigators have also used these cells to isolate and characterize novel, pathogenic, obligate intracellular strains of Ehrlichia minasensis⁴⁷ and Rickettsia tillamookensis⁴³. The DH82 cell line was also used as a model in the discovery of two Ehrlichia virulence genes that are responsible for host infection⁴⁸. Since this cell line is considered permissive for Bartonella, previous studies have utilized these cells in Bartonella serological testing using immunofluorescence^49–51 and in testing and determination of effective drug combinations that could potentially be used to treat bartonelloses⁵².

Similar to B. quintana and the human body louse, B. henselae adapts to very diverse environments, such as the cat flea vector where the temperature is low and the vertebrate host, where temperature is higher. Very little is known about how B. henselae modifies gene expression profiles to efficiently transition from the conditions of the cat flea vector to the mammalian host³. As mentioned above, a few reports have proposed a role for several different regulatory proteins in host adaptation by B. henselae and B. quintana^37,38,53,54. B. henselae can exist in both intracellular and extracellular forms during its infection cycle. These bacteria are found extracellularly, especially during initial stages of infection and transmission. BadA is used to attach the bacterium to host cells and ECM proteins⁵⁵. Then, the bacteria can stimulate angiogenesis through secreted factors like BafA, which interacts with host VEGF receptors⁵⁶. After the entry into the host cells, the bacteria use the VirB/VirD T4SS for survival, injecting effector proteins into host cells^55,57. Inside the cells, B. henselae can form aggregates within structures called invasomes^17,56. This ability of transition between extracellular and intracellular states helps Bartonella adapt to different host environments and evade immune responses. In this study, we used RNA-seq analysis to compare gene expression of B. henselae within DH82 cells to expression by the free-living bacteria. The compiled transcriptome of this pathogen provides a novel and essential contribution to the study of Bartonella infections and clinical interventions.

Results

Differentiation of invasive bacteria from surface-adhered bacteria

Cells were infected with different B. henselae multiplicities of infection (MOI), as described in our previous study⁵². An MOI of 1:100, where the highest percentage (100%) of DH82 cells were infected with Bartonella, was used in this study. To differentiate intracellular bacteria from the surface-adhered bacteria, we used MemBrite Fix staining to stain the host cell membrane, along with Bartonella-specific antibody to stain for the bacteria. Initially, DH82 cells were stained with either DAPI alone or in combination with MemBrite Fix staining to determine if the dye would stain the canine cell cytoplasm (Supplementary Fig. 1A, B). To determine the specificity of the secondary antibody, a secondary antibody alone control (Supplementary Fig. 1C) was included. Permeabilization was the only additional step that was added for staining intracellular bacteria. Permeabilization caused the internalization of the anti-Bartonella antibody into the cell, which distinguished intracellular bacteria from the bacteria attached to the cell surfaces. Comparison of fixed/non-permeabilized cells stained for Bartonella to the cells where permeabilization was employed clearly documented the intracellular localization of the bacteria (Fig. 1A, B; Supplementary Fig. 2). Confocal microscopy generated an image of the intracellular bacteria (Fig. 1C).

Fig. 1. Infection of DH82 cells with Bartonella henselae.

Indirect immunofluorescence images showing staining of Bartonella henselae in Panels A & B. The staining in Panel B was differentiated from Panel A by including an extra permeabilization step. Bacteria were processed for immunofluorescence assay using rabbit anti-Bartonella antibody and consequently stained using secondary antibody Goat Anti-rabbit IgG-Alexa Fluor 594 (red) and MemBrite Fix stain was used to stain the DH82 cell membrane (Green). The extracellular bacteria stained in red colocalized with the cell membrane stained in green. C A confocal image showing the intracellular bacteria in three dimensions. D The viable intracellular bacteria was stained using RNAscope.

A subsequent set of experiments was conducted to quantify the fraction of extracellular versus intracellular bacteria. Immunofluorescent staining was repeated, using different fluorescently-tagged antibodies before and after permeabilization. The counts from the individual fields were shown Table S1 and the representative images are shown in Supplementary Fig. 3. A total of at least 50 images were taken from each individual experiment and yellow dots (extracellular bacteria) were counted manually from each field. A total of four individual experiments were performed to finalize the average extracellular bacteria percentage, which was 12.795%

RNAscope technique to determine bacterial RNA viability

To optimize the protocol and to test the viability of the bacteria inside of DH82 host cells, a positive control RNA probe was used alongside a negative control probe. Supplementary Fig. 4 panel illustrates RNAscope staining of the positive control (PPIB), which is considered a standard to determine the RNA quality of the cells and the negative control probe targeting the dapB gene of Bacillus subtilis. To validate the presence of viable B. henselae inside the cells, we used an RNA probe targeting the B. henselae 23s ribosomal RNA gene (Fig. 1D).

RNA integrity check and enrichment of bacterial RNA

Using the osmotic lysis method, we were able to isolate 5 × 10⁵± 2.05 bacteria from the infected DH82 cells. After lysing the bacteria, RNA integrity was assessed using a Bioanalyzer (Agilent) and their nano chip (Fig. 2A). Electrophoretic analysis revealed strong bands corresponding to the 18S and 28S ribosomal RNA (rRNA) of the DH82 host cells, with faint bacterial RNA bands corresponding to the 16S and 23S rRNA in samples, which confirmed the predominance of eukaryotic ribosomal RNA in the samples and suggested that the host mRNA would probably substantially predominate over bacterial transcripts. To eliminate host RNA from the samples for bacterial transcriptomic analyses, the MICROBEnrich™ kit was used, which selectively removes 18S, 28S rRNA, and polyadenylated mRNA. To our knowledge, this is the first time this kit was tested on a Canine species cell line for the removal of RNA. Following MICROBEnrich™ treatment, electrophoretic analysis of the total RNA obtained with this method revealed the presence of bands corresponding to the 16S and 23S bacterial rRNA, whereas the host 28S and 18S rRNA were not detected. Since the MICROBEnrich™ kit doesn’t remove 5S and 5.5S rRNA, these bands were clearly seen in the enriched samples (Fig. 2A). Enriched samples 6, 7, and 8 from Fig. 2A were used as intracellular/experimental (inside DH82 cells) samples. Samples labeled controls 1, 2 and 4 from Fig. 2B were used as control samples (planktonic bacteria). The RNA integrity number of the control and the experimental samples before enrichment is listed in Fig. 2C. Since bacteria-enriched samples were run on the same chip as the pre-enriched samples, a RIN value was not generated for the enriched samples, possibly because of corresponding 5 and 5.5S bands from host cells.

Fig. 2. Evaluation of the RNA extracted from planktonic and intracellular bacteria.

A Image of the bioanalyzer gel for three independent RNA samples that were extracted from the DH82 cells that were infected with B. henselae. Sample 1 represents the RNA from the DH82 cells alone. Samples 2, 3, and 4 represent the RNA isolated from DH82 cells infected with B. henselae before enriching the sample for bacterial RNA. Sample 5 represents the RNA from B. henselae alone and samples 6, 7, and 8 represent the RNA after enrichment for bacteria. L, ladder. The 28S and 18S (host) rRNA bands, 23S and 16S (bacteria) rRNA bands are indicated by the black arrows. Before enrichment, faint bacterial rRNA bands can be visualized in samples 2, 3, and 4. After enrichment, 5 and 5.5S rRNA bands can be visualized in samples 6, 7, and 8. B Image of the bioanalyzer gel for three independent RNA samples that were extracted from B. henselae. Control 1, Control 2, and Control 4 represent the RNA from the B. henselae. C Table showing the RNA integrity values of the samples. In this manuscript, Control 1, 2 and 4 were used as planktonic samples and samples 6, 7, and 8 were used as experimental (intracellular) samples.

RNA-seq analysis of B. henselae isolated from DH82 cells

Using Lexogen RiboCop rRNA Depletion kits for DH82 cells and bacteria simultaneously, rRNA was removed from the in vitro grown bacteria directly and from enriched intracellular bacterial RNA. Libraries were generated, and Illumina sequencing was performed on RNA samples from B. henselae grown blood agar plates and in the enriched, intracellular bacteria (inside DH82 cells). Fig. 3 depicts the RNAseq pipeline for steps in the analysis process. Since a well-annotated genome of B. henselae San Antonio 2 strain is not available, we used B. henselae Houston-1 strain genome as the nearest annotated neighboring genome. The planktonic bacteria samples resulted in mapping more than 70–73% of the reads after processing for low-quality reads (Table 1). However, only 4–6% of the reads were mapped from the intracellular bacteria samples to the bacterial reference genome, suggesting that the rest of the reads were host RNA. Reads mapped to 99.57% of the genes on the B. henselae Houston-1 reference genome, giving sufficient data to describe the transcriptome. Only 7 genes out of 1665 had no read counts. However, due to large differences in the total number of mapped reads between planktonic and intracellular bacteria samples (Tables S2 and S3), we downsampled the in vitro reads (Table 1). This involved selecting a subset of data from the larger dataset to reduce the chances of overestimation errors generated by comparison of large to small number reads. We performed two rounds of downsampling, initially by reducing the total number of reads to 4.5 million and subsequently to 3 million reads. Downsampling the reads to 4.5 million generated more total number of reads from controls after processing, compared to the experimental samples, than we anticipated (Table S4). This served as the rationale behind doing another round of down sampling to 3 million reads from the original reads. After this processing, this second round resulted in a similar total number of reads among all the samples (Tables 2 and S4), which allowed for a more appropriate analysis with less saturation effects due to the excess data depth of the in vitro

Fig. 3. An overview of the RNA-sequencing pipeline.

This figure was created with BioRender.com.

Table 1.

Total number of reads that were generated per sample

Samplea	Number of input reads	Mapped reads	Mapped reads %	Reads left after UMI deduplication
C1	61,366,102	45,305,773	69.41	38,152,844
C2	71,867,118	52,676,704	71.35	43,622,424
C3	62,836,096	47,755,209	73.98	40,008,051
E1	57,392,557	4,081,782	5.84	2,615,517
E2	61,250,989	3,465,895	4.59	2,011,867
E3	63,862,636	3,937,077	5.11	2,502,664

Values in bold highlight the effect of down-sampling

^aSamples C1, C2, and C3 represent the in vitro (planktonic) bacteria and samples E1, E2, and E3 represent intracellular bacteria

Table 2.

Reads from control samples were downsampled to 3 million to generate similar number of reads across all the samples

Sample	Number of input reads	Mapped reads	Mapped reads %	Reads left after UMI deduplication
C1	3,000,000	2,226,020	74.77%	2,104,650
C2	3,000,000	2,208,581	74.23%	2,091,311
C3	3,000,000	2,294,677	77.12%	2,166,946
E1	57,392,557	4,082,878	7.18%	2,362,193
E2	61,250,989	3,478,325	5.74%	1,799,680
E3	63,862,636	3,941,004	6.23%	2,288,239

Values in bold highlight the effect of down-sampling

Gene expression of intracellular bacteria

We have included the B. henselae gene expression data from the use of 4.5 million reads and 3 million reads (Tables S5 and S7) and raw total counts from 3 million reads (Table S6). Even after rRNA depletion, we observed reads mapping to the rRNA. Prior to normalizing the counts, rRNA counts and genes that had zero read counts were removed and then were normalized. The results from the counts that didn’t contain rRNA were documented (Table S8 and the genes that were excluded from the analysis were added to the Table S9). We observed that the changes in the expression levels of 289 genes were changed ≥1.5-fold and 114 genes had log2fold change of ≥2-fold and were significantly (p < 0.05) increased or decreased when planktonic and intracellular growth conditions were compared. Overall, 70 genes were repressed, and 44 genes were activated with two or more-fold change (Fig. 4A). We have listed the top 20 up- and downregulated genes in Table 3 along with their normalized counts. A heatmap was generated using the Z-scores and presented as Fig. 4C. Based upon principal component analysis (PCA), the variation between the growth conditions (planktonic vs intracellular) was ~80% whereas, within the samples there was a variance of 12.4% (Fig. 4B). A total of 1652 gene normalized counts were used to generate the PCA plot (Table S10).

Fig. 4. Total number of genes that were differentially expressed between planktonic and intracellular conditions of B. henselae.

A Total number of genes that were statistically upregulated or downregulated in a comparison between B. henselae grown on a blood agar plate (planktonic) to B. henselae grown inside DH82 cells. B A PCA plot that was generated using ClustVis is shown. Normalized read counts from all the genes were used. The values were ln(x)-transformed and Pareto scaling is applied to rows and Nipals PCA was used to calculate principal components. The X and Y axes show principal component 1 and principal component 2 that explain 79.9% and 12.4% of the total variance, respectively; N = 6 data points. C Heatmap showing the Top 20 up and top 20 down genes that changed their gene expression significantly. Here we used Z-score of the normalized counts to generate a heatmap. ChatGPT 3.5 was used to write a script for generating this heatmap.

Table 3.

The top 20 upregulated genes and top 20 downregulated genes in the intracellular B. henselae along with their normalized counts

Top 20 upregulated genes in the intracellular bacteria							Normalized counts
ID	gene name	baseMean	log2FoldChange	PValue	PAdj	FDR	C1dedup	C2dedup	C3dedup	E1dedup	E2dedup	E3dedup
CAF27495-1	hypothetical genomic island protein	6943.3	6.9	0.00E + 00	0.00E + 00	0	55	64	50	6941	6498	7222
CAF27497-1	hypothetical genomic island protein	2864	5.7	0.00E + 00	0.00E + 00	0	54	55	52	2896	2627	2908
CAF27494-1	hypothetical protein	583.3	5.6	8.70E–166	0.00E + 00	0	19	8	9	559	568	587
CAF27498-1	hypothetical protein	352.3	5	4.10E–77	0.00E + 00	0	16	10	6	345	299	381
BH08960-1	tRNA-Asp	532	4.9	1.40E–94	0.00E + 00	0	21	19	12	548	600	396
BH12460-1	tRNA-Leu	223.3	4.3	3.10E–40	0.00E + 00	0	12	13	6	189	257	193
CAF28127-1	Chaperonin protein groES	8312	4.1	0.00E + 00	0.00E + 00	0	432	469	504	7935	7465	8131
CAF28204-1	hypothetical protein	10541.7	4	2.70E–237	0.00E + 00	0	587	639	651	9681	11588	8479
CAF28332-1	trwH1 (hypothetical protein)	235	3.9	3.30E–35	0.00E + 00	0	23	13	10	247	187	225
BH12490-1	tRNA-Ser	254.7	3.9	5.80E–34	0.00E + 00	0	21	10	17	249	294	173
BH01420-1	tRNA-Ser	298.7	3.6	6.70E–44	0.00E + 00	0	31	19	17	302	291	236
CAF28335-1	trwH2 (hypothetical protein)	92.7	3.2	9.90E–19	0.00E + 00	0	8	9	10	79	72	100
CAF28048-1	hypothetical protein	787	3.1	1.30E–95	0.00E + 00	0	102	71	75	639	740	734
BH01340-1	tRNA-Arg	186.7	3.1	2.20E–22	0.00E + 00	0	28	15	16	178	185	138
CAF27493-1	Anti-repressor protein	285.7	3.1	1.10E–20	0.00E + 00	0	49	29	13	240	257	269
BH09590-1	Prophage integrase	58.3	3.1	1.40E–14	0.00E + 00	0	6	7	5	49	53	55
BH09190-1	tRNA-Ser	74.7	2.9	1.20E–11	0.00E + 00	0	10	9	7	56	97	45
CAF27177-1	hypothetical prophage protein	973	2.8	1.90E-107	0.00E + 00	0	129	145	102	859	849	835
CAF27180-1	hypothetical prophage protein	65.3	2.7	3.20E–11	0.00E + 00	0	9	10	6	46	76	49
CAF28126-1	Chaperonin protein groEL	19433	2.5	3.50E–209	0.00E + 00	0	2918	2755	2879	16833	15536	17378
Top 20 downregulated genes in the intracellular bacteria							Normalized counts
ID	gene name	baseMean	log2FoldChange	PValue	PAdj	FDR	C1dedup	C2dedup	C3dedup	E1dedup	E2dedup	E3dedup
CAF27413-1	50S ribosomal protein l7 /l12 (rplL)	851.7	–3.5	4.40E–109	0.00E + 00	0	735	863	744	90	69	54
CAF28057-1	Glycine cleavage system protein h (gcvH)	616.3	–3.3	3.80E–98	0.00E + 00	0	558	586	539	69	51	46
CAF27262-1	Adenine DNA methyltransferase protein (ccrM)	1754.7	–3.1	9.20E–150	0.00E + 00	0	1560	1573	1592	222	140	177
CAF27005-1	nrdI protein (nrdI)	1685.7	–3	4.80E–149	0.00E + 00	0	1565	1543	1407	220	158	164
CAF27826-1	50S ribosomal protein l18 (rplR)	1145.7	–3	3.00E–145	0.00E + 00	0	977	1078	990	144	132	116
CAF27004-1	Glutaredoxin-like protein nrdH (nrdH)	599.3	–3	9.40E–77	0.00E + 00	0	469	568	565	64	66	66
CAF28170-1	hypothetical protein	327	–3	4.40E–43	0.00E + 00	0	309	250	314	34	40	34
CAF28362-1	SUN protein (FMU protein) (sun2)	6607	–2.9	1.70E–171	0.00E + 00	0	5592	5721	6121	918	703	766
CAF27412-1	50S ribosomal protein l10 (rplJ)	1646	–2.9	7.40E–164	0.00E + 00	0	1366	1489	1499	222	187	175
CAF27430-1	Phosphatidate cytidylyltransferase (cdsA1)	1166.3	–2.9	8.30E–140	0.00E + 00	0	990	1069	1040	157	117	126
CAF28294-1	ATP synthase epsilon chain (atpC)	512.7	–2.9	1.20E–68	0.00E + 00	0	476	461	419	56	64	62
CAF27381-1	hypothetical protein	1709.7	–2.8	2.00E–138	0.00E + 00	0	1511	1465	1515	258	180	200
CAF28342-1	Succinate dehydrogenase hydrophobic membrane anchor protein (sdhD)	1884.7	–2.8	4.50E–121	0.00E + 00	0	1577	1789	1598	285	204	201
CAF28173-1	hypothetical protein	844.7	–2.8	2.00E–110	0.00E + 00	0	746	738	726	111	114	99
CAF27819-1	30S ribosomal protein s11 (rpsK)	820.7	–2.8	7.20E–104	0.00E + 00	0	688	727	748	115	93	91
CAF28058-1	Glycine cleavage system protein t (gcvT)	1547.7	–2.7	7.50E–168	0.00E + 00	0	1338	1360	1311	225	201	208
CAF27688-1	NADH dehydrogenase I, F subunit (nuoF)	1974.3	–2.7	8.70E–158	0.00E + 00	0	1783	1641	1717	297	238	247
CAF27411-1	50S ribosomal protein l1 (rplA)	2318.3	–2.7	1.50E–149	0.00E + 00	0	1861	2120	2033	355	312	274
CAF28169-1	hypothetical protein	1802	–2.7	1.00E–126	0.00E + 00	0	1520	1606	1579	282	191	228
CAF28167-1	hypothetical protein	299.7	–2.6	8.80E–29	0.00E + 00	0	239	254	280	52	47	27

qRT-PCR analysis

Since the RNA-seq was performed for the first time on intracellular B. henselae bacteria, numerous steps in both the experimental and analytical processes required optimization. To validate the RNA-seq findings, we employed qRT-PCR on selected random genes. Initially, for cDNA synthesis, RNA samples were considered acceptable only if the CT value was ≥30 using the housekeeping (16S) primers. This ensures that any genomic DNA contamination is minimal. Subsequently, to normalize template DNA for both planktonic bacteria and intracellular bacteria, serial dilutions were performed with the same 16S primers, to produce a C_T value between 10 and 15, so that low-expressed genes could be detected. A total of 0.05 ng of cDNA for intracellular bacteria and 0.01 ng of cDNA for planktonic bacteria was used for each sample. The genes that were tested by qRT-PCR are highlighted in Table 3. The downregulated genes were chosen randomly, while the upregulated genes were selected based on previously characterized genes, including an uncharacterized hypothetical gene. There was increased groEL, groES, and trwH1 gene expression in intracellular bacteria compared to expression levels in planktonic bacteria (Fig. 5A and Table S11). Expression of rplJ, ccrM, atpC, and gcvH genes were decreased in intracellular bacteria compared to the level in extracellular B. henselae (Fig. 5A and Table S11). The fold change results of the 7 genes from both qRT-PCR and RNA-seq experiments are presented (Fig. 5B and Table S12), revealing a consistent pattern of expression across the two methodologies.

Fig. 5. Validation of RNA-sequencing data using qRT-PCR.

A qRT-PCR analysis was performed on 7 genes that were either up or downregulated in the intracellular bacteria. Fold change in the gene expression levels between planktonic and intracellular conditions is shown. Data is from three biological replicates with three technical replicates each. B Fold change in the gene expression levels of 7 genes from the RNA-seq data was plotted along with qRT-PCR data.

Functional enrichment analysis of differentially expressed genes

When the up and downregulated genes were uploaded to the STRING database (Table S13), it generated an interaction network that consisted of 79 nodes with 188 edges, with an enrichment p-value 7.36e-06. The network is presented in Fig. 6A. We also observed that two KEGG pathways, oxidative phosphorylation (bhs00190) and ribosome (bhs03010), were enriched with a false discovery rate of 0.0016 and 0.0161, respectively. We have included this subset of protein interactions in Fig. 6B, C. Additionally, STRING had predicted 6 clusters with a false discovery rate <0.05. A noteworthy cluster among these 6 clusters is CL:3089, where four proteins among six were annotated for entry of the bacteria into the host cell. These particular genes are activated in planktonic bacteria and are downregulated more than twofold when B. henselae is in the intracellular environment. While these gene products have not been characterized, their function was predicted by genomic context, computational predictions, and transferred evidence from other organisms. We plotted expression differences within a heatmap (Fig. 6D) and the normalized counts were included in Table 4. In the heatmap visualization, gene expression data is represented by Trimmed Mean of M values (TMM). The heatmap utilizes a 10-color scale, where each color corresponds to a specific range of TMM values evenly distributed across the entire dataset of expression values. Since oxidative phosphorylation and ribosome genes were downregulated in the intracellular environment, we hypothesize that the B. henselae might be growing more slowly compared to the planktonic/extracellular growth conditions. Additionally, B. henselae may be undergoing a metabolic shift to utilize host-derived nutrients more efficiently, leading to the reduction in overall protein synthesis through a change in energy production methods. Our hypothesis is that intracellular bacteria downregulate these genes as a strategy for long-term survival within the host cell, resulting in the entry of the bacteria into a dormant or persistent state where energy and resource utilization are minimized.

Fig. 6. String network of Up and down regulated genes that were significantly changed by ≥2-Fold.

A shows the predicted protein-protein interactions (PPIs) between all the up and down regulated genes that were changed by ≥ 2-Fold. Among these interactions, two KEGG pathways were predicted oxidative phosphorylation and Ribosome. A network of predicted proteins that are involved in the oxidative phosphorylation is presented as a sub network (B) and ribosome in (C). Figure 6D. Heatmap visualization of genes that were downregulated in the intracellular bacteria and organized by function. The trimmed mean of M values (TMM) is denoted by the color scale. The maximum value of the scale is set at 4000 (dark blue), and the minimum is set at 30 (brown). The distribution of the color scale of 10 colors is divided by the total number of data points per range. Downregulated groups in intracellular bacteria compared to planktonic bacteria include genes that are involved in entry into the host, oxidative phosphorylation, and ribosome. All genes shown are statistically significant. The heatmap was generated using Prism V9.4.1.

Table 4.

Normalized read counts of the downregulated genes that were predicted as a network by the STRING database

		Average normalized read counts (TMM)
	Gene name	in_vitro	intracellular
Entry into host	BH13960	764.33	146.33
	BH13970	1349.67	309.67
	BH13980	377.67	86.67
	BH14010	3885.67	768.33
Oxidative phosphorylation	cyoC	1555.00	382.33
	atpE	366.67	66.00
	nuoK	136.67	31.00
	nuoH	544.00	121.00
	nuoG	1421.00	294.67
	nuoF	1713.67	260.67
	nuoC	1003.33	220.33
	atpC	452.00	60.67
	sdhD	1654.67	230.00
	sdhC	1213.00	212.00
Ribosome	rpsA	3427.67	838.00
	rplK	968.67	195.00
	rplA	2004.67	313.67
	rplJ	1451.33	194.67
	rplL	780.67	71.00
	rplQ	443.33	77.67
	rpsK	721.00	99.67
	rplR	1015.00	130.67
	rpsG	1621.00	418.00
	rplS	477.67	110.00

Discussion

The goal of this study was to compare the transcriptome of B. henselae under intracellular versus extracellular conditions. Quantitative RT-PCR and microarray technologies have provided high-quality data regarding the gene expression of Bartonella under various conditions³⁷, and whole genome sequencing had also been used to study the expression of badA, an outer membrane protein in different strains of B. henselae⁵⁸. A previous study reported the gene expression profile of B. henselae during human endothelial cell infection using the microarray approach³⁸. In the study presented here, our objectives were to establish a protocol for isolating Bartonella residing within the intracellular environment, to conduct RNA-seq, and to and gain a more comprehensive understanding of the bacterial gene expression profile following invasion of DH82 cells. Using osmotic lysis, we were able to separate bacteria from the cells. We tried multiple methods to separate bacteria from cells, such as using mammalian cell lysis buffer (M-per), 0.1% saponin, and osmotic lysis to disrupt DH82 cells. We tried the filtration method⁵⁹, to isolate the bacteria from the host cell but overall bacterial counts were low compared to the osmotic lysis (data not shown). This is likely due to the presence of microcolonies⁶⁰ that were attached to the cells that might not have been filtered through the 5 μm pore-size filter. In our hands, osmotic lysis worked better than other methods, since we were able to separate more viable bacteria comparatively (data not shown). Remarkably, more than 90% of genes were significantly changed in the intracellular DH82 cell environment, suggesting the influence of the host environment in triggering the adaptive responses of bacteria to survive and replicate. Indeed, this is reflected in the notable energy conservations by downregulation of genes associated with Bartonella entry into the host cell.

After overcoming the hurdle of separating bacteria from the cell, we characterized the transcriptome of intracellular B. henselae using an RNA-seq approach. However, after sequencing the RNA from the intracellular Bartonella, more than 90% of reads were mapped to the host. It was shown previously that to reliably identify the differentially expressed genes of bacteria by RNA-seq, 2–3 million reads per sample is considered an optimal sequencing depth⁶¹. Since we were able to obtain 3–4 million reads per sample from the intracellular bacteria after the removal of host RNA reads, our experiment generated sufficient reads to describe a reliable transcriptome of the intracellular Bartonella. Previous studies that aimed to separate the bacteria from the host have characterized the transcriptome of bacteria with just 1 million of overall bacterial reads^59,62. Taking these numbers into consideration, we could possibly aim to characterize gene expression of B. henselae from smaller populations, which help preserve the bacterial phenotype of intracellular environment. If the challenge of low bacterial read continues to pose an issue in the intracellular system, opting for a dual RNA-seq analysis of both the bacteria and host cell could be an alternative. To get enough reads from the bacteria, the depth coverage has to be around 2-billion reads⁶³.

The type IV secretion system (T4SS) VirB/D4 genes involved in the translocation of Bartonella effector proteins (Beps) into the cytoplasm of the infected endothelial cells^64,65, were slightly downregulated in the intracellular bacteria. However, this fold change was not statistically significant. Similarly, BatR that controls the expression of virulence genes during infection³⁸ was slightly downregulated (–0.1-fold) with no significant difference in the expression. The lack of statistical significance suggests that the observed change may not be reliable and could be attributed to the noise in the experimental system. Furthermore, the effector proteins BepB, C, and D were downregulated but had no statistical difference in the expression when compared to the extracellular growth environment, while bepA and E were significantly downregulated by –1 and –0.4-fold, respectively. It is well-established that each Bep protein is involved in modulation of the host cell^66–68. During the course of infection, BepD and BepE translocation into the host cell appears to result in the downregulation of the host’s innate immune response. Concurrently, BepG, BepC together with BepF contribute to the induction of internalization of the bacterial aggregates. BepA’s ability to inhibit cell apoptosis indicates a broader impact on host cellular processes^66,69. In a host environment, the downregulation of Beps in the intracellular bacteria could be a factor that contributes to persistence. Similarly, BadA (BH01490), which is involved in the adhesion of Bartonella to the host cells^22,70 was significantly downregulated. This is in accordance with the previous study that essential virulence factors BadA and VirB/D4 are likely to exhibit differentially expression patterns throughout the infection cycle of Bartonella⁶⁴. This intricate control of gene expression may ensure Bartonella’s adaptability to vector and mammalian host environments, facilitating optimal interaction with the host cell throughout different stages of transmission and infection.

As facultative intracellular bacteria, Bartonella spp. thrive inside host cells, where they can access essential substrates for growth and replication. By downregulating oxidative phosphorylation, these bacteria minimize their reliance on aerobic respiration, which is particularly advantageous in the often-hypoxic conditions found within host tissues⁷¹. This metabolic shift facilitates the utilization of host-derived nutrients, such as amino acids and glucose, enhancing the bacteria’s ability to survive and replicate. For instance, Bartonella species have been shown to sequester heme from the host, which is critical for their metabolic needs and contributes to their virulence⁷². The ability to adapt their metabolism not only aids nutrient acquisition but also helps Bartonella evade host immune responses by reducing the production of reactive oxygen species (ROS) associated with oxidative phosphorylation, thus minimizing detection by the host immune system⁷³.

Furthermore, Bartonella employs sophisticated mechanisms such as Type IV secretion systems (T4SS) to translocate effector proteins into host cells. These effector proteins modulate host cellular functions, promoting bacterial survival and facilitating nutrient acquisition while simultaneously dampening immune responses⁷⁴. The interplay between Bartonella metabolism and host cell responses highlights a complex relationship where both parties adapt to each other’s strategies.

Overall, the downregulation of oxidative phosphorylation genes in Bartonella is a clear indication of its evolutionary adaptation to exploit host resources efficiently while surviving within the host environment. This insight underscores the importance of targeting metabolic pathways in developing therapeutic strategies against Bartonella infections.

Previously, it was demonstrated that the alternative sigma factor RpoH1 is essential for the expression of the VirB/D4 T4SS and its secreted effector proteins⁵⁴. The activation is dependent on RpoH1 that requires the presence of an active BatR/BatS two-component system. RpoH1 levels are in turn modulated by the components of the stringent response (SR), namely DksA and SpoT. BatR expression is controlled by both RpoH1 and the components of SR⁵⁴. Given the significant downregulation of rpoH1 (–0.7), dksA (–1.1) and spoT (–0.6) expression levels in the intracellular bacteria, we hypothesize that this downregulation could cause the observed decrease in the expression levels of most virulence-associated genes. The inhibition of these virulence genes could be a possible strategy of the Bartonella to combat the host ROS, that are released by the neutrophils and macrophages⁷⁵.

It is well known that the SR is a regulatory system in bacteria that is triggered by nutrient limitation, amino acid starvation, or other stress conditions. The SR response is mediated by the release of guanosine tetraphosphate (ppGpp) or guanosine pentaphosphate (pppGpp)^76,77. This nucleotide signals bacteria to re-allocate cellular resources by inhibiting DNA synthesis, RNA stability, and ribosomal protein synthesis. The (p)ppGpp molecule also promotes upregulation of key regulatory factors for stress resistance, glycolysis, and amino acid synthesis^77,78. SpoT is considered as bifunctional synthase/hydrolase that controls the level of the (p)ppGpp^79,80 in alphaproteobacteria. When Caulobacter crescentus, an alphaproteobacteria, was analyzed for the adaptive response, the initiation of DNA replication was inhibited by the increasing levels of (p)ppGpp⁷⁹. Control of the DNA replication is maintained by the two essential regulators of the cell: CtrA and DnaA^81,82, where inhibition of DnaA and activation of CtrA inhibits DNA synthesis⁷⁹. In our study, CtrA was upregulated (0.5) and DnaA was downregulated (–1.3) significantly, which suggests that the levels of (p)ppGpp might be altered and consequently induced the downregulation of genes encoding ribosomal proteins in the intracellular bacteria. However, the levels of (p)ppGpp in the intracellular bacteria need to be evaluated by performing SpoT mutation experiments. In response to stress, Bartonella appears to have increased the synthesis of tRNA as a mechanism to maintain the essential cellular functions, even though overall protein synthesis is reduced. The upregulation of genes encoding tRNA supports the translation process, ensuring the efficacy of the remaining protein synthesis, thus allowing the bacteria to prioritize the synthesis of essential proteins for survival. Subsequently, it can be hypothesized that the increase in the tRNA regulation signals that the bacteria might be preparing for the upregulation of the pathogenic genes, increasing the ability to productively infect the host. Experiments utilizing longer durations of intracellular invasion by the bacteria could offer more insight into the roles of tRNA, SpoT, and (p)ppGpp. It is worth noting that ribosomal synthesis is an energy-intensive process, consuming up to 40% of the cell’s energy⁸³, which possibly explains the downregulation of ribosomal gene expression in intracellular bacteria.

The upregulation of phage-associated genes in intracellular bacteria compared to planktonic bacteria suggests increased phage activity within host cells. Previous studies have shown the presence of bacteriophage-like particles in B. henselae^84,85, hinting a potential interaction between phages and the bacteria. Additionally, a decrease in B. henselae growth rate, observed after repeated subcultures has been linked to defective bacteriophages^86,87. During the log-phase growth, Bartonella exhibits a low number of phage particles⁸⁶, resembling the ratio seen during lysogenic phage reversion⁸⁸. This suggests lysogenic phages may be present, impacting bacterial growth. The increased expression of phage genes in the intracellular environment suggest that the phage genes might be entering the lytic phase, with implications for disease progression. However, further investigation is needed to understand the mechanisms behind this phenomenon, which could provide insights into phage-bacteria interactions and potential therapeutic strategies.

In this study, we infected DH82 cells with B. henselae to model the host intracellular environment, enabling characterization of the infection transcriptome of the bacteria. By using a cell line, Bartonella was not subjected to the host-specific environment that consists of functional macrophages, mature lymphocytes, signaling molecules, complement activation, and host microbiota. The lack of these cells is a limitation of the study because a true host environment can modify the transcriptome of B. henselae during infection. As such, the bacteria do not encounter immune effector mechanisms to eliminate the infection, thereby potentially influencing the activation and repression of specific virulence factors⁵⁹. Moving forward, we aim to conduct ex vivo and in vivo RNA-seq in susceptible mammalian hosts once we address various technical challenges of working with these intracellular bacteria. Another limitation that must be considered is the bacterial burden. A previous study reported the influence of the multiplicity of infection rates on the overall transcriptional signal⁶³. In a human host, it is highly unlikely that Bartonella grows to high density⁸⁹. Given the potential impact of cell density on overall gene expression, our future studies will focus on analyzing smaller populations (lower MOIs) of B. henselae to infect the host.

Conclusions

Comparison of the transcriptome of B. henselae between extracellular and intracellular environments offers a window into the ability of this genus of bacteria to host-adapt and survive inside host cells. We previously reported reduced antibiotic susceptibility of Bartonella living within monocytic cells⁵². Subsequent studies should aim at interrupting survival of Bartonella in all of the microenvironments in which they exist, so it is essential to know the corresponding phenotypes of the bacteria within different cells and tissues. This study will provide a key resource to the Bartonella research field through the availability of comprehensive transcriptomic data on which to support future research.

Methods

Infection and isolation of B. henselae from DH82 cells

The quantity of 2 × 10⁵ DH82 cells were seeded into individual wells of 6-well plate and allowed to adhere to the plate overnight. The next day, 5–7-day old B. henselae cultures grown on blood agar plates were used for infecting the DH82 cells. The B. henselae strain used in this study is San Antonio 2, 267HO04, a human clinical isolate. Initially, bacteria were resuspended in the EMEM + 15% FBS media and dilutions were made. Individual wells of DH82 cells were infected with an MOI of 1:100 to ensure that 100% of the host cells were infected. Six-well plates were then centrifuged at 1200 × g for 10 min. at room temperature (RT). Plates were then incubated at 37 °C + 5%CO₂ for 48 h. After 48 h, cells were washed three times with 1X PBS containing calcium and magnesium for 5 min each to remove extracellular bacteria. After the wash, cells were exposed to 500 μl of ice-cold sterile filtered distilled water for 5 min to lyse the cells. A volume of 100 μl of the lysate was plated on blood agar plates for bacterial counting.

The following is the experimental evidence and rationale for the selected isolation procedure. Using the filtration method, after passing the lysate through 25 microns and subsequently through 5 microns, the CFU count was 6 (±5.34) × 10³, whereas passing the lysate through only 5-micron filter yielded a CFU of 8(±2.86) × 10³. However, using the osmotic lysis method, yielded a CFU of 5.3(±1.6) × 10⁵ of the intracellular bacteria and a CFU of 5.07(±1.73) × 10⁶ for the extracellular bacteria or the cell-exposed bacteria.

When gentamicin (100 µg/ml) antibiotic was used to kill the extracellular bacteria, the assay yielded a CFU count of 1.26 (±0.16) × 10⁵. CFU counts are included in Table S14. The extracellular bacteria were incubated with the antibiotic for 1 h before cells were lysed and plated.

While the gentamicin protection assay was used to kill the extracellular bacteria, it has yielded almost the same number of intracellular bacteria as the osmotic lysis. However, this assay relies on gentamicin to eliminate extracellular bacteria and membrane-bound bacteria, with the assumption that gentamicin cannot penetrate the eukaryotic cells. However, there are previous studies that suspected that higher concentrations of gentamicin for long incubation times possibly kills the intracellular bacteria presumably by internalized gentamicin through pinocytosis^90–94. A study has also reported that aminoglycosides slowly penetrate eukaryotic cells and can even reach intracellular concentrations⁹⁵. This suggests that some gentamicin might enter cells even within 1 h, though likely not bactericidal levels. Even if the intracellular concentration doesn’t reach high enough to kill bacteria, it might be sufficient to cause stress or alter gene expression of the intracellular bacteria. Since a most recent study⁹⁶ that assessed the bacterial effects of three amino acids in combination with gentamicin, upregulated the metabolites, increased the amino acid abundance, and significantly activated metabolisms. This indicates that even sublethal antibiotic exposure can cause metabolic changes in bacteria. A study has quantitatively shown the internalization of gentamicin and the adverse effects on measuring the invasion potential and enumeration of surviving intracellular bacteria⁹⁷.

DH82 cell surface staining

To stain the cell membrane of the DH82 cells, MemBrite Fix staining (Biotium, 30093) was used. After removing culture media, cells were washed with 1X PBS buffer and incubated with 1X pre-staining solution for 5 min at 37 °C. After 5 min, pre-staining solution was removed, and 1X MemBrite Fix dye solution was added to cover the cells. Cells were incubated at 4 °C for 30 min. with pre-chilled staining solution to prevent dye internalization. Cells were then rinsed with 1X PBS and were fixed with 4% paraformaldehyde (PFA) in PBS for 20 min at RT.

Immunofluorescence staining

To stain intracellular bacteria, cells were permeabilized with PBS containing 0.1% TritonX-100 for 10 min at RT. Cells were then stained against B. henselae following the previously published protocol⁵². To stain surface-adhered bacteria, the permeabilization step was omitted and followed by staining against B. henselae.

Imaging was performed using a Nikon Ti2-E motorized fluorescence microscope. Infected and uninfected cells with dye were imaged during the same session with identical acquisition parameters. Fluorescence intensity was optimized on uninfected cells to eliminate the autofluorescence from the cells, and remained constant for all the infected and uninfected cells with the dye.

For differential staining of intracellular and extracellular bacteria, after 48 h, infected DH82 cells were washed and fixed with 4% paraformaldehyde and subsequently incubated with blocking buffer (10% Normal Goat Serum-Gibco, 16-210-072, in PBS) for 1 h. to prevent non-specific binding. After blocking, cells were incubated with the primary antibody: rabbit polyclonal anti–B. henselae (serum derived from a hyperimmunized rabbit eight weeks after inoculation with in vitro propagated B. henselae) at a dilution of 1:300 for 1 h followed by three washes using wash buffer (1X PBS, 0.2% fish skin gelatin). Cells were then incubated with the secondary antibody Goat Anti-rabbit IgG-Alexa Fluor 594 (Thermo Fisher Scientific, R37117) diluted 1:1000 in blocking buffer. After secondary antibody incubation, cells were washed with wash buffer and were permeabilized using 0.1% Triton X-100 in 1X PBS for 20 min. Cells were then incubated with blocking buffer for 1 h and incubated with the primary antibody: rabbit polyclonal anti–B. henselae and counterstained with secondary antibody Goat Anti-rabbit IgG-Alexa Fluor 488 (Thermo Fisher Scientific, A11034) diluted 1:1000 in blocking buffer. Slides were washed and cells were counterstained with DAPI for 10 min to stain nuclei (EMD millipore, 2160), then mounted with the anti-quenching solution (Thermofisher, P36934) and coverslipped.

Imaging was performed using a Nikon Ti2-E motorized fluorescence microscope using a 40X objective and three fluorescence channels (DAPI, FITC, and TRITC) were used to capture an image field. A total of at least 50 images were taken from each individual experiment and yellow dots (extracellular bacteria) were counted manually from each field. The counts from the individual fields were shown (Table S1) and the representative images are shown in (Supplementary Fig. 3)

This experiment was repeated by staining with secondary antibody Goat Anti-rabbit IgG-Alexa Fluor 488 first and then with Goat Anti-rabbit IgG-Alexa Fluor 594. Switching the order of secondary antibodies provided a way to validate specificity and enhance the reliability of our immunofluorescence results. This experiment was repeated two times with secondary antibody combinations. The average percentage of extracellular bacteria from four independent experiments was 12.795%.

RNAscope (RNA in-situ Hybridization)

RNAscope is commercially available from Advanced Cell Diagnostics, Inc. (ACDBio). RNAscope was performed according to the manufacturer’s guidelines from RNAscope Multiplex Fluorescent Detection Kit version 2 (MK-50 010 and 323100).

DH82 cells that were infected with B. henselae for 48 h were fixed with 4% PFA at RT for 30 min. After removing PFA, cells were gently rinsed twice with 1X PBS and ~5 drops of RNAscope Hydrogen Peroxide was added to cover the cells followed by incubation at RT for 10 min. Slides were then washed with distilled water. After washing slides with distilled water, excess liquid was removed by tapping the slide, and 2–4 drops of 1:15 diluted protease III was added, then incubated for 10 min at RT in a humidity control tray. Slides were washed with 1X PBS and probed for B. henselae using 23S rRNA (ACDBio, 23S rRNA transcript, 485621-C3). A positive control probe was used to target the PPIB gene (ACDBio, 437441) of DH82 cells. As a negative control, a probe targeting the dapB gene (ACDBio, 320871) of B. subtilis was used. After addition of probes, slides were incubated in the EZ hybridization oven (ACDBio, 321710) using a humidity control tray (ACDBio, 310012) and slide rack (ACDBio, 310017) for 2 h. After probe hybridization, amplification and detection steps were performed. Amplifier 1 was initially incubated on slides for 30 min at 40 °C and washed with wash buffer three times each for 2 min. Slides were then incubated with Amplifier 2 for 30 min as before, washed, and incubated with a final layer of amplifier, Amplifier 3, and incubated for 15 min at 40 °C and washed as before. For the development of the signal, HRP (horseradish peroxidase)-C1 was added to the slide to target the probes (PPIB and dapB) assigned to channel C1 and HRP-C3 was added to target the probe (23S rRNA against B. henselae) and incubated in the hybridization oven for 15 min at 40 °C. Finally, slides were incubated with fluorescent molecules from PerkinElmer [Tyramide Signal Amplification (TSA) plus Cyanine 3 system: NEL744001KT or TSA plus Cyanine 5 system: NEL745001KT]. These fluorescent molecules bind to the cascade of signal amplification molecules resulting in signal detection. Imaging was performed using a Nikon Ti2-E motorized fluorescence microscope as mentioned above.

RNA extraction and purification from both lab cultured bacteria and host cell adapted bacteria

Intracellular bacteria. Immediately after osmotic lysis, 1 ml of bacterial RNA protect reagent (Qiagen, 76104) was added and incubated for 15 min at RT. After 15 min, the resulting lysate was pelleted at high speed 14,000 rpm for 15 min. Supernatant was removed and the pellet was resuspended in 1X PBS and pelleted again at high speed. The resulting supernatant was removed. To the pellet containing bacteria, a mixture of lysozyme (15 mg/ml) containing proteinase K prepared in 1X TE buffer was added and incubated for 15 min by vortexing for every 2–3 min. After incubation with lysozyme, RLT buffer from Qiagen was added for complete lysis of bacteria. The lysed bacteria were purified by following manufacturer’s protocol of Qiagen RNA mini kit. Bacteria cultured on blood agar plates. Bacteria were suspended in EMEM media +15% FBS and pelleted. After pelleting, media was removed, and bacteria were treated as above and RNA was isolated.

Enrichment of bacterial RNA

Using MICROBEnrich kit

To eliminate most of the host ribosomal RNA and mRNA from the RNA purified from the host-adapted bacteria, the MICROBEnrich™ kit (Thermofisher, AM1901) was adopted and the manufacturer’s protocol was followed. Briefly, a total of 8.5 µg of the host-adapted bacteria total RNA mixture was used per sample. To the RNA mixture, binding buffer and capture oligo mix was added, which captures and removes 18S rRNA, 28S rRNA, and polyadenylated mRNAs of host RNA. This mixture was then heat denatured at 70 °C. Later, the mixture was incubated at 37 °C, which allows the capture oligonucleotides to hybridize to homologous regions of the 18S and 28S rRNAs. Next, Oligo MagBeads were employed to selectively remove the capture oligonucleotides from the reaction mixture. The resulting supernatant after Oligo MagBeads removal contains the enriched bacterial RNA. After enrichment, equal amounts of RNA from pre- and post-enrichment steps were run on a 2100 Bioanalyzer (Agilent Technologies) using Agilent RNA 6000 Nano chip. The MICROBEnrich™ kit does not remove 5S rRNA of the host.

Ribosomal depletion of the enriched bacterial total RNA

To remove the host and bacterial rRNA, RiboCop rRNA Depletion kits for Human/Mouse/Rat (Lexogen, 144) and for bacteria (Lexogen, 126) were simultaneously used. Manufacturer’s user guides 144UG288 and 125UG246 were followed. Briefly, a total of 500 ng of total RNA from the bacteria control samples and 1000 ng of bacterial-enriched RNA from the mixture sample was utilized. A combined probe mix from the above-mentioned kits was added to the RNA and samples were denatured at 75 °C for 5 min. In this step, only control samples were denatured because in the MICROBEnrich step, enriched RNA samples were already denatured. From the next step, all the samples, both enriched and control bacterial RNA were treated according to the protocol. After denaturation, all the samples were subjected to a temperature of 65 °C for 30 min on a thermomixer with agitation at 1250 rpm. This step helps to hybridize the probes to the target sequences of the rRNA. Later, depletion beads were used to remove probes that were hybridized to rRNA from the solution. The final purified product that is supposed to be free from rRNAs was used for the Next-Generation Sequencing (NGS) library preparation.

Total RNA-sequencing Library preparation

RNA libraries were prepared following Lexogen’s user manual of CORALL Total RNA-Seq Library Prep Kit (Lexogen, 095). This library preparation kit is fragmentation-free, which provides complete transcript coverage including start and end sites. It seamlessly integrates Unique Molecular Identifiers (UMIs) while maintaining protocol-inherent strand specificity (>99%).

Briefly, rRNA-depleted RNA was used as the starting material. The library generation started by random hybridization of displacement stop primers (DSP) to the RNA template. These primers contain partial Illumina-compatible sequences at the 5′ ends. Reverse transcription extends each DSP to the next DSP where transcription is effectively stopped. This random priming of the DSP to DSP determines the insert size of the library. In the next step, a highly efficient ligation of linker oligos to the 3′ ends of first strand cDNA fragments introduces UMIs and partial Illumina-compatible P5 sequences. This primary library is purified using magnetic beads to remove ligation reaction components. Since the mRNA content affects the number of PCR cycles needed for the final library amplification, a qPCR assay was utilized to optimize the number of cycles that are required for each sample for endpoint PCR. This step normalizes the total amount of input across all the samples. This step will also prevent both under- and over-cycling, the latter of which may bias sequencing results. After determining the number of cycles required for the endpoint PCR, the library was amplified to add the complete adaptor sequences required for cluster generation and to produce sufficient material for quality control and sequencing. The i7 indices were added during this step in order to uniquely multiplex the samples for the sequencing run. The resulting libraries were then purified using magnetic beads to get rid of PCR components, which could interfere with quantification and sequencing on an Illumina NGS instrument. Finally, the library samples were quality controlled (QC’d) on a bioanalyzer and quantified. The libraries were sequenced on an Illumina NextSeq 500 using the 75 bp single-end protocol at the Genomics core, University of Birmingham, Alabama. A total of 55–70 million single end reads were generated per sample.

Bioinformatic analysis

For the initial analysis of the RNA-seq data, Lexogen’s BlueBee pipeline was used. Briefly, UMI sequences were trimmed and added to the read identifiers. Next, cutadapt was used to remove artifacts and trim adapter sequences from the sequencing reads. These trimmed reads were quality checked using FastQC. Reads were then aligned to the Bacterial (B. henselae) genome using STAR aligner. UMI tools were used to remove reads with identical mapping coordinates and UMI sequences were collapsed to remove PCR duplicates in addition to only keeping mapped reads in Binary Alignment Map (BAM) files. Mix² then was used to compile transcript and gene abundance estimates. The read counts were then normalized and differentially expressed genes were estimated using DeSeq2.

For down sampling of the raw reads, the Seqtk tool⁹⁸ (https://github.com/lh3/seqtk) with the command “sample” was used. Using Seqtk, we performed two rounds (3 million and 4.5 million) of downsampling. The downsampled reads were then quality checked using FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Since the sequences contains UMI sequences, we performed UMI extraction using UMI-tools⁹⁹. The extracted reads were processed using the tool Cutadapt¹⁰⁰ to find and remove adapter sequences. Along with the adapter trimming, Cutadapt was also used to remove low quality bases below the cutoff value of 30. These trimmed reads were then aligned to the reference genome (https://ftp.ensemblgenomes.ebi.ac.uk/pub/bacteria/release-57/fasta/bacteria_0_collection/bartonella_henselae_str_houston_1_gca_000046705/) using BWA-MEM¹⁰¹. UMI tools were used to remove PCR duplicates as mentioned above. Read counts for each gene were determined using featureCounts¹⁰². We used B. henselae gff3 file from ensembl (https://ftp.ensemblgenomes.ebi.ac.uk/pub/bacteria/release-57/gff3/bacteria_0_collection/bartonella_henselae_str_houston_1_gca_000046705/). Differential expression analysis and visualization was performed using the R package edgeR¹⁰³. An overview of the analysis is presented as Fig. 3. Genes were considered to be significantly differentially expressed when the relative expression had a fold change (FC) ≥2.0 and a P value of ≤0.05, unless otherwise mentioned. The computer setup for the analysis of the RNA-seq and scripts for the edgeR were from “RNA-Seq by Example” volume of the Biostar Handbook ISBN: 978-0-578-80435-4(https://www.biostarhandbook.com/books/rnaseq/index.html). To generate a PCA plot, the ClustVis webtool¹⁰⁴ was used.

STRING database analysis

Enrichment of the genes that were up or downregulated by twofold or more was performed using STRING¹⁰⁵. The genes were uploaded onto the database and the interactions between the genes were predicted. The analysis pipeline was set to have a cut-off interaction confidence score ≥0.4 and the predicted cluster would have cutoff of p < 0.05 corrected for multiple testing within each category using the Benjamin–Hochberg procedure.

qRT-PCR

Validation of the expression profiles of selected genes was performed by real-time quantitative RT-PCR (qRT-PCR) on total RNA samples prepared from both intracellular and in vitro grown/extracellular bacteria. First-strand cDNA was synthesized using the High-Capacity RNA-to-cDNA kit (Thermofisher#4387406) from 500 ng of total RNA as per manufacturer’s instructions. Each qPCR reaction was performed in a total volume of 10 µL with 2X PowerUp SYBR Green I Master Mix (Thermofisher#A25742). A non-template control and a reverse transcriptase negative reaction was included with every round of qPCR. A total of three technical replicate reactions were run per gene target per sample. Reactions were run using the Applied Biosystems QuantStudio 6 Flex Real-Time PCR instrument under fast cycling mode for 40 cycles. Primers were designed using Primer3web V4.1.0^106–108. To check for specificity, melt curve analysis as well as subsequent agarose gel electrophoresis was performed on each normalized to the 16S reference gene using the 2^-(∆∆Ct) method¹⁰⁹, and sequence-specific primers used for qPCR are listed in (Table S15). The average of 2^-(∆∆Ct) values from three biological replicate experiments was calculated and the graph was plotted. Standard error of mean was calculated based on the variability of the 2^-(∆∆Ct) value of three biological replicates.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Abbreviations

CSD

cat scratch disease

ECM

extracellular matrix

RNA

ribonucleic acid

RNAseq

RNA sequencing

PCR

polymerase chain reaction

qPCR

quantitative PCR

rRNA

ribosomal RNA

T4SS

Type IV secretion systems

DAPI

4’,6-diamidino-2-phenylindole

PCA

principal component analysis

TMM

trimmed mean of M values

EMEM

Eagle’s minimal essential media

MOI

multiplicity of infection

PBS

phosphate buffered saline

UMIs

unique molecular identifiers

Author contributions

S.K.G.G. and M.E.E. conceptualized the project; S.K.G.G., M.E.E., and J.R.C. established the methodology; S.K.G.G. and JRC performed validation; S.K.G.G., J.R.C., R.G.M., and M.E.E. formally analyzed the data; S.K.G.G. and R.G.M. engaged in the investigation; M.E.E. and E.B.B. provided the resources; S.K.G.G. provided data curation and original draft preparation; All authors contributed to review and editing; M.E.E. oversaw project administration; M.E.E. and E.B.B. were responsible for funding acquisition. All authors read and approved the final manuscript.

Peer review

Peer review information

Communications Biology thanks Burt Anderson and Karthika Rajeeve [name] for their contribution to the peer review of this work. Primary Handling Editor: Tobias Goris.

Funding

This research was funded by the Steven and Alexandra Cohen Foundation and NIH base grant P51 OD011104-61; RRID: SCR_008167. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

The datasets generated and/or analysed during the current study are available in the NCBI repository (https://submit.ncbi.nlm.nih.gov/subs/sra/SUB14175801/overview) with locator information file accession numbers BioSample accessions: SAMN40140122, SAMN40140123, SAMN40140124, SAMN40140125, SAMN40140126, SAMN40140127, SAMN40140128, SAMN40140129, SAMN40140130. Object IDs and corresponding URLs: 40140122, 40140123, 40140124, 40140125, 40140126, 40140127, 40140128, 40140129, 40140130.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-025-07535-9.