To quantify the temporal interactions between L. monocytogenes and host cells we used live cell microscopy approaches. We infected HeLa cells, a well-established model of InlB-mediated L. monocytogenes infection (Dramsi et al., 1995), with L. monocytogenes EDGe: InlA^m (Wollert et al., 2007) expressing chromosomally integrated promoter PactA driving expression of green fluorescent protein (GFP), in addition to constitutively expressed red fluorescent protein (dsRed), referred herein as Lm-dsRed PactA-GFP (see Materials and Methods, as well as Tables 1 and 2 for list of plasmids and strains used in the study). We used a membrane impermeant gentamicin protection assay, with host cells exposed to L. monocytogenes for 2 h, before antibiotic treatment, with cells followed by live-cell microscopy for additional 4 h. Based on previous work (Wang et al., 2015), we used MOI of 20 (20:1 pathogen to host cell ratio), which result in spatially separated replicative invasion events and thus allow quantitative investigation of the infection process ( Figure 1A ). We found that changes in MOI were generally associated with intracellular growth, with significantly highest growth at MOI 20 ( Supplementary Figure S1 ). Further increase of MOI to 100 resulted in a reduction of growth, consistent with previous reports (Rengarajan et al., 2016).

We found that during the time-course of the infection bacteria aggregate into distinct clusters ( Figures 1B , see also 1C and Supplementary Movie 1 ). Initially, bacteria appear to be moving freely in small clusters in the culture media, but immediately arrange into larger multi-cellular structures, which significantly increase in size over time with of up to 50 bacteria at the time of the gentamicin treatment (see Figure 1D ). Although L. monocytogenes replicates in the culture media, which generally contributes to the effect, replication alone cannot explain existence of the large aggregates, since given the doubling time of 45-60 mins in rich broth (Jones and D’Orazio, 2013) approximately 2 divisions may occur before the addition of gentamicin. These aggregates not only form in the culture media but adhere to host cells and continue to increase in size after host cell binding. Simultaneously with formation of aggregates we observed upregulation of PactA-GFP signal, with an onset around 60 mins ( Figure 1B, C ). Bacteria appear yellow in composite images, which represents the induction of PactA-GFP outside host cells, in addition to constitutively expressed dsRed. During the experiment, PactA-GFP becomes bright green in composite images, which is especially evident at times between 2 and 6 h post infection ( Figure 1B ). This is consistent with bacteria being internalised and subsequently escaping to the cytosol, resulting in strong intracellular induction of actA (previous reports demonstrating 200-fold increase in transcription (Shetron-Rama et al., 2002; Chatterjee et al., 2006; Toledo-Arana et al., 2009). Overall, in agreement with the literature, our data show that prior to gentamicin treatment a subset of bacteria is able to invade host cells, where it can subsequently escape vacuole into the host cytoplasm (which can occur at the order of 10 mins (Peron-Cane et al., 2020)) resulting in rapid intracellular replication. The latter typically occurred between 2 to 4 h from the gentamicin treatment ( Figure 1B ; Supplementary Movie 2 ) and coincided with elevated PactA-GFP expression. Interestingly, given the high MOI, while most, if not all host cells interacted with multiple aggregates, we found that only few interactions resulted in intracellular replication of L. monocytogenes. Importantly, aggregation as well as ActA upregulation was also induced upon infection of primary HUVEC cells, which similarly to HeLa cells lack E-cadherin expression (Kalender et al., 2022). This demonstrates that aggregation is a general InlB-mediated phenomenon, which is physiologically relevant for human infection ( Figure 1E ).

ActA expression has been typically associated with intracellular L. monocytogenes (Pillich et al., 2017) and therefore the induction in actA transcription during aggregation suggests a role for PrfA-mediated virulence in the process. Also, aggregates of L. monocytogenes have been previously observed in vitro in culture broth and in vivo in the intestine during infection of mice, and have been shown to be PrfA and ActA-mediated, and specifically facilitated by the extracellular C terminal domain (amino acids-393-585) of ActA ( Figure 1F ) (Travier et al., 2013). We found that wildtype, but not ΔprfA nor ΔactA mutant strains, exhibited aggregation ( Supplementary Figure S2 ). We also confirmed that the actA-ΔC-dsRed strain in which the C-terminus of ActA has been deleted, did not exhibit aggregation upon infection of HeLa cells ( Figures 1G, H ). Therefore, these data demonstrate that intracellular InlB-mediated invasion of L. monocytogenes into human non-phagocytic cells involves the formation of ActA-dependent bacterial aggregates.

We hypothesised that aggregation represents an invasion strategy. We used time-lapse microscopy movies to identify replicative invasion events by tracking host cells that were harbouring replicating bacteria, i.e., those increasing in number and robustly expressing PactA-GFP, confined within individual cell boundaries over the 6 h duration of the experiment. We only counted primary infection events associated with aggregate binding prior to gentamicin treatment and subsequent replication in target cell, while excluding rare and late secondary infection events associated with cell-to-cell spread or division of infected host cells. Subsequently, we compared the fraction of individual HeLa cells harbouring replicating bacteria with wildtype Lm-GFP to that of the non-aggregating actA-ΔC and ΔactA strains as a measure of their ability to establish replicative infections ( Figure 2A ). We found that on average 8.8% (±1.1% standard deviation) of HeLa cells facilitated intracellular replication of wildtype bacteria (calculated as a fraction of total HeLa cells) ( Figure 2A ). In contrast, only 2.3% (±1.2%), i.e., approximately 4-times less, HeLa cells were infected with the non-aggregating actA-ΔC strain, with 0.9% (±0.07%) cells being infected with a complete ΔactA mutant ( Figure 2A ).

To directly compare wild type and actA mutant strains, we simultaneously co-infected cells with wildtype Lm-GFP and actA-ΔC-dsRed bacteria, each at MOI=10 (see Figure 2B for an example of an event where both strains established replicative infections in proximity). In agreement with previous analyses, we found that the number of host cells harbouring replicative infection of the non-aggregating strain was significantly reduced in comparison to the wildtype (4.8 ±0.5% vs. 0.5 ±0.1%, respectively, Figure 2C ). Bacteria may cooperate to manipulate host cells, for example cooperation between Salmonella allows otherwise non-invasive strains to enter host cells (Ginocchio et al., 1992; Kazmierczak et al., 2001; Misselwitz et al., 2012; Lorkowski et al., 2014). Although we found that some individual actA-ΔC-dsRed bacteria were incorporated into wildtype Lm-GFP aggregates ( Figure 2B , see arrowheads), we did not find evidence that actA-ΔC-dsRed was able to establish replicative invasion in these aggregates. This suggests that ActA C-terminus not only facilitates aggregation but also has a role in invasion.

Finally, we wanted to determine if aggregation plays a role in invasion of primary cells. We infected primary HUVEC cells with wildtype and actA-ΔC reporter strains. Due to increased cytotoxicity of HUVEC cells when a large bacterial inoculum was used (Rengarajan et al., 2016), we employed a lower MOI of 5, which was sufficient to induce replicative invasions ( Figure 2D ). In agreement with assays in HeLa cells, we found that approximately 2.1% (±0.3%) of HUVEC cells harboured replicative invasions of the wildtype bacteria, whilst significantly less, 0.7% (±0.6%), harboured replicative invasions of the non-aggregating actA-ΔC strain ( Figure 2E ).

Overall, these data demonstrate that while replicative infections are relatively rare events (in cells negative for E-cadherin), bacterial aggregates facilitate more efficient replicative invasions compared to non-aggregating bacteria.

To understand how aggregates facilitate robust intracellular replication of L. monocytogenes, we first characterised adherence and intracellular invasion. We used anti-Lm antibody staining in cells that were washed prior to the fixation protocol thus removing bacteria not associated with host cells ( Figure 3A ). This allowed distinguishing bacteria that were intracellular at 2 h post-infection from those that were adhering to the cell surface (having washed away bacteria that were not bound to host cells). In agreement with live-cell imaging ( Figure 1 ), wildtype Lm-dsRed, but not actA-ΔC-dsRed strain, exhibited aggregation upon infection of HeLa cells in those assays ( Figure 3B ; Supplementary Figure S3A ). On average, each host cell had multiple spatially resolved interactions with adhering bacteria (see the correlation between aggregate area and number of bacteria, Supplementary Figure S3B ), but we found significantly more interactions for wildtype (5.4 ±0.6) than the mutant bacteria (3.1 ±0.3), calculated as a function of all host cells and aggregates ( Figure 3C ). In terms of actual adherent bacteria per host cell, each host cell was bound by significantly more individual wildtype Lm-dsRed bacteria (20.7 ±1.7) than for the actA-ΔC-dsRed strain (3.8±0.5) ( Figure 3D ). This demonstrates that in terms of absolute numbers, aggregated wildtype bacteria produce approximately 5-fold more host cell adhesions than the non-aggregating actA-ΔC mutants.

Next, we used the extracellular anti-Lm antibody staining to quantitatively distinguish between adherence (positive staining) and intracellular invasion (lack of staining). We found that whilst most of the bacteria were only adhering to host cells, there was clear evidence of intracellular invasion for both strains as indicated by red fluorescent signal ( Figure 3B ). These data demonstrate that for the wildtype, 70% of individual bacteria that were internalised at 2 h post-infection originated from aggregates where whole or part of an aggregate was intracellular ( Supplementary Figure S3B ). In terms of absolute numbers, we found on average 4.1 ±2 wildtype bacteria were internalised per host cell (in those host cells that were infected, i.e., harboured intracellular bacteria). In contrast, in the case of less frequent infections with actA-ΔC-dsRed where only one bacterium was internalised ( Figure 3E ). Based on negative and positive anti-Lm staining, we then counted the number of intracellular and total number of associated bacteria across recoded images to account for the different level of adherence observed in the wildtype and actA-ΔC strains. We found that on average, a wildtype bacterium, which was associated with a host cell, had a 0.02 ±0.006 probability of establishing an intracellular invasion ( Figure 3F ). This was approximately 3-fold higher than the corresponding probability for the actA-ΔC bacterium (0.007 ±0.003). Given that multiple wildtype bacteria invade from a single aggregate, probability of an intracellular invasion per aggregate was 6-fold higher (0.05 ±0.01) than that of a non-aggregating strain. Overall, this shows that aggregation not only promotes more efficient interactions with host cells, but also more efficient intracellular invasion per bacterium.

Overall, in our internalisation experiments, a host cell had 0.17 ±0.04 probability of harbouring (at least 1) intracellular invasion event, which was ~5 fold more than that of the actA-ΔC strain (0.03 ±0.01, Figure 3G ). We wanted to understand if aggregation had a direct effect on intracellular replication or rather simply increased the chance of adherence and intracellular invasion. Given that we already estimated the overall probability of host cell harbouring replicative invasion ( Figure 2A ), we applied conditional probabilities to quantify if internalised wildtype and actA-ΔC bacteria had different abilities to replicate in the host. Using this approach, the conditional probability of harbouring a replicative invasion given that the bacteria are internalised (probability of replication of intracellular bacteria) P(R/I)=P(R)/P(I) is defined by the ratio of the overall replication event probability P(R) and the probability of internalisation P(I) (Moran et al., 2023). We found that the probability of a host cell harbouring replicative infection of intracellular wildtype L. monocytogenes was 0.53 ±0.14, while the equivalent probability for the actA-ΔC strain was 0.66 + 0.4 (based on error propagation). These numbers cannot be distinguished statistically, which suggest, that aggregation does not affect the ability of cells to replicate following invasion but that does lead to greater invasion.

In HeLa cells the internalisation pathway occurs mostly through the InlB-MET clathrin-mediated uptake pathway (Vessey et al., 1995; Braun et al., 1998). To validate the degree to which InlB is required in cell invasion we used a gentamicin protection assay. HeLa cells were infected with wildtype L. monocytogenes and the ΔinlB deletion mutant and after 2 hours gentamicin was added to kill extracellular bacteria. We found a significant reduction in CFU per well in the mutant bacteria, comparing to the wild type, which was maintained from addition of gentamicin for at least 8 h, consistent with the InlB-MET-mediated internalisation ( Figure 4A ). For example, at 3 hours the total number of CFU per well in the wildtype was 8.52 (± 3.71) x 10⁴ CFU/mL. In the ΔinlB mutant there was a significant 9-fold decrease in viable intracellular bacteria to 9.29 (± 5.74) x 10³ CFU/ml. This residual effect is potentially associated with previously described EDGe: InlA^m tropism for N-cadherin expressed by HeLa cells (Tsai et al., 2013). Overall, these analyses confirmed the invasion is predominantly occurring via the InlB-MET pathway.

Previous data suggest that InlB acts as a ‘‘molecular clamp’’ that forces the otherwise flexible MET protein into a rigid conformation through receptor clustering events, a key step leading to MET activation (Jonquières et al., 1999; Banerjee et al., 2004; Niemann et al., 2007). However, it is unclear, whether aggregates behave quantitatively different than individual bacteria, i.e., perhaps by engaging more receptors thus induce more clustering and uptake. Therefore, to further investigate the interactions between aggregates and MET we used immunostaining to measure and manipulate the cell-surface MET expression in HeLa cells upon infection with wildtype Lm-dsRed and actA-ΔC-dsRed strains ( Figure 4B ). As previously, cells were washed during the fixation protocol thus removing bacteria that were not associated with host cells. We found that adherence of wildtype aggregates was correlated with areas where MET localisation was more abundant than that of single non-aggregating actA-ΔC bacteria and levels of MET ( Figure 4C ; Supplementary Figure S4A ). This suggests that bacterial aggregates promote a formation of dense MET clusters on the cell membrane, which are not present following a single bacterium interaction. To further investigate this interaction, we used a pulse and chase method to deplete the cells of available MET receptors. We hypothesised that by reducing or depleting the available MET on the cell surface we would see a reduction in the ability of wildtype L. monocytogenes to associate with cells. After treating HeLa cells with anti-MET for 1 hour we saw an almost total reduction in the level of MET on the cell surface ( Supplementary Figure S4B ), in agreement with a previous study (Li et al., 2019). Infection of MET-depleted cells with wildtype L. monocytogenes resulted in a reduction in the number and size of aggregates associated with host cells after 2 h infection, compared to control cells in which MET had not been depleted ( Figure 4D ). In these experiments, cells were also washed to remove non-adherent bacteria. Quantification of these images showed that there was a significant reduction in the average number of bacteria associated per host cell in the MET-depleted cells, from 8.17 (± 0.27) in the untreated control cells to 3.25 (± 1.07) in the MET-depleted cells ( Figure 4E ). The average size of aggregates was also reduced, from 12.6 (± 17.6) µm² to 7.1 (± 8.4), equivalent to the reduction in the median from 3.7 to 2.6 and range from 64.7 to 24.4, respectively ( Figure 4F ). Overall, these data demonstrate that binding of L. monocytogenes is MET-dependent, while aggregates, but not individual bacteria, facilitate MET clustering resulting in more robust host cell invasion.

Our microscopy data show that L. monocytogenes aggregation occurs within the first 2 hours of exposure to host cells ( Figure 1C ); therefore, we hypothesised that bacteria are responding to secreted factors produced by host cells. To test this hypothesis wildtype Lm-dsRed-PactA-GFP was incubated for 2 hours in spent tissue culture media collected from HeLa cells and compared against bacteria incubated in fresh tissue culture media (see Materials and Methods for the generation of spent tissue culture media). We found that bacteria incubated in spent tissue culture media exhibited aggregation as well as upregulation of virulence gene expression, observed by GFP fluorescence from the PrfA-regulated PactA promoter ( Figures 5A, B ; Supplementary Movie 3 ). In contrast, when the bacteria were exposed to fresh tissue culture media, there was no aggregation nor upregulation of virulence reporter expression ( Supplementary Movie 4 ).

A major advantage of aggregation being induced by spent tissue culture media was that it allowed investigations without the presence of host cells. We therefore used this system to uncover global gene expression changes in L. monocytogenes in response to spent and fresh tissue culture media, in the wildtype and non-aggregating actA-ΔC strain. We performed RNA-seq analyses of 16 samples across 4 different conditions (media vs. genotype) ( Figure 5C ). Principal component analysis (PCA) showed a robust separation between type of media (spent or fresh), but not between wildtype and non-aggregating actA-ΔC strain ( Figure 5D ). In addition, we observed a batch effect associated with data being collected and sequenced in two independent experiments (8 samples including two biological replicates each). A formal differential gene expression analysis (Love et al., 2014) showed 366 genes that were upregulated and 444 genes downregulated in the wildtype in spent media (compared to the fresh media control, Figures 5E, F , see also Supplementary Data Dheet 1 for gene lists and analyses). In contrast, there were only 14 genes that were differentially regulated between wildtype and the actA-ΔC mutant strain in spent media, and none in fresh tissue culture media. This strongly suggests that spent media robustly activates gene transcription in L. monocytogenes, regardless of its ability to aggregate.

In terms of gene expression patterns, we found that spent media significantly induced transcription of the PrfA regulon including actA, inlA, inlB, inlC, plcB, hly, mpl as well as prfA ( Figure 5G ). For example, consistent with the imaging data, actA mRNA levels in the wildtype exhibited an average 4.2-fold increase in spent media, comparing to fresh media, while plcB and hly mRNA levels, required for efficient invasion (Grundling et al., 2003; Cossart, 2011) exhibited 3.8 and 3-fold increases, respectively. Both inlA and inlB exhibited statistically significant fold changes > 2.2, and while also regulated by SigB in the intestine (Toledo-Arana et al., 2009). No upregulation was observed of sigB regulon or other virulence genes involved in intracellular growth such as codY.

In addition to the PrfA regulon, we found several other systems ( Figure 5G ), which were upregulated in response to spent tissue culture media. The greatest fold increase was seen with the genes involved in ethanolamine and 1,2-propanediol metabolism (pduB, pduL, pduM, cbiB, cbiD, cobC). Genes (srtB, isdC, isdE, fepB, hbp2) encoding proteins involved in iron scavenging were also upregulated but to a lesser extent than the PrfA regulon ( Figure 5G ) as were the tryptophan biosynthesis genes (trpB, trpC, trpD and trpE), previously shown to be involved in host cell invasion (Mraheil et al., 2011). Ethanolamine and 1,2-propanediol metabolism genes were previously associated with persistent L. monocytogenes strains (Fox et al., 2011). Their expression was suggested to provide competitive advantage over commensal bacteria in the host due to utilisation of ethanolamine as a sole carbon source (Srikumar and Fuchs, 2011) and enhanced anaerobic growth (Zeng et al., 2019; Zeng et al., 2021). The srtB gene encodes Sortase B (SrtB), a second class of sortase in L. monocytogenes, involved in the attachment of a subset of proteins to the cell wall by recognising an NXZTN sorting motif (Bierne et al., 2004). The isdC and isdE genes present in the svpA-srtB operon encode for proteins that may be involved in a high affinity haem uptake system required for scavenging and Fe²⁺/Fe³⁺ binding, although the published data is somewhat conflicting regarding their function (Bierne et al., 2004). The fepB gene encodes for deferrochelatase, which is involved in the recovery of exogenous haem iron while hbp2 gene encodes for a protein involved in haem acquisition (Lechowicz and Krawczyk-Balska, 2015).

Significantly downregulated genes in spent tissue culture media included genes with roles in amino acid (hisA, hisE, metC, metE, ilvA, ilvB, ilvD) and pyridoxal 5’- phosphate P5P (pdxS, pdxT) biosynthesis, phosphate transport (pstA, pstB, pstC) and anaerobic metabolism (nrdG, nrdD) ( Figure 5G ). P5P is an essential cofactor for numerous metabolic enzymes (Belitsky, 2004). The pstA, pstB and pstC genes encode transmembrane proteins involved in the phosphate transport system (Moreno-Letelier et al., 2011; Choi et al., 2015). Catalytic subunit nrdD and activase nrdG encode class III anaerobic ribonucleotide reductase used in anaerobic conditions, however this NrdD protein has impaired function in EGD-e due to a 6 amino acid deletion (Ofer et al., 2011).

Taken as a whole, these data indicate responses by L. monocytogenes to likely changes in available nutrients present in the spent tissue culture media. In particular, the observed activation of iron uptake systems suggests that bacteria are adapting to changes in iron concentration in the media. By inductively coupled plasma mass spectrometry (ICP-MS) we determined that the iron concentration in spent media was 0.2 (± 0.007) μM, which was significantly lower than in the fresh media (0.39 ± 0.007 μM, Supplementary Figure S5 ). Given the well-known effect of iron depletion on upregulation of PrfA activity (Conte et al., 2000; McLaughlin et al., 2011), this suggests that depletion of iron levels in the spent media may contribute to PrfA-mediated aggregate formation.

Overall, these data demonstrate the aggregation is an adaptive response to factors produced by host cells, resulting in metabolic changes in L. monocytogenes and upregulation of PrfA virulence, which drives aggregation and facilitates intracellular invasion.

We used L. monocytogenes EDGe: InlA^m (Wollert et al., 2007), suitable for animal studies; this mutation is not affecting the binding with MET investigated herein. Wildtype and mutant L. monocytogenes EDGe: InlA^m were grown at 37°C shaking (200 rpm) in tryptone soya broth (TSB) (Oxoid) or with additional 1.5% (w/v) agar (Oxoid) unless otherwise stated. L. monocytogenes mid-log (OD₆₀₀ 0.5-0.6) aliquots stored at -80 °C in PBS glycerol (15% v/v) were used for infections. Escherichia coli DH5α grown in Luria-Bertani broth (LB) was used for cloning and when required media was supplemented with antibiotic (Erythromycin 5 μg/ml, Chloramphenicol 7 μg/ml).

Plasmids were electroporated into L. monocytogenes to generate fluorescently tagged and fluorescent reporter strains (see Tables 1 and 2 for plasmids and strains used in the study). Chromosomal integration of integrative plasmids at the tRNA^Arg-attBB site was confirmed by PCR as described previously (Lauer et al., 2002). Correct fluorescence of strains was confirmed by microscopy. L. monocytogenes ΔactA and actA-ΔC mutant strains were constructed using the temperature sensitive shuttle plasmid pAUL-A as described previously (Wang et al., 2015). Viability of stored bacteria was confirmed by routine measurements of colony forming units of the inoculum.

HeLa cells were grown in Dulbecco’s modified eagle media (DMEM, Sigma) supplemented with 10% (v/v) foetal calf serum (FCS, Gibco) and 1% (v/v) non- essential amino acids (Gibco) at 37°C 5% CO₂ (v/v). Cells were maintained by sub- cultivating at a ratio between 1:2 to 1:6 split 2-3 times a week depending on cell density. HUVEC cells (Thermo Fisher) were recovered from cryopreservation and grown in Human Large Vessel Endothelial Cell Basal Medium (HLVEM, Thermo Fisher) supplemented with Large Vessel Endothelial Supplement (Thermo Fisher) at 37°C 5% CO₂ (v/v) for no more than 16 doublings after the initial recovery, as per the provider recommendations. Infections are performed in serum-free media to avoid opsonisation of bacteria preventing host cell binding.

HeLa cells were seeded in DMEM supplemented with 10% FCS at a density of 2.0 x 10⁵ in a 6 well plate and incubated overnight at 37°C 5% CO₂ (v/v). Subsequently media was aspirated, and the cells were washed in PBS. Serum-free DMEM containing 1.0 x 10⁷ CFU/ml L. monocytogenes was added at a MOI of 20. The plates were incubated for 2 hours at 37°C 5% CO₂ (v/v). After washing in 2 ml of PBS twice, 2 ml of serum-free DMEM was added except the wells being analysed for the first time point. Remaining plates were then incubated with gentamicin (at 10 μg/ml final concentration) until the required time point. Bacteria were retrieved and enumerated by washing the wells in 1 ml of PBS three times and then adding 1 ml of PBS with 0.5% (v/v) triton to lyse the host cells for 2 mins. This solution was then mixed and transferred to a 1.5 ml Eppendorf tube. A serial dilution up to a dilution factor of 10^-5 was performed in PBS and 20 μl of all the dilutions was transferred in triplicate to a TSB agar plate. Plates were left to dry and then transferred to an incubator and incubated overnight at 37°C. Colonies were counted and multiplied by the dilution factor to retrieve the CFU/ml for each well.

Cells were seeded in media (DMEM for HeLa and HLVNM for HUVEC) supplemented with 10% (v/v) FCS into a 35 mm imaging dish (Griner Bio One) at 2.0 x 10⁵ and incubated overnight. Cells were infected with L. monocytogenes prepared in 3 ml of pre-warmed serum-free media at MOI 20 (or 5 for HUVEC cells) for 2 h and washed three times with PBS prior to the addition of gentamicin containing serum-free media (at final concentration 10 μg/ml). Infections were imaged by live-cell time-lapse microscopy using a Zeiss LSM880 microscope using 40x (1.4NA) oil immersion objective and 488 nm laser for GFP, and 561 nm for dsRed fluorophores. Z-stack function was used for imaging, taking 4 slices across 12.24 μm between the first and last slice, every 3 mins per location. Zeiss Zen Black software was used to generate a maximum intensity projection single plane image from the Z-stacks. Replicative invasion events refer to individual host cells exhibiting increasing number of PactA-GFP-expressing bacteria inferred by visually tracking host cells and bacteria in the captured live-cell microscopy movies over the duration of the experiment. Initial infection events were typically spatially separated, allowing robust identification of secondary infections associated with cell-to-cell spread and division of infected host cells. Secondary infections were not considered in our analyses. In some bacteria the constitutive dsRed signal (expressed from non-integrative plasmid) may disappear due to rapid proliferation and presumably insufficient production of the fluorescent marker. Probability of replication was calculated as the number of successful replicative invasions in relation to the total number of host cells present in the image at the first time point. Aggregate size and intensity were measured by using the ROI (region of interest) manager in FIJI (Schindelin et al., 2012). Individual aggregates were segmented in the image by drawing around the edge using the freehand tool. These were recorded in the ROI manager and measured for their size or intensity, when relevant.

2.0 x 10⁵ HeLa cells were seeded into an 35mm imaging dish (Greiner Bio One) in DMEM supplemented with FCS and L-glutamine and incubated in 37°C 5% CO₂ (v/v). The next day cells were washed with PBS and supplemented with serum-free DMEM (no FCS). The following day, 1 ml of spent media (at 37°C) from above cultures was added to imaging dish and the bacterial inoculum was added to a final cell density of 1.0 x 10⁷ CFU/ml. In some experiments a control was also performed with fresh serum free DMEM media. The dish was sealed with parafilm and either incubated at 37°C 5% CO₂ (v/v) in a cell culture incubator or on the incubation unit of LSM 880 while imaged. A time series was taken every 15 seconds in an individual tile with brightfield, 488 nm and 561 nm lasers using the z-stack function taking 4 slices with 12.24 μM distance between the first and last slice. In non-fluorescent strains the 488 nm laser and brightfield was used for analysis. Aggregate size (in μm²) were measured by using the ROI manager in FIJI, individual bacteria were segmented in the image by drawing around the edge using the freehand tool. Single cell virulence expression was obtained using mean intensity of the fluorescent reporter signal per segmented object.

HeLa cells were cultured in an imaging dish to a density of 2.0 x 10⁵ cells and incubated overnight. Cells were washed with PBS and 1 ml of the media containing bacteria was added to the imaging dish at the desired MOI and incubated at 37°C 5% CO₂ (v/v) for 2 hours. Media was removed and the cells were washed twice with PBS. 1 ml of 4% (v/v) paraformaldehyde (PFA) suspended in PBS was added and the dishes at room temperature for 20 mins. The PFA was removed, and the cells were washed 3x in PBS. The cells were incubated with blocking buffer (1% (v/v) BSA, 22.52 mg/ml glycine, 0.1% (v/v) tween-20) for 30 mins and washed with PBS three times. 300 μl of diluted (1:500 from supplier stock - 8 ug/ml) rabbit anti-Listeria antibody (Abcam, ab35132) was added and incubated at room temperature for 20 mins. Dishes were washed 3x in PBS and then secondary antibody donkey anti-rabbit IgG Brilliant Violet 421 (Biolegend) was added and incubated at room temperature for 20 mins and then washed 3x in PBS. Images were taken with a Zeiss 880 confocal microscope with 40x (1.4 NA) immersion oil objective using appropriate lasers. Z-stack images were taken with 5 slices and 12.24 μm between the first and last slices. Images were analysed in FIJI by manually counting the number of bacteria in each aggregate or using the ROI manager to obtain aggregate size (see Supplementary Figure S3B for correlation between aggregate size and number of bacteria). The size (in μm²) of each external/internal aggregate, number of bacteria, total number of aggregates and total number of host cells were recorded. Internalised bacteria were recorded as those objects that were excited by the 561 nm laser (dsRed) but were not excited by the 405 nm laser settings, i.e., they had been internalised as not stained with the anti-Listeria antibody. The size of each external/internal aggregate, number of bacteria, total number of aggregates and total number of host cells were recorded. The number of associated bacteria per cell was calculated by taking the total number of associated bacteria and dividing the total number of host cells. Probability of invasion was calculated by dividing the number of intracellular bacteria by the total number of associated bacteria.

HeLa cells cultured in an imaging dish to a density of 2.0 x 10⁵ were incubated overnight. Cells were washed twice with PBS and 1 ml of 4% paraformaldehyde (v/v) suspended in PBS was added and incubated at room temperature for 20 mins. The PFA was removed, and the cells were washed 3x in PBS. The cells were incubated with blocking buffer (1% (w/v) BSA, 22.52 mg/ml glycine, 0.1% (v/v) tween-20) for 30 mins and washed with PBS three times. 5 μg/ml of primary goat Anti-MET antibody (Abcam) suspended in blocking buffer was added to the dish and left for 1 min. The cells were washed 3x with PBS by adding the PBS and leaving it for 20 mins before removing and changing the PBS. After the washes the secondary antibody 10 μg/ml Alexafluor 488 anti-Goat (Abcam) was added and incubated at room temperature for 30 mins. The cells were washed 3x with PBS for 20 mins per wash. DAPI Vectashield was added to the dish to preserve fluorescence and to stain cell nuclei with DAPI. The cells were imaged using a Zeiss 880 confocal microscope and a 40x objective (1.4NA) oil immersion lens. 488 nm laser was used for excitation settings and background signal was removed by calibrating against a negative control with no staining. To deplete the cell surface of MET, 2 μg/ml of primary antibody was added to serum free DMEM and this was incubated on the cells for 1 hour at 37°C 5% CO₂ (v/v). After MET depletion L. monocytogenes was added to the host cells at a MOI of 20 in serum free media and incubated for 2 hours at 37°C 5% CO₂ (v/v) before fixing and imaging as described above. ROI manager in FIJI was used to quantify MET staining associated with bacteria, using freehand tool to segment bacteria and measure GFP intensity in the corresponding region.

RNA extraction was performed using Purelink RNA Extraction Kit (Invitrogen) and Lysing Matrix E tubes (MP Biomedicals. Fresh lysozyme solution (10 mM Tris-HCl, 0.1 mM EDTA, 10 mg/ml lysozyme), 10% SDS (v/v) in RNAse-free water and lysis buffer (supplemented with 10 μl of 2-mercaptoethanol per 1 ml of buffer) were prepared. Bacteria were pelleted by centrifuging at 13000 x g for 3 mins and the supernatant was discarded. 100 μl of lysozyme solution was added and the pellet was resuspended by vortex. 0.5 μl of 10% (v/v) SDS solution was added and mixed via vortex. 350 μl of lysis buffer was added and mixed by vortex. The lysate was transferred to a Lysing Matrix E tube and homogenised in a rotor stator homogeniser (FastPrep FP120) at 6.5 speed setting for 45 seconds. The tube was centrifuged at 2600 x g for 5 mins and the supernatant was transferred to a RNAse-free microcentrifuge tube. 250 μl of 100% (v/v) ethanol was added to the tube and mixed by vortex to remove precipitate. The sample was transferred to a kit spin cartridge and centrifuged at 12000 x g for 15 seconds to bind RNA to column. Flow-through was discarded and 700 μl of Wash Buffer 1 was added to the column. The column was centrifuged at 12000 x g for 15 sec and both the flow through and collection tube were discarded. The column was placed in a new collection tube and 500 μl of Wash Buffer 2 was added to the column. This was centrifuged at 12000 x g for 15 sec, and flow-through was discarded. The wash step with Wash Buffer 2 was repeated and after flow-through was discarded again the column was centrifuged at 12000 x g to remove any residual wash buffer. The column was transferred to a recovery tube (RNAse free 1.5ml tube provided by kit) and 50 μl of RNAse free water was carefully added directly to the column membrane. This was incubated for 1 min at room temperature to elute RNA and then centrifuged at 12000 x g for 2 mins. All downstream work with RNA after this point was performed on ice.

Genomic DNA (gDNA) was removed using TURBO DNAse kit. 0.1 volume of 10x TURBO DNAse buffer and 1 μl of TURBO DNAse was added the RNA samples and gently mixed. The samples were incubated in a 37°C water bath for 30 mins. The solution was mixed by vortex and 2 μl of DNAse inactivation reagent was added. The samples were incubated for 5 mins at room temperature and flicked 2-3 times during incubation to keep inactivation reagent suspended throughout the mixture. Samples were centrifuged at 10,000 x g for 90 sec and the supernatants were transferred to a fresh RNAse free collection tube. Samples were frozen at -80°C until ready for sequencing. RNA integrity, concentration, and presence of gDNA using an Agilent Tapestation System.

RNA-Seq was performed by the Genomic Technology Core Facility. A total 16 samples across 4 different conditions (media vs. genotype) were prepared in two independent experiments performed (and sequenced) at different times (8 samples each including two independent replicates). A ribosomal Stranded Total RNA Prep with Ribo-Zero Plus depletion kit (Illumina) was used with custom probes (Integrated DNA Technologies) against the L. monocytogenes ribosomal RNA subunits. These were combined with the standard depletion probes (DP1) at the ‘Hybridise Probes’ step of the protocol, as per manufacturer’s instructions. An RNA sample prep kit (Illumina Stranded Total RNA Prep) was used to fragment and denature RNA, synthesise first and second strands, adenylate 3’ prime ends, ligate anchors, clean up fragments and amplify and clean up the library. Additional probes to efficiently deplete the rRNA from L. monocytogenes were added during the probe hybridization step on recommendation from in silico analysis performed by Illumina. Libraries were added to an equimolar pool which was quantified by qPCR (KAPA). The pool was then denatured and loaded onto one lane of an SP NovaSeq 6000 flowcell using the XP workflow, at 200pM with 1% (v/v) PhiX spiked in. The NovaSeq 6000 was run in XP mode with (sequencing lengths: Read 1: 59bp, i7 index: 10bp, i5 index: 10bp, Read 2: 59bp). Unmapped paired-end sequences were tested by FastQC. Sequence adapters were removed, and reads were quality trimmed using Trimmomatic_0.39 (Bolger et al., 2014). The reads were mapped against the reference L. monocytogenes EGDe Genome and annotation. Counts per gene were calculated using featureCounts (subread_2.0.0) (Liao et al., 2013). Normalisation, Principal Components Analysis, and differential expression was calculated with DESeq2_1.36.0 (Love et al., 2014).

The levels of iron were determined by ICP-MS as described before (Corbett et al., 2011). 1 ml of fresh and spent media samples were diluted into 2% nitric acid and analysed in three replicate experiments.

Statistical analysis was performed using GraphPad Prism 9 software. Data were tested for normal distribution using Shapiro-Wilk test. Two-sample comparisons were conducted using paired t-test for data that was normally distributed, otherwise non-parametric Mann Whitney test was used. Multiple comparisons were analysed using ANOVA for normally distributed data, otherwise Kruskal-Wallis ANOVA was used.