Hill D.P.,The Jackson Laboratory |
Adams N.,University of Cambridge |
Adams N.,European Bioinformatics Institute |
Adams N.,CSIRO |
And 24 more authors.
BMC Genomics | Year: 2013
Background: The Gene Ontology (GO) facilitates the description of the action of gene products in a biological context. Many GO terms refer to chemical entities that participate in biological processes. To facilitate accurate and consistent systems-wide biological representation, it is necessary to integrate the chemical view of these entities with the biological view of GO functions and processes. We describe a collaborative effort between the GO and the Chemical Entities of Biological Interest (ChEBI) ontology developers to ensure that the representation of chemicals in the GO is both internally consistent and in alignment with the chemical expertise captured in ChEBI.Results: We have examined and integrated the ChEBI structural hierarchy into the GO resource through computationally-assisted manual curation of both GO and ChEBI. Our work has resulted in the creation of computable definitions of GO terms that contain fully defined semantic relationships to corresponding chemical terms in ChEBI.Conclusions: The set of logical definitions using both the GO and ChEBI has already been used to automate aspects of GO development and has the potential to allow the integration of data across the domains of biology and chemistry. These logical definitions are available as an extended version of the ontology from http://purl.obolibrary.org/obo/go/extensions/go-plus.owl. © 2013 Hill et al.; licensee BioMed Central Ltd.
No statistical methods were used to predetermine sample size. The investigators were not blinded to allocation during experiments and outcome assessment. Salmonella enterica serovar Typhimurium strain SL1344 constitutively expressing GFP from a chromosomal locus (strain JVS-3858) was previously described51 and is referred to as wild type throughout this study. The complete list of bacterial strains used in this study is provided in Supplementary Table 1. Routinely, bacteria were grown in Lennox broth (LB) medium at 37 °C with shaking at 220 r.p.m. When appropriate, 100 μg ml−1 ampicillin (Amp), 50 μg ml−1 kanamycin (Kan), or 20 μg ml−1 chloramphenicol (Cm) (final concentrations) were added to the liquid medium or agar plates. Chromosomal mutagenesis of Salmonella SL1344 was performed as previously described52. To construct a non-polar pinT mutant strain (YCS-034, GFP−; or JVS-10038, GFP+), the first ~60 nt of the gene were removed and replaced by a resistance cassette, while keeping the Rho-independent terminator intact. Then, the resistance cassette was eliminated using the FLP helper plasmid pCP20 at 42 °C52. All mutations were transduced into the wild-type background using P22 phage53. For plasmid transformation the respective Salmonella strains were electroporated with ~10 ng of DNA. The following cell lines were used in this study: human cervix carcinoma cells (HeLa-S3; ATCC CCL-2.2), human epithelial colorectal adenocarcinoma cells (CaCo-2; ATCC HTB-37), human epithelial colorectal adenocarcinoma cells (HT29; DSMZ No. ACC-299), human stomach adenocarcinoma cells (AGS; ATCC CRL-1739), human epithelial colon metastatic cells (LoVo; ATCC CCL-229), human embryonic kidney 293 cells (HEK293; ATCC CRL-1573), human monocytic cells (THP-1; ATCC TIB-202), murine fibroblast cells (L929; ATCC CCL-1), murine embryonic fibroblast cells (MEF; ATCC SCRC-1040), mouse leukaemic monocyte/macrophage cells (RAW264.7; ATCC TIB-71), porcine intestinal epithelial cells (IPEC-J2)54, porcine macrophage-like cells (3D4/31)55. HeLa-S3, CaCo-2, THP-1, HEK293; RAW264.7 and MEF cells were obtained from the group of Thomas Rudel (Biocentre, Würzburg). AGS cells were provided by Cynthia Sharma (Research Center for Infectious Diseases, Würzburg). L929 cells were obtained from Thomas Meyer (Max Planck Institute for Infection Biology, Berlin). HT29, LoVo, IPEC-J2 and 3D4/31 cells were provided by Karsten Tedin (Centre for Infection Medicine, Berlin). Cell lines have not been authenticated in our laboratory, but were routinely tested for mycoplasma contamination (MycoAlert Mycoplasma Detection Kit, Lonza). HeLa-S3 cells were cultured according the guidelines provided by the ENCODE consortium (http://genome.ucsc.edu/encode/protocols/cell/human/Stam_15_protocols.pdf). Briefly, cells were grown in DMEM (Gibco) supplemented with 10% fetal calf serum (FCS; Biochrom), 2 mM l-glutamine (Gibco) and 1 mM sodium pyruvate (Gibco) in T-75 flasks (Corning) in a 5% CO , humidified atmosphere, at 37 °C. Further cell lines used in this study (THP-1, CaCo-2, AGS, HT29, LoVo, HEK293, MEF, L929, RAW264.7, IPEC-J2 and 3D4/31) were cultured in RPMI (Gibco) supplemented with 10% FCS, 2 mM l-glutamine, 1 mM sodium pyruvate and 0.5% β-mercaptoethanol (Gibco) in a 5% CO , humidified atmosphere, at 37 °C. To differentiate THP-1 monocytes, seeded cells (1 × 106 cells per well; six-well format) were treated with 50 ng ml−1 (final concentration) of phorbol 12-myristate 13-acetate (PMA) (Sigma) for 72 h (after 48 h fresh PMA at the same concentration was added to the culture). For the differentiation of murine bone marrow derived macrophages (BMDMs), the marrow of femur and tibia was isolated from 8–12-week-old female C57BL/6 wild-type mice and stored in RPMI supplemented with 10% FCS. The cell suspension was centrifuged for 5 min at 250g and the leukocyte pellet was resuspended in differentiation medium consisting of X-vivo-15 medium (Lonza) supplemented with 10% FCS and 10% L929-conditioned DMEM medium (same composition as above). Cells were cultured at 3 × 106 cells per 10 ml in a T-75 flask. At day 3, another 3 ml of differentiation medium were added and cells were further cultured until day 5. Successful macrophage differentiation was validated by microscopy before the cells were detached using a rubber scraper (Sarstedt) and seeded into six-well plates at 105 cells per well in fresh differentiation medium. Infection was carried out on day 7 as described below. In vitro infection of HeLa-S3 cells was carried out following a previously published protocol56 with slight modifications. Two days before infection 2 × 105 HeLa-S3 cells were seeded in 2 ml complete DMEM (six-well format). Overnight cultures of Salmonella were diluted 1:100 in fresh LB medium and grown aerobically to an OD of 2.0. Bacterial cells were harvested by centrifugation (2 min at 12,000 r.p.m., room temperature) and resuspended in DMEM. Infection of HeLa-S3 cells was carried out by adding the bacterial suspension directly to each well. If not mentioned otherwise, infections were performed at a multiplicity of infection (m.o.i.) of 5. Immediately after addition of bacteria, the plates were centrifuged for 10 min at 250g at room temperature followed by 30 min incubation in 5% CO , humidified atmosphere, at 37 °C. Medium was then replaced for gentamicin-containing DMEM (final concentration: 50 μg ml−1) to kill extracellular bacteria. After a further 30 min incubation step, medium was again replaced by fresh DMEM containing 10 μg ml−1 of gentamicin, and incubated for the remainder of the experiment. Time point 0 was defined as the time when gentamicin was first added to the cells. Further cell types were infected as described for Hela-S3 cells except that infection was carried out in RPMI medium and that infection was with an m.o.i. of 10 (THP-1, CaCo-2, HT29, AGS, HEK293, MEF, L929 and RAW264.7) or 20 (IPEC-J2, 3D4/31), respectively. Infection of BMDMs was carried out with an m.o.i. of 10 and using X-vivo-15 medium (10% fetal calf serum, 10% L929-conditioned medium). Infection was carried out as described above, except that HeLa-S3 cells had been seeded onto coverslips (24-well format). At the respective timepoint, coverslips with infected HeLa-S3 were washed twice with PBS (Gibco) and fixed in 4% paraformaldehyde (PFA) for 15 min in a wet chamber. After two additional PBS washing steps, cells were stained with Hoechst 33342 (Invitrogen; diluted 1:5,000 in PBS) for 15 min in a wet chamber and again washed twice with PBS. After coverslips had been air-dried, they were embedded in Vectashield Mounting Medium (Biozol) and analysed using the Leica SP5 confocal microscope (Leica) and the LAS AF Lite software (Leica). To stain human mitochondria, MitoTracker Orange CMTMRos (Life Technologies; kindly provided by V. Kozjak-Pavlovic, Biocentre, Würzburg) was used. The dye was added in the dark to a final concentration of 200 nM directly into the medium of the infected cells in the 37 °C incubator, 30 min before their harvest. After the 30 min incubation with the dye, the plates were covered with aluminium foil to prevent bleaching during the following steps. The supernatant was aspirated and the cells were washed with PBS and fixed with 4% PFA at 4 °C overnight. Hoechst staining and sample preparation was performed as described above. For flow cytometry-based analyses, infected cultures were washed twice with PBS, detached from the bottom of the plate by trypsinization and resuspended in complete DMEM. Upon pelleting the cells (5 min at 250g, room temperature), they were resuspended in PBS and analysed by flow cytometry using a FACSCalibur instrument (BD Biosciences) and the Cyflogic (CyFlo Ltd; version 1.2.1) or Flowing (Cell Imaging Core, Turku Centre for Biotechnology, Finland; version 2.5.0) software, respectively. Selection of intact HeLa-S3 cells was achieved by gating based on cell diameter (forward-scatter) and granularity (side-scatter) (linear scale). Of those, infected (GFP-positive) and non-infected (GFP-negative) sub-fractions were defined based on GFP signal intensity (FITC channel) versus auto-fluorescence (PE channel) (logarithmic scale). For cell sorting, RNAlater-fixed cells (see below) were first passed through MACS Pre-Separation Filters (30 μm exclusion size; Miltenyi Biotec) and then analysed and sorted using the FACSAria III device (BD Biosciences) at 4 °C (cooling both the input tube holder and the collection tube rack) and at a medium flow rate using the same gating strategy as described above, except that the gates for GFP-positive and GFP-negative fractions were conservative in order to prevent cross-contamination (as exemplified in Extended Data Fig. 1d). Typically ~2 × 105 cells of each fraction were collected for RNA isolation. To detect apoptotic cells, HeLa-S3 cells were washed twice with PBS and resuspended in 1× binding buffer (BD Pharmingen) to a concentration of 106 cells per ml. 100 μl of this cell suspension were mixed with 5 μl of APC-labelled annexin V (BD Pharmingen) and 1 μl of 500 mg ml−1 propidium iodide (PI; lyophilized stock from Sigma). Upon incubation for 15 min at room temperature, (light-protected) cells were subjected to flow cytometry using the MACSQuant Analyzer (Miltenyi Biotec). Upon gating of the fraction of intact cells based on cell diameter (forward-scatter) and granularity (side-scatter), the annexin-positive/PI-negative sub-population was determined by comparison against the appropriate single-stained controls in the APC vs PerCP channels, and quantified. Necrosis was evaluated by quantifying released lactate dehydrogenase (LDH) via the Cytotox96 assay (Promega) according to the manufacturer’s instructions. The absorbance at 490 nm was measured using a Multiskan Ascent instrument (Thermo Fisher). In order to convert the measured absorbance values into the relative proportion of dead cells, the maximal absorbance was determined by using 1× lysis solution (Promega) following the manufacturer’s instructions and referred to as 100% cytotoxicity. For both apoptosis and cytotoxicity measurements each biological replicate comprised three technical replicates. To quantify bacterial intracellular replication (Extended Data Fig. 1b), infected host cells were analysed by flow cytometry as described above, except that the increase in GFP intensity (geometric mean) was measured in the GFP-positive sub-population over time and normalized to that of the non-infected population in the same sample (example in Extended Data Fig. 1c). Alternatively, infected HeLa-S3 cultures were solubilized with PBS containing 0.1% Triton X-100 (Gibco) at the respective time points. Cell lysates were serially diluted in PBS, plated onto LB plates and incubated at 37 °C overnight. The number of colony forming units (c.f.u.) recovered was compared to that obtained from the bacterial input solution used for infection. In all cases, each biological replicate comprised three technical replicates. Infected cells were washed twice with PBS, trypsinized and pelleted. For ethanol fixations, cell pellets were re-dissolved in 0.1 volume of ice-cold PBS and then 0.9 volume of ice-cold ethanol (either 70% or 100%; as indicated) were added in single droplets during shaking (400 r.p.m., 4 °C) to avoid cell clumping. Fixation using stop solution (95% EtOH/5% water-saturated phenol)57 was performed by resuspending the cell pellet in PBS before the addition of 0.2 volume of stop solution and mixing. When PFA was used, the pellet was resuspended in the respective PFA concentration (0.5% or 4% PFA, pH 7.4, with or without 4% sucrose) and shaken for 15 min at 400 r.p.m., room temperature. PFA-induced crosslinks were reverted by an additional heating step for 15 min at 70 °C (refs 58, 59). For fixation with RNAlater (Qiagen), cell pellets were directly resuspended in RNAlater (1 ml per 5 × 106 cells). For systematic evaluation of different fixation protocols (Extended Data Fig. 1e–g), fixed cells had not been sorted but were either directly analysed upon fixation (30 min) or stored at −20 °C (ethanol-based fixatives) or 4 °C (others), respectively, overnight. To prepare RNAlater-fixed samples for sorting, tubes containing ~5 × 106 fixed cells were filled up with 10 ml of ice-cold PBS, centrifuged (5 min, 500g, 4 °C) and cell pellets resuspended in 2 ml of cold PBS. This cell suspension was filtered and sorted (as described above). In the dual RNA-seq experiments, as a reference for gene expression changes in host cells upon infection, a non-infected yet mock-treated control was included. The bacterial reference samples were derived from Salmonella grown in LB to an OD of 2.0, which either were then shifted to DMEM for 15 min, pelleted and fixed in RNAlater (see above) or were fixed directly (that is, without a medium exchange step) as indicated. Fixed Salmonella cells were pelleted and lysed using the lysis/binding buffer of the mirVana kit (Ambion). In order to maintain the approximate ratio of bacterial to host transcripts during RNA isolation, Salmonella lysates were mixed with host cell lysate in a way that the calculated proportion of individual Salmonella cells per infected host cell at the latest time point (see Extended Data Fig. 1h) was matched. The resulting mixture was then processed collectively. RNA was extracted from cells using the mirVana kit (Ambion) following the manufacturer’s instructions for total RNA isolation. To remove contaminating genomic DNA, samples were treated with 0.25 U of DNase I (Fermentas) per 1 μg of RNA for 45 min at 37 °C. If applicable, RNA quality was checked on the Agilent 2100 Bioanalyzer (Agilent Technologies). For qRT–PCR experiments total RNA was isolated using the TRIzol LS reagent (Invitrogen) according to the manufacturer’s recommendations and treated with DNase I (Fermentas) as described above. qRT–PCR was performed with the Power SYBR Green RNA-to-CT 1-Step kit (Applied Biosystems) according to the manufacturer’s instructions. Fold changes were determined using the 2(−ΔΔC ) method60. Primer sequences are given in Supplementary Table 1 and their specificity had been confirmed using Primer-BLAST (NCBI). For the estimation of Salmonella RNA within infection samples (Extended Data Fig. 1h), a dilution series of separately isolated Salmonella and HeLa-S3 total RNA was set up and in each case the ratio of rfaH/ACTB mRNAs was determined. The same was done for biological samples from infected cells as well as for the Salmonella reference controls. From the resulting trend-line equation the approximate proportion of the Salmonella transcriptome within mixed prokaryotic and eukaryotic total RNA samples could be deduced. Where indicated (Supplementary Table 1), Salmonella and eukaryotic host rRNA were removed using the Ribo-Zero Magnetic Gold Kit (Epidemiology) purchased from Epicentre/Illumina. Following the manufacturer’s instructions, ~500 ng of total, DNase-I-treated RNA from infection samples was used as an input to the ribosomal transcript removal procedure. rRNA-depleted RNA was precipitated in ethanol for 3 h at −20 °C. cDNA libraries for Illumina sequencing were generated by Vertis Biotechnologie AG, Freising-Weihenstephan, Germany. For dual RNA-seq of total RNA, at least 100 ng RNA were used for cDNA library preparation. DNase-I-treated total RNA samples were first sheared via ultra-sound sonication (4 pulses of 30 s at 4 °C each) to generate ~200–400 bp (average) fragmentation products. Fragments <20 nt were removed using the Agencourt RNAClean XP kit (Beckman Coulter Genomics). As an internal quality control for the pilot experiment (shown in Fig. 1), spike-in RNA (5′-AAAUCCGUUCGUACGGGCCC-3′; 5′-monophosphorylated and gel-purified) was added to a final concentration of 0.5%. The samples were poly(A)-tailed using poly(A) polymerase and the 5′ triphosphate (or eukaryotic 5′ cap) structures were removed using tobacco acid pyrophosphatase (TAP). Afterwards, an RNA adaptor was ligated to the 5′ monophosphate of the RNA fragments. First-strand cDNA synthesis was performed using an oligo(dT)-adaptor primer and the M-MLV reverse transcriptase (NEB). The resulting cDNA was PCR-amplified to about 20–30 ng μl−1 using a high fidelity DNA polymerase (barcode sequences for multiplexing were part of the 3′ primers). The cDNA library was purified using the Agencourt AMPure XP kit (Beckman Coulter Genomics) and analysed by capillary electrophoresis (Shimadzu MultiNA microchip electrophoresis system). cDNA libraries for dual RNA-seq on rRNA-depleted samples were constructed as described above, except for the following modifications. Upon RNA fragmentation, dephosphorylation with Antarctic Phosphatase (AP, NEB) and re-phosphorylation with T4 Polynucleotide Kinase (PNK, NEB) were performed. Oligonucleotide adapters were ligated to both the 5′ and 3′ ends of the RNA samples. First-strand cDNA synthesis was performed using M-MLV reverse transcriptase and the 3′ adaptor as primer. cDNA libraries from Salmonella-only samples were generated by fragmenting 5 μg of total RNA using ultrasound and RNAs <20 nt were removed using the Agencourt RNAClean XP kit (Beckman Coulter Genomics) as above. The RNA samples were poly(A)-tailed and 5′ppp structures were removed as before. RNA adapters were ligated to the 5′ monophosphate of the RNA and first-strand cDNA synthesis was performed using an oligo(dT)-adaptor primer and the M-MLV reverse transcriptase. The resulting cDNAs were PCR-amplified, purified using the Agencourt AMPure XP kit (Beckman Coulter Genomics) and analysed by capillary electrophoresis (Shimadzu MultiNA microchip). Generally, for sequencing cDNA samples were pooled in approximately equimolar amounts. The cDNA pool was size-fractionated in the size range of 150–600 bp using a differential clean-up with the Agencourt AMPure kit. For the dual RNA-seq pilot experiment (Fig. 1), single-end sequencing (100 cycles) was performed on an Illumina HiSeq 2000 machine at the Max Planck Genome Centre Cologne, Cologne, Germany. For dual RNA-seq on rRNA-free samples as well as for conventional RNA-seq of Salmonella-only samples, single-end sequencing (75 cycles) was performed on a NextSeq500 platform at Vertis Biotechnologie AG, Freising-Weihenstephan, Germany. All RNA-seq data discussed in this publication have been deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO Series accession number GSE60144. For the accession numbers of individual experiments, see Supplementary Table 1. Total RNA prepared with TRIzol LS reagent (Invitrogen) was separated in 6% (vol/vol) polyacrylamide-8.3 M urea gels and blotted as described11. We loaded per lane either 5–10 μg of RNA from pure bacterial samples (Extended Data Figs 3d and 9a), 2 μg total RNA from sorted cell samples (Extended Data Fig. 8b), or 50 μg total RNA from unsorted infection samples (Fig. 2b). Hybond XL membranes (Amersham) were hybridized at 42°C with gene-specific [32P] end-labelled DNA oligonucleotides (see Supplementary Table 1 for sequences) in Hybri-Quick buffer (Carl Roth AG). The pinT promoter region was amplified by PCR using primers JVO-7036/-7037 and inserted via the AatII and NheI sites in the backbone of plasmid pAS093, resulting in plasmid pYC65. To identify the PhoP binding sites in a minimal fragment, the pinT promoter region was truncated by amplifying pYC65 using Phusion polymerase (NEB) with JVO-9393/-7387. The critical residues in the PhoP binding motif (T T ) were mutated to adenines by site-directed mutagenesis with JVO-12461/-12462 and Phusion polymerase (NEB). For pulse-expression of PinT in in vitro grown Salmonella, we used arabinose-induced overexpression of PinT from a pBAD plasmid previously described10, 51, 61 with minor modifications. Briefly, wild-type Salmonella that carried either a pKP8-35 (pBAD control), pYC5-34 (pBAD-PinT) or pYC60 (pBAD-PinT*) plasmid were grown overnight in LB and, the next day, the cultures were 1:100 diluted and further grown in LB to an OD of 2.0. l-arabinose (Sigma) was added to a final concentration of 0.2%; 5 min later RNA was extracted using TRIzol LS reagent (Invitrogen) and analysed by RNA-seq (~3–5 million reads/library). For the same experiment under SPI-2-inducing conditions, overnight cultures of the three strains were washed 2× with PBS and 1× with SPI-2 medium28, diluted 1:50 in SPI-2 medium and grown to an OD of 0.3 before PinT expression was induced as above. For the pulse-expression of PinT inside host cells (Extended Data Fig. 6d, e), HeLa-S3 cells were infected with the same three strains as above and 4 h after infection, 0.2% l-arabinose was supplemented directly into the DMEM medium. Activation of inducible sRNA expression in intracellular bacteria was confirmed by qRT–PCR over a time-course of 20 min (Extended Data Fig. 6d), demonstrating full induction levels to be reached already at 5 min. Thus, for Extended Data Fig. 6e the host cells were lysed at 5 min after induction with ice-cold 0.1% Triton X-100/PBS and further incubated for 30 min on ice with pipetting up and down from time to time to improve host cell lysis efficiency. Then the intact bacterial cells were pelleted by centrifugation for 2 min at 16,100g (4 °C) and resuspended in RNAlater (Qiagen). The fixed bacterial cells were further enriched against the host background via cell sorting (FACSAria III, BD Biosciences) and selective gating for the fraction of GFP+ bacterial cells released from their hosts. From those, total RNA was isolated and analysed by RNA-seq as above except that sequencing was to a depth of ~20 million reads per library as necessitated by remaining host-derived RNA fragments. Immunoblotting of Salmonella proteins was done as previously described62. Briefly, samples from Salmonella in vitro cultures were taken corresponding to 0.4 OD , centrifuged for 4 min at 16,100g at 4 °C, and pellets resuspended in sample loading buffer to a final concentration of 0.01 OD per μl. After denaturation for 5 min at 95 °C, 0.05-OD equivalents of the sample were separated via SDS–PAGE. Gel-fractionated proteins were blotted for 90 min (0.2 mA per cm2; 4 °C) in a semi-dry blotter (Peqlab) onto a PVDF membrane (Perkin Elmer) in transfer buffer (25 mM Tris base, 190 mM glycin, 20% methanol). Blocking was for 1 h at room temperature in 10% dry milk/TBST20. Appropriate primary antibodies (see Supplementary Table 1) were hybridized at 4 °C overnight and – following 3 × 10 min washing in TBST20 – secondary antibodies (Supplementary Table 1) for 1 h at room temperature. For western blotting of human proteins, infected cells were harvested in sample loading buffer (500 μl per well; six-well format), transferred to 1.5 ml reaction tubes, boiled for 5 min at 95 °C and 20 μL per lane were loaded onto a 10% PAA gel for SDS–PAGE as above. After blotting and blocking (as above), the membrane was probed with the respective primary antibody at 4 °C overnight and—upon washing (as above)—with the secondary antibody for 1 h at room temperature (a full list with information about all antibodies and sera used is given in Supplementary Table 1). After three additional washing steps for each 10 min in TBST20, blots were developed using western lightning solution (Perkin Elmer) in a Fuji LAS-4000. In Fig. 3e, intensities of protein bands were quantified using the AIDA software (Raytest, Germany) and normalized to GroEL levels. To mimic the early stages of the infection of a host cell in vitro, the indicated Salmonella strains were grown in LB overnight, diluted 1:100 in LB and grown to an OD of 2.0 (that is, a condition under which SPI-1 is highly induced4, 11), washed twice with PBS and once with SPI-2 medium28 at room temperature, diluted 1:50 in pre-warmed SPI-2 medium (defined as t ) and grown further in Erlenmeyer flasks at 37 °C for the indicated time periods. At the respective time points, samples were taken for RNA-seq, western blotting, and GFP fluorescence measurements. To measure the GFP intensity of reporter strains, bacteria were grown in LB in presence of Amp and Cm until an OD of 2.0 was reached. Salmonella cells corresponding to 1 OD were pelleted and fixed with 4% PFA. GFP fluorescence intensity was quantified for each 100,000 events by flow cytometry with the FACSCalibur instrument (BD Biosciences). Data were analysed using the Cyflogic software (CyFlo). To monitor SPI-2 activation in real time, a transcriptional gfp reporter was constructed by inserting the SPI-2-dependent ssaG promoter into plasmid pAS0093 via AatII/NheI sites as previously described8. The resulting plasmid pYC104 was co-transformed with either the pBAD-ctrl. or pBAD-PinT plasmid into the indicated strain backgrounds. The resulting strains were grown overnight in LB (+Amp + Cm) and then diluted 1:100 and further grown in the same medium to an OD of 2.0. A volume of 1 ml of the culture was pelleted and the collected cells shifted to SPI-2 medium28 (defined as t ) as described above, except that the growth experiment was conducted in 96-well plates (Nunc Microwell 96F, Thermo Scientific). After measuring the OD and GFP intensity at t , l-arabinose was added to each well to final concentration of 0.2% for sRNA induction and bacteria were grown for 20 h at 37 °C (with shaking) with measurements of both the OD and GFP fluorescence in 10 min intervals using the Infinite F200 PRO plate reader (Tecan). HeLa-S3 cells were infected with wild-type Salmonella, ΔpinT or pinT+ mutant strains at an m.o.i. of 5 as described above. Culture supernatant samples were taken at 20 h p.i. and analysed using the ELISA kit for human CXCL8/IL-8 (R&D Systems). Code availability. In order to document the details and parameters of the (dual) RNA-seq data analyses and to make the biocomputational approaches reproducible for others, we implemented the workflows as Unix Shell scripts. These scripts are deposited at Zenodo (DOI: 10.5281/zenodo.34695, https://zenodo.org/record/34695). Please refer to Supplementary Table 1 for descriptions of the analyses. For all RNA-seq experiments listed in Supplementary Table 1, Illumina reads in FASTQ format were trimmed with a Phred quality score cut-off of 20 by the program fastq_quality_trimmer from FASTX toolkit version 0.0.13 (http://hannonlab.cshl.edu/fastx_toolkit/). Reads shorter than 20 nt after adaptor- and poly(A)-trimming were discarded before the mapping. The reads were aligned to the Salmonella enterica SL1344 genome (NCBI RefSeq accession numbers: NC_016810.1, NC_017718.1, NC_017719.1, NC_017720.1) and—where applicable—the human (hg19 – GRCh37; retrieved from the 1000 Genomes Project63), the mouse (GENCODE M2, GRCm38.p2), or the porcine genome sequence (ENSEMBL, Sscrofa10.2), in parallel. The mapping was performed using the READemption pipeline (version 0.3.5)64 and the short read mapper segemehl and its remapper lack (version 0.2.0)65 allowing for split reads66. Mapped reads with an alignment accuracy <90% as well as cross-mapped reads, that is, reads which could be aligned equally well to both host and Salmonella reference sequences, were discarded. The resulting data were used for visualization (see for example, Fig. 1b and Extended Data Fig. 2b). Reads of the high resolution time-course experiment (cDNA libraries numbers 27–77 in Supplementary Table 1) that were detected as cross-mapped by READemption (see above) were further inspected: their median percentage over the entire time-course was 0.25% with increased fractions for the later time points, implying that those reads are mainly contributed by Salmonella cells. We observed that the majority of the cross-mapped reads aligned to Salmonella rRNA or tRNA loci, while on the human side no gene class preference was observed (data not shown). For dual RNA-seq experiments (cDNA libraries 1–184, 215–256 in Supplementary Table 1) after mapping differential expression analysis was carried out separately for the host and the pathogen. Strand-specific gene-wise quantifications for each data subset were performed by READemption64. Host transcript expression analyses are based on annotations from GENCODE (version 19)67, NONCODE (version 4)68 and miRBase (version 20)69 after removing redundant entries. The annotation for Salmonella genes was retrieved from NCBI (under the above mentioned accession numbers) and manually extended with small RNA annotations4, 70. In either organism, multi-mapped reads were removed and only uniquely mapped reads were considered for the expression analysis. Differential gene expression analyses were performed with the edgeR package (version 3.10.2)71 using an upper-quartile normalization and a prior count of 1. Where needed (that is, to correct for batch effects in the comparisons between wild-type and mutant infections; the comparisons displayed in Figs 3 and 4 and Extended Data Figs 5, 7,8,9), sequencing data were further normalized using the RUVs correction method72 with k = 3. For this purpose, we treated the samples time-point-wise to remove unwanted nuisance factors. At each time point our covariate of interest was the pinT status of the infecting bacterium. This is constant within replicate blocks, which are used for the RUVs correction. Host or bacterial genes with at least 10 uniquely mapped reads in three replicates were considered detected. Genes with an adjusted P value < 0.05 were considered differentially expressed. Differential expression analysis for conventional (bacteria only) RNA-seq experiments (cDNA libraries numbers 185–214 in Supplementary Table 1) was done similarly, except that a cut-off of ≥50 uniquely mapped reads was used as a detection threshold. Based on the obtained BAM files, coverage files in wiggle format were generated by READemption64 in a strand-specific manner and split by organism. In each case, coverage files are based on uniquely mapped reads and normalized by the total number of uniquely aligned reads per organism. For Fig. 4e, wiggle files were visualized using the Integrated Genome Browser (version 8.4.4)73. A database of pathways, regulons, and genomic islands was constructed using information obtained from the KEGG database74 (organism code sey), the SL1344 genome annotation70, and relevant literature sources (see Supplementary Table 1). Pearson correlation coefficients between changes in PinT expression and changes in expression of each gene within each regulon over the time-course of wild-type Salmonella infection (cDNA libraries number 27, 30, 33, 36, 39, 42, 44, 47, 50, 53, 56, 59, 61, 64, 67, 70, 73, 76 in Supplementary Table 1) were plotted in Fig. 2d. To assess enrichment of differentially expressed transcripts in pathways in the comparative infection experiments (cDNA libraries numbers 27–77 and 152–184 in Supplementary Table 1) and the in vitro assay (cDNA libraries numbers 185–202 in Supplementary Table 1), gene set enrichment analysis (GSEA; version 2.1.0) was run on the log fold changes reported by edgeR. The GSEA was performed in ranked list mode (with statistic classic) and gene sets containing less than 15 or more than 100 entries were excluded. Extended Data Fig. 5a reports all pathways significant at an FDR-corrected P value of at most 0.05 in at least one time point. Host pathway enrichment studies were performed consistently with bacterial analyses using GSEA on human pathways available in the KEGG database (downloaded January 22, 2014) using the same settings described above. Pathways with an adjusted P value ≤ 0.05 were considered to be significantly modulated. Data visualization for Extended Data Fig. 8a was produced using the Bioconductor package Pathview75. Genes displayed in Fig. 1d, that is, genes whose transcription is known or predicted to be regulated by the binding of nuclear factor κB (NF-κB) to their promoter or genes whose products have been shown to promote an NF-κB response, were retrieved from the GeneCards76 and Boston University Biology (http://www.bu.edu/nf-kb/gene-resources/target-gene) databases or refs 77, 78. STAT3 target genes denoted in Fig. 4b were retrieved from ref. 79. We used Cufflinks/Cuffdiff (version 2.2.1)80, 81 to test for differentially expressed isoforms in the high-resolution, comparative dual RNA-seq time-course data set (cDNA libraries number 27–77 in Supplementary Table 1). In a first step, we used Cufflinks to quantify transcript isoforms in the mapped read data. Afterwards, all transcript annotations were merged using Cuffmerge and differentially expressed isoforms were called using Cuffdiff. To identify bacterial and human genes with similar expression kinetics across the time-course of the infection of HeLa-S3 cells (cDNA libraries number 27–77 in Supplementary Table 1), we used RUVs-corrected, abundance-filtered and normalized read counts (see above). Absolute counts were then transformed into standard z-scores for each gene over all considered samples as follows: for each gene, the z-score was calculated as the absolute read count minus the mean read count over all samples, divided by the standard deviation of all counts over all samples. Genes with a standard deviation <2 were excluded from further analysis. Pearson correlation coefficients were calculated between all remaining bacterial genes and all remaining human genes, and P values were calculated using the function cor.test in R. To account for a possible temporal delay between Salmonella expression changes and effect manifestation in the host cell, a time-shift was allowed. This means the expression of Salmonella genes at each time point was compared to host expression at the subsequent time point. Human genes were considered to be correlated with a bacterial gene if they had a P value of less than 10−4 and a Pearson’s r greater than 0.65. This resulted in a total of 751 clusters of human genes showing correlation in expression with a bacterial gene, approximately half of which (see Supplementary Table 1) had at least one enriched GO term associated with them (adjusted P value < 0.05) as tested using the software tool Ontologizer 2.0 (build: 20100310-351)82 with the gene ontology definition obtained from the Gene Ontology Consortium (data-version: releases/2015-09-26) and the Universal Protein Resource (UniProt) gene annotation (generated: 2015-09-14). To account for the possibility that multiple bacterial genes might be associated with a human gene cluster a correlation analysis was performed for all against all bacterial genes as described above, with the only exception that no time-shift was allowed. For this, we focused on seventeen gene clusters that were built on bacterial genes encoding for secretion-associated gene products (according to UniProt; see Supplementary Table 1). Detailed inspection of these clusters revealed the one depicted in Fig. 4b (centred on the bacterial SPI-2 gene sseC) which contained many further (bacterial and human) genes with pronounced PinT-dependent expression changes – that is, genes that showed differential expression between wild-type and ΔpinT infection at several time points p.i. In all RNA-seq-based analyses, transcript expression changes that were associated with an adjusted P value < 0.05 (reported by edgeR) were considered significantly differentially expressed. For Fig. 3b, a Monte Carlo permutation test was performed on the median fold change of genes in the SPI-2 regulon, using 105 randomly selected gene sets of the same size. This indicated the significant de-repression (P < 0.05) of the SPI-2 regulon in the absence of PinT at 2 and 8 h after the infection of HeLa cells, at 2, 6 and 16 h after the infection of 3D4/31 cells, and in the in vitro assay. Tests for the evaluation of increased host cell death in Extended Data Fig. 1a were performed using a one-tailed Student’s t-test. *P values ≤ 0.05 were considered significant and ***P values ≤ 0.001 were considered very significant. The significance of gene activation in qRT–PCR results in Fig. 4c and Extended Data Figs 5b, c and 7c, d or the ELISA assay in Extended Data Fig. 7e was assessed using a one-tailed Mann–Whitney U-test. The significance of differences in intracellular replication between the ΔpinT strain and wild-type Salmonella (Extended Data Fig. 4d) was evaluated using a two-tailed Mann–Whitney U-test.