E. histolytica generates significantly higher traction force on fibronectin-coated micropillars compared to uncoated ones.

Polydimethylsiloxane (PDMS) micropillar-array is an effective system to study cellular traction forces across a range of physiological stiffnesses, including those found in gut tissues encountered by E. histolytica trophozoites due to the excellent biocompatibility and high structural flexibility of PDMS [32–34]. The elastic PDMS micropillars provide a surface for cells to adhere to and can be bent by the cells. Tracking cell movements and pillar displacements with live-cell imaging enables the analysis of cell motility and cellular forces (Fig 1A, see Materials and methods section for details). We coated the surface micropillar-arrays with human fibronectin, as interaction with the ECM, particularly fibronectin, plays a paramount role in E. histolytica invasion and virulence [35,36]. Indeed, histology slides from the Human Protein Atlas (HPA, www.proteinatlas.org) [37] revealed that fibronectin is abundant in human intestine, with particularly prominent expression in the subepithelial basement membrane and lamina propria of mucosa (S1 Fig).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. E. histolytica generates significantly larger traction forces on fibronectin-coated micropillars compared to uncoated ones.

(A) Schematic diagram showing how a trophozoite interacts with the micropillar-array surface and thereby generating traction force, leading to the deflection of micropillars. The traction force as a function of F(t) is calculated by multiplying the spring constant of a micropillar k_pillar and micropillar displacement over time δ(t). The cartoon illustration of the E. histolytica trophozoite was adopted and modified from Servier Medical Art (smart.servier.com), licensed under CC BY 4.0. (B) Representative graphs showing the temporal changing of traction force on a micropillar as a trophozoite passing on a micropillar with (left) or without (right) fibronectin (FN) coating. The grey curve indicates background noise of micropillars untouched by the cells. (C) The maximum traction force exerted by E. histolytica on micropillars was compared between fibronectin-coated (n = 44 from 9 cells) and uncoated micropillars (n = 45 from 9 cells) using t-test. (D) The maximum traction force exerted by E. histolytica trophozoites on fibronectin-coated micropillars was analyzed between DMSO-treated controls and those treated with the cysteine protease inhibitor E64d, metronidazole (MNZ), or the F-actin polymerization inhibitor Cytochalasin D (CytD), using the Kruskal-Wallis test followed by post-hoc pairwise comparisons. The analyses included 45 micropillars from 9 cells for each treatment condition, with 5 pillars analyzed per cell. * P < 0.05, *** P < 0.001, **** P < 0.0001.

https://doi.org/10.1371/journal.ppat.1012618.g001

We have tested several arrays of micropillars with the same diameter, 2 μm, and different heights including 5.3, 9.4, and 13.2 μm, corresponding to rigidities (spring constant, k_pillar) of 31, 6 and 2 pN/nm, respectively, calculated by Euler-Bernoulli beam theory [38]. Consequently, only the micropillar-array of height 5.3 μm was employed in this study, since E. histolytica generated forces that led to very large micropillars deflections on softer/higher micropillars-arrays (9.4 and 13.2 μm height), resulting in tracking difficulties for later analyses. The effective elastic modulus of the micropillars-array of height 5.3 μm is approximately 22.7 kPa, which is within the human physiological range [39].

To assess the biomechanical relevance of this system, we first compared the traction forces applied by E. histolytica trophozoites on pillars with and without fibronectin coating. Curves of traction force as a function of time F(t) of specific micropillars were plotted showing the magnitude and characteristics of cellular force (Fig 1B). For comparisons between different conditions, we used the maximum force value from each curve, representing the highest traction force exerted by the cell on the corresponding micropillar. The results show that E. histolytica trophozoites apply significantly greater traction forces on fibronectin-coated micropillars compared to uncoated ones. (Fig 1B and 1C).

Analyses of the motility properties of E. histolytica on the micropillars—including measurements of the Directionality Ratio, Average Speed, Maximum Euclidean Displacement (MED) Speed, and Normalized Area Coverage Rate (see definitions of these quantities and details on cell tracking and segmentation for live-imaging videos in the Materials and methods section)—revealed higher motility on fibronectin-coated compared to uncoated micropillars (S2 Fig).

Next, we employed three inhibitors and/or medications for E. histolytica that are expected to affect cellular force generation on fibronectin-coated micropillars by influencing cell adhesion and the actin cytoskeleton. The treatment concentrations for E. histolytica were determined based on published literature [40–42]. Only visibly active migrating cells observed in the live-imaging videos were included in the analysis. We applied the F-actin polymerization inhibitor Cytochalasin D (CytD) since traction force generation depends on F-actin cytoskeleton contraction and adhesion formation on ECM [14–16]. We used the cysteine protease inhibitor E64d since E64 derivatives can significantly decrease the adhesion of E. histolytica to ECM proteins [43]. We also applied the antibiotic metronidazole (MNZ), a drug currently used to treat amebiasis, as its treatment induces the formation of oxidized proteins in E. histolytica, including cytoskeletal proteins such as actin. This oxidative stress not only impairs the formation of F-actin [42], but also reduces the adhesion of E. histolytica to substrate [44].

Traction forces exerted by E. histolytica on fibronectin-coated micropillars significantly decreased when treated with these inhibitors (Fig 1D). Taken together, depriving the ECM component fibronectin from the substrate, preventing F-actin polymerization (CytD), suppressing cysteine protease (E64d), and inducing oxidative stress (MNZ), all significantly reduce traction forces of E. histolytica on the external substrate.

Exposure to enteropathogenic E. coli led to significant alterations in both motility properties and cellular contraction force of E. histolytica.

E. histolytica resides in a bacteria-rich environment within the intestine, where it not only interacts with but also feeds on various bacterial species, which in turn can modulate its virulence [45,46]. We performed cytopathic assays to compare the ability of E. histolytica trophozoites to destroy a HeLa cell monolayer across different groups after a short co-incubation period (30 minutes) with various E. coli strains, including enteropathogenic O55 and nonpathogenic K12. Control conditions included WT (wild type, no bacterial exposure), WT co-incubated with paraformaldehyde (PFA)-fixed (dead) O55, as well as seeding O55 alone onto the monolayer without trophozoites. The results confirmed that co-incubation of trophozoites with live E. coli resulted in a greater percentage of destruction of a HeLa cell monolayer, with the enteropathogenic O55 strain showing the highest ability in enhancing the parasite’s destructive effect (Fig 2).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Cytopathic assay of E. histolytica trophozoites in different conditions following short period of E. coli exposure.

The cytopathic behavior of E. histolytica trophozoites and E. coli in different combinations, including WT (wild type, no bacterial exposure) trophozoites (n = 7), enteropathogenic E. coli O55 only (n = 10), as well as trophozoites incubated with enteropathogenic E. coli O55 (n = 10), dead E. coli O55 (n = 7), and nonpathogenic E. coli K12 (n = 7), was assessed and compared based on their ability to destroy a monolayer of HeLa cells. Statistical analyses, one-way ANOVA with post-hoc pairwise comparisons. * p < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001.

https://doi.org/10.1371/journal.ppat.1012618.g002

To better replicate the conditions E. histolytica trophozoites encounter during the invasion along fibronectin-rich basement membrane beneath the human intestine epithelium, as well as the further invasion to the lamina propria and even deeper structures (S1 Fig), our subsequent experiments were all conducted using human fibronectin-coated micropillars.

To investigate the impact of bacterial exposure on the invasiveness of E. histolytica, which is partly manifested through changes in motility, we pre-incubated the trophozoites with E. coli (enteropathogenic O55 or nonpathogenic K12) at 37°C for 30 minutes, using the same conditions as the previously described cytopathic assay. This was followed by live imaging using the micropillar system. We then analyzed and compared the motility features of trophozoites under three conditions: 1) WT control (no bacterial exposure), 2) incubation with E. coli O55, and 3) incubation with E. coli K12. Additional control experiments included trophozoite incubation with dead E. coli O55, as well as with fibroblasts, which are abundant and widely distributed stromal cells throughout the intestinal layers, making them highly likely to interact with trophozoites.

To obtain accurate positional, area, and morphological information of cells, we developed and employed a two-phase artificial intelligence (AI)-based tracking and segmentation workflow to track and segment cells in live-imaging videos. This method effectively ensures accuracy and avoids interference from complex backgrounds.

Analyses of movement trajectories revealed that trophozoites exposed to bacteria migrated significantly farther than the control groups, with those incubated with enteropathogenic E. coli O55 exhibiting the highest speed and covering the greatest distance in the shortest time (Fig 3A and S3 Fig, see also S1 and S2 Videos). The median recorded durations were 1030, 324, 560, 640, 740 seconds for WT control, WT + O55, WT + K12, WT + dead O55, and WT + fibroblasts, respectively. Further quantitative analyses indicated that the migration Average Speed, Maximum Euclidean Displacement Speed (MED speed, defined as Maximum Euclidean Displacement divided by the corresponding time), Directionality Ratio, and Normalized Area Coverage Rate (by time and cell size) of trophozoites incubated with live E. coli O55 were all significantly higher than the other groups (Fig 3B–F). The motility characteristics of trophozoites incubated with E. coli K12 were intermediate between the control and those incubated with E. coli O55. There was no significant change in E. histolytica motility when co-incubated with fibroblasts compared to the WT control group.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Significant differences in motility properties of E. histolytica following short period of E. coli exposure.

(A) Trajectory plots of trophozoite migration under different conditions: WT control (n = 82), incubation with enteropathogenic E. coli O55 (n = 141), incubation with nonpathogenic E. coli K12 (n = 41), incubation with dead E. coli O55 (n = 79), as well as incubation with fibroblasts (n = 37). Note that WT + O55 group spreads and occupies the entire plot within a relatively short time (red and yellow), whereas the other groups, even over a prolonged period (cyan, blue, and purple), remain confined to a limited area, clearly indicating that the WT + O55 group reaches a greater distance in less time. (B-E) Quantitative comparison of the migration Directionality Ratio (B), Average Speed (C), Maximum Euclidean Displacement (MED) Speed (D), and Normalized Area Coverage Rate (by time and cell size) (E) of E. histolytica under the different conditions described in (A), with Kruskal-Wallis test and post-hoc pairwise comparisons. (F) The schematic diagram illustrates Accumulated Distance, Euclidean Distance, MED, and Area Coverage of cells (with cell overlapping regions not counted multiple times) along with their corresponding time periods. The parameters in Fig B-E are derived from these calculations (details are in Materials and methods section). (G) The relationship between mean square displacements (MSD) and the corresponding time (seconds) was plotted on a logarithmic scale for E. histolytica locomotion under the different conditions. The solid lines and shaded areas represent the linear fit for each group and the corresponding 95% confidence interval. The slopes (α) and standard error of the mean (SEM) from the linear fits are summarized on the right. (F) Comparison of the slopes (α) for each individual cell/plot under the different conditions, with Kruskal-Wallis test with post-hoc pairwise comparisons Notably, the average slope α of O55 group is the highest, indicating that it exhibits the strongest superdiffusive movement pattern. * P < 0.05, ** P < 0.01, *** P < 0.001.

https://doi.org/10.1371/journal.ppat.1012618.g003

To better understand the cell migration mode, we calculated the mean square displacements (MSD) of the cells, a well-established method in cell migration analysis [47,48], including in E. histolytica studies [49,50]. MSD quantifies how far a cell travels from its starting point over time, reflecting migration efficiency. It provides insights into both speed and directional persistence and is typically shown in a log(MSD)-log(time) plot [47]. The slope of this plot, α, indicates the motion type: α = 2 for ballistic motion (persistent in direction), α > 1 for superdiffusion (e.g., Lévy walk, Correlated Random Walk, Fractional Brownian Motion), α = 1 for normal diffusion (Brownian motion/unbiased random movement), and α < 1 for subdiffusion (restricted movement) [47–49,51–53].

Analyses of the averaged log(MSD)-log(time) plots across all cells (Fig 3G), along with the slope α-values for individual cells from different conditions (Fig 3H), demonstrated that trophozoites incubated with E. coli O55 exhibited the highest α-value (α = 1.852 ± 0.009), followed by those incubated with E. coli K12 (α = 1.575 ± 0.055) and the control groups (e.g., α = 1.191 ± 0.039 for WT control group). This suggests that in the control groups, E. histolytica trophozoites exhibit movement close to normal diffusion or Brownian motion, whereas those incubated with E. coli O55 display enhanced superdiffusion, characterized by significantly greater directionality and speed. Notably, a subset of cells (50 out of 141) exhibited α values > 1.900, indicating a tendency toward ballistic motion. The potential contribution of stage drift during imaging, which could have artificially skewed the diffusive motion toward apparent superdiffusion, was ruled out, as it was at least two orders of magnitude smaller than the cell movements (S4 Fig).

To explore whether cellular traction forces can be affected following E. coli exposure, we further analyzed and compared the traction forces on micropillars by E. histolytica under the different conditions mentioned above: WT trophozoite control, as well as trophozoites incubated with enteropathogenic E. coli O55, and with nonpathogenic E. coli K12 (Fig 4A and 4B). In a separate independent experiment, we additionally included comparisons of the traction forces of trophozoites incubated with dead E. coli O55 and with fibroblasts (Fig 4C). This revealed that trophozoites incubated with live E. coli O55 exhibited the lowest traction forces, followed by those incubated with E. coli K12, whereas the control groups (including WT alone, WT with dead E. coli O55, and WT with fibroblast co-incubation) displayed higher traction forces on the micropillar-arrays (Fig 4).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Significant difference of cellular traction force on micropillars of E. histolytica following short period of E. coli exposure.

(A) Representative graphs showing the changing in traction force exerted on micropillars over time for control trophozoites (left), trophozoites incubated with enteropathogenic E. coli O55 (middle), and trophozoites incubated with nonpathogenic E. coli K12 (right), during a trophozoite passing on a micropillar. The grey curve indicates background noise of micropillars untouched by the cells. (B-C) Quantitative comparison of the max traction forces on micropillars exerted by E. histolytica between the different conditions: (B) WT control (n = 44), E. coli O55-incubated (n = 43), and E. coli K12-incubated (n = 44) trophozoites, as well as (C) WT control (n = 45), E. coli O55-incubated (n = 45), dead O55-incubated (n = 45), and fibroblast-incubated (n = 44) trophozoites. Nine cells were analyzed per condition, with 4–5 micropillars assessed per cell. Statistics, one-way ANOVA with post-hoc pairwise comparisons. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001.

https://doi.org/10.1371/journal.ppat.1012618.g004

The trend in cellular traction forces and that in motility characteristics across the different E. histolytica-E. coli co-incubation conditions were opposite, prompting further consideration of the relationship between cellular traction forces and motility. In the group without exposure to E. coli, the cells exhibited non-directional exploratory behavior, actively extending pseudopods to tug on the surrounding pillars. At the same time, the cells seemed to exhibit greater adhesion, resulting in more significant deflection of micropillars (S3 Video). However, when exposed to enteropathogenic E. coli O55, the cells transitioned to an enhanced superdiffusive pattern (rapid and goal-directed), with a reduction in randomly exploring behavior and displaying noticeably smaller deflections of micropillars (S4 Video).

To illustrate the exact correlation between cell motility and traction force quantitively, linear regressions and Spearman correlations were performed for cells from WT and WT-E. coli O55 co-incubation conditions. This showed that all the four cell motility indicators (Directionality Ratio, Average Speed, MED Speed, and Normalized Area Coverage Rate) were negatively correlated with traction force (Fig 5).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Correlation between cell motility indicators and traction force in WT control and enteropathogenic E. coli O55-incubated E. histolytica.

(A) Directionality Ratio, (B) Average Speed, (C) MED Speed, (D) Normalized Area Coverage Rate (by time and cell size). n = 16 in WT group (blue) and 15 in WT + O55 group (red).

https://doi.org/10.1371/journal.ppat.1012618.g005

Exposure to enteropathogenic E. coli altered the F-actin morphology and spatial localization in E. histolytica

Given that co-incubation of enteropathogenic E. coli led to remarkable variations in the motility and cellular traction forces of E. histolytica, we next sought to investigate whether the actin cytoskeleton plays a crucial role in explaining the underlying mechanism, as it is essential for both motility and cellular traction force. To explore how E. coli exposure affects F-actin organization in E. histolytica, WT control trophozoite, trophozoites incubated with E. coli O55, and trophozoite incubated with E. coli K12 were seeded on fibronectin-coated micropillars and incubated at 37°C for one hour, followed by fixation and subsequent staining with AF-488 Phalloidin (specifically binding to F-actin with high affinity). Notably, all images scanned by confocal microscopy were focused on the plane of the micropillars’ top surface. Consequently, a modest yet statistically significant reduction of F-actin intensity in E. histolytica following E. coli exposure was observed, with no significant difference found between trophozoites incubated with E. coli O55 and E. coli K12 (Fig 6A and 6C).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Alteration of F-actin morphology in E. histolytica following E. coli short-period exposure.

(A) Quantitative comparison of Phalloidin intensity in WT control (n = 40), enteropathogenic E. coli O55-incubated (n = 41), and E. coli K12-incubated (n = 40) E. histolytica trophozoites. (B) Comparison of the proportion of F-actin-polarized cells in control (n = 160), E. coli O55-incubated (n = 202), and E. coli K12-incubated (n = 209) E. histolytica. (C) Representative confocal images (20x objective) showing the Phalloidin signal in E. histolytica on micropillars, both with polarized (partially and completely) and unpolarized samples. (D) Left panel: high-resolution confocal images (63x objective) captured by focusing on the plane of micropillar-array top surface, showing that F-actin formed rings that tightly encircle the corresponding micropillars, as well as filaments linking the rings. Right panel: z-stack reconstruction of F-actin spatial distribution near the micropillars, presenting podosome-like structure. The schematic diagram in the dashed-line frame illustrates the spatial relationship between an F-actin podosome-like structure and its corresponding micropillar. Scale bar in C and D: 10 μm. Statistical analyses in A and B, one-way ANOVA with post-hoc pairwise comparisons. * P < 0.05, ** P < 0.01, **** P < 0.0001.

https://doi.org/10.1371/journal.ppat.1012618.g006

To provide additional evidence, we performed Phalloidin staining on cells plated in ibidi µ-Slide 8 Well plates. F-actin staining of E. histolytica trophozoites on fibronectin-coated substrates showed significantly higher intensity compared to those on uncoated micropillars (S5A and S5E Fig). Moreover, on the fibronectin-coated substrate, E. histolytica trophozoites exposed to E. coli O55 exhibited significantly lower F-actin intensity than those in the control group (S5C and S5E Fig), consistent with the results observed on the micropillars. These findings align with previous results, showing that E. histolytica trophozoites exerted greater traction forces on fibronectin-coated micropillars compared to uncoated ones (Fig 1B and 1C), as well as decreased traction forces when exposed to E. coli (Fig 4). The positive relation between F-actin intensity at micropillar top surface and cellular traction force suggests that elevated traction force on micropillars may be dependent on enhanced F-actin polymerization at pseudopods.

Given that the intracellular local polarization of F-actin is closely correlated with traction force on micropillars [38], we further analyzed the polarity of the Phalloidin signal in different conditions, categorizing it into completely polarized (towards the cell edge), partially polarized (off-center but not reaching the cell edge), and unpolarized (centered) F-actin distributions (Fig 6B and 6C). This analysis showed a significant reduction in the proportion of completely polarized cells when exposed to E. coli O55 and K12 (Fig 6B and 6C). This F-actin polarization pattern is consistent with the Phalloidin intensity comparisons between trophozoites incubated with E. coli.

Finally, we analyzed the spatial relationship between F-actin and its corresponding micropillars using high-resolution confocal scanning combined with the z-stack reconstruction mode. We discovered that F-actin accumulated near the micropillars, forming podosome-like adhesion structures that extend downward, wrapping around the top portion of the micropillars (Fig 6D, right panel). When observed from the plane of the micropillar’s top surface, F-actin formed rings that tightly encircle the corresponding micropillars, as well as filaments linking the rings (Fig 6D, left panel). This arrangement helps concentrate the force for gripping a micropillar, enabling more effective application of traction force to the micropillar [38].

Next, we examined the spatial localization difference of F-actin between three conditions: E. histolytica WT control trophozoite, trophozoites incubated with enteropathogenic E. coli O55, and trophozoite incubated with E. coli K12, which might shed light on the mechanism regarding the negative correlation between cell locomotion and traction force on micropillars. It can be observed that without exposure to E. coli, F-actin primarily accumulated at podosomes (Fig 7A and 7C), possibly enabling the cells to grip the micropillars firmly and effectively. Additionally, F-actin tended to polarize toward pseudopods at the cell edge. However, upon exposure to E. coli, a significant amount of F-actin shifted to the upper part of the cytoplasm (Fig 7A and 7B). Consequently, the F-actin signal at the podosomes corresponding to the micropillars weakened, resulting in shorter and less prominent adhesions.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. The spatial localization difference of F-actin between following E. coli short-period exposure.

(A) Z-stack representative images showing the spatial localization differences of F-actin among three conditions: E. histolytica WT control trophozoites (n = 54), trophozoites incubated with enteropathogenic E. coli O55 (n = 32), and trophozoites incubated with E. coli K12 (n = 43). (B) Comparison of the height of F-actin structures in the upper part of the cell, distinct from adhesion structures on micropillars, under the different conditions. (C) Comparison of the height of F-actin adhesion structures under the different conditions. Statistical analyses in B and C, one-way ANOVA with post-hoc pairwise comparisons, ** P < 0.01, *** P < 0.001, **** P < 0.0001.

https://doi.org/10.1371/journal.ppat.1012618.g007

Additional analysis based on Phalloidin staining of E. histolytica trophozoites on fibronectin-coated ibidi µ-Slide 8 Well plates revealed that when trophozoites were incubated with E. coli O55, the ratio of basal surface area (determined by confocal F-actin imaging) to the total area in the bright field (non-confocal imaging) was significantly smaller compared to control trophozoites. This suggests that interaction with E. coli O55 reduces the contact surface area with the fibronectin-coated substrate (S5D and S5E Fig).

E. histolytica strains

The E. histolytica strain HM-1:IMSS was graciously supplied by Prof. Samudrala Gourinath from Jawaharlal Nehru University, New Delhi, India. Trophozoites of E. histolytica strain HM-1:IMSS was cultured at 37°C in 13 × 100 mm screw-capped Pyrex glass tubes containing Diamond’s TYI-S-33 medium until they reached the exponential growth phase. Trophozoites were then harvested from the culture tubes using the previously outlined method of tapping followed by centrifugation, as detailed in prior protocols [82].

Bacterial strains

The enteropathogenic E. coli strain TW04062 O55 is from the Thomas S. Whittam STEC Center. The nonpathogenic E. coli strain MG1655 K12 was provided by Sima Yaron from the Department of Biotechnology and Food Engineering at the Technion. E. coli cultures were incubated at 37°C in Luria-Bertani (LB) medium.

Preparation of dead bacteria

A total of 1 × 10^9 bacteria were mixed with 1 mL of 4% formaldehyde (buffered to pH 7.2–7.4) and incubated at room temperature for 30 minutes. The bacteria were then centrifuged, washed once with PBS, followed by three washes with 0.1 M glycine, and finally washed once with TYI medium before resuspension.

Measurement of cytopathic activity

The assessment of trophozoites’ destruction of cultured HeLa cell monolayers was conducted utilizing a previously outlined methodology [45]. In brief, a concentration of 1 × 10^6 trophozoites/mL was incubated with E. coli O55 (live or dead) or E. coli K12 (1 × 10^9/mL) in serum-free Diamond’s TYI-S-33 medium at 37°C for 30 minutes while shaking. Afterwards, most bacteria were removed by washing the trophozoites with TYI medium through centrifugation at 726 × g for three cycles. Then trophozoites (10^5 per well) were then seeded onto and incubated with HeLa cell monolayers (~90% confluency) in 24-well tissue culture plates with serum-free Diamond’s TYI-S-33 medium at 37°C for 60 minutes. The incubation process was stopped by transferring the plates to ice, followed by removal of unattached trophozoites through washing with cold phosphate-buffered saline (PBS). The remaining HeLa cells adhered to the plates were subsequently stained with methylene blue solution (0.1% in 0.1 M borate buffer, pH 8.7). Extraction of the dye from stained cells was performed using 0.1 M HCl, and the color intensity of the extracted dye was measured spectrophotometrically at OD660. For the bacterial-only treatment group, 5 × 10^7 of E. coli O55 were directly seeded onto and incubated with HeLa cell monolayers, with all other conditions kept consistent with the other groups.

Micropillar-array system and traction force calculation

PDMS micropillar-arrays were employed in this study, which allowed us to detect and measure the traction force that cells applied on the micropillars [32,38]. Based on the microscope time-series imaging of micropillars within a particular period, we spatiotemporally tracked and recorded the location of every micropillar, which was then calculated to micropillar displacement (δ) and then transferred to traction force (F) [32] (Fig 1A).

The rigidity of PDMS micropillar-array is only altered by changing micropillar height, while the cross-sectional area and chemical properties of the micropillars remain constant. The diameter of the circular, micropillars cross-section is two μm, and the center-to-center spacing between adjacent micropillars is four μm.

Tracking of micropillars movements over time was performed with NanoTracking plugin of Fiji/ImageJ (National Institutes of Health), the detailed algorithm of which is described previously [83]. Briefly, it employed a cross-correlation technique that enabled the software to obtain the relative x and y position of every micropillars in every frame of the time-series movie in nanometer-level. We only selected for analysis the cell migration-associated micropillars that were not in contact with any cells (zero position) at the beginning of the movie. Micropillar displacement (δ), which is caused by traction force (F) when cell edge/body contact the micropillars, at each frame, is calculated based on the distance from current to the zero position. Traction force (F) is calculated by multiplying micropillar displacement (δ) by the spring constant of the corresponding pillar height (Fig 1A). Micropillar arrays with heights of 5.3, 9.4, and 13.2 μm have external rigidities (spring constant, k_pillar) of 2, 6, and 31 pN/nm, respectively, as calculated using the Euler-Bernoulli beam theory [38]. The curve of traction force as a function of time F(t) of a micropillar was plotted using MATLAB (MathWorks, v2018b) with Smooth function, using a smoothing parameter of 0.01.

PDMS micropillars-array fabrication and fibronectin coating

Pillar fabrication was achieved by pouring PDMS, which was mixed at 10:1 with its curing agent (SYLGARD 184, Dow Corning), over silicon molds with holes at fixed depths and geometric distance. The mold was then put upside down, onto a glass-bottom 35-mm dish (D35-20–0-N, Cellvis), which was followed by incubation at 60°C for 12 hours. After cooling to room temperature, the mold was peeled off while both the mold and the PDMS pillar cast were immersed in absolute ethanol to prevent micropillars from collapsing. Afterwards, the ethanol was replaced by PBS, and then human plasma full-length fibronectin (FC010–10MG, Merck) was added to the dish, making a coating solution with final concentration of 10 μg/mL for a one-hour surface-coating at 37°C [32]. Finally, residual fibronectin was washed away by four serial replacements with serum-free Diamond’s TYI-S-33 medium. For the experiments using micropillars without fibronectin coating, absolute ethanol was replaced by PBS, which was then replaced by serum-free Diamond’s TYI-S-33 medium.

Cell seeding on micropillar-arrays and time-series imaging

A total of 1 × 10^5 E. histolytica trophozoites were collected through tapping and subsequent centrifugation. These trophozoites were then seeded onto a fibronectin-coated (or non-coated) micropillar-array in a 35 mm glass-bottom dish filled with serum-free Diamond’s TYI-S-33 medium. The system was incubated in a heating chamber at 37°C located on Zeiss LSM800 confocal microscope. Regarding E. histolytica trophozoites treated with bacteria, a concentration of 1 × 10^6 trophozoites/mL was incubated with E. coli O55 (live or dead) or E. coli K12 (1 × 10^9/mL) in serum-free Diamond’s TYI-S-33 medium at 37°C for 30 minutes while shaking. Afterwards, most bacteria were removed by washing the trophozoites with TYI medium through centrifugation at 726 × g for three cycles. The trophozoites were then resuspended in TYI medium, followed by seeding (1 × 10^5 trophozoites) onto a micropillar-array described above. Regarding E. histolytica trophozoites treated with fibroblasts, 1 million cells were pre-seeded per well in a six-well plate for two hours before adding 0.5 million E. histolytica trophozoites. Co-incubation was carried out in serum-free Diamond’s TYI-S-33 medium at 37°C with 5% CO₂ for 30 minutes. The trophozoites were then collected by pipetting, centrifuged, washed once with TYI medium, and resuspended in TYI medium. Subsequently, 1 × 10^5 trophozoites were seeded onto a micropillar array as described above. Bright-field time-lapse imaging for E. histolytica on micropillars was performed using a 20 × objective at 37°C. Images were scanned every four seconds with a time-series mode. Fibroblasts (mouse embryonic fibroblasts) used in this study is a gift from M. Sheetz (MBI Singapore and the University of Texas Medical Branch) [84].

Phalloidin staining on micropillars

A concentration of 1 × 10^6 trophozoites/mL was incubated with or without E. coli O55 or E. coli K12 (1 × 10^9/mL) in serum-free Diamond’s TYI-S-33 medium at 37°C for 30 minutes while shaking. Afterwards, most bacteria were removed by washing the trophozoites with TYI medium through centrifugation at 726 × g for three cycles. The trophozoites were then resuspended in TYI medium, followed by seeding onto a micropillar-array and incubated at 37°C for an hour. Then trophozoites were fixed by 4% paraformaldehyde (PFA) for 15 minutes at room temperature. After washing away PFA by PBS, the cells were permeabilized with 0.2% Triton-X for 10 minutes at room temperature. Afterwards, the cells were stained with Alexa Fluor 488 Phalloidin (A12379, Invitrogen) with a dilution of 1:1000 for 15 minutes at room temperature. After another PBS wash, the cells were kept in PBS solution with SlowFade Gold Antifade Mountant (#S36937, Invitrogen), and then imaged by Zeiss LSM800 confocal microscope with a magnification of 20x with Airyscan processing.

Regarding three-dimensional reconstruction of F-actin structures, the cell is imaged with 63x objective scanned with Airyscan and Z-stack mode (0.2 µm interval between neighboring slices with a total thickness around 5 µm), followed by Airyscan processing and deconvolution.

Cell motility analysis

The cell motility analysis begins with converting the time-series imaging in CZI format to 8-bit image stacks (TIF format) in Fiji. The observed experimental domain measures 266.21 × 266.21 μm, with a resolution of 0.144 μm per pixel.

To measure and evaluate stage drift of imaging (S4 Fig), five randomly selected micropillars, unaffected by any disturbances, were used for nano-tracking. Drift in both the x and y directions was calculated for each frame. Within a complete video of approximately 300 frames (4 seconds per frame), the drift is within 250 nm, which is less than 2 pixels, far smaller than the size of the cells and the distance the cells migrate within 300 frames. To note, the average cell size is about 481.88 ± 6.65μm² in area and 24.75 ± 0.17 μm in diameter.

To obtain the mask of every cell in each frame of time-series image stacks, we developed a sophisticated cell segmentation and tracking framework that harnesses the zero-shot generalization capabilities of the Segment Anything Model 2.1 Large (SAM 2.1 Large) (https://ai.meta.com/sam2/). Our approach implements a novel double segmentation methodology designed to overcome persistent challenges in cell segmentation and tracking, including stability issues, contrast-dependent segmentation errors, rapid morphological changes, and efficiency degradation when tracking multiple objects simultaneously.

Our framework employs two sequential segmentation phases for each frame, implemented in Python 3.12 with the SAM 2.1 Large model serving as the foundation for both phases. The first phase utilizes the SAM 2.1 model in high sensitivity mode to detect new objects entering the Field of View (FOV) and to update the positions of previously tracked objects. The second phase employs bounding-box based segmentation in SAM 2.1 to achieve high accuracy cell delineation. This dual approach effectively addresses low contrast conditions and mitigates accuracy loss in cell segmentation, particularly at cell boundaries.

Although cells exhibit rapid morphological changes, their inter-frame movement remains relatively slower, enabling the implementation of similarity-based cell tracking. For comprehensive analysis, we monitor multiple parameters for each cell, including centroid position, intensity, area, and intercellular interactions. These parameters are systematically recorded alongside annotated videos and cell masks to facilitate thorough analysis and validation. The examples of cell tracking with cell contours for all groups is shown in S1 Appendix.

By using the cell masks, the position (x,y) coordinate and area of each cell at every frame was calculated and recorded by scikit-image package (version 0.23.2) in Python 3.12. Trajectory plots for cells in an experimental group are generated by R4.4.0 with ggplot2, with the coordinates of the cells’ starting points transformed to position (0,0).

The time duration of the recorded trajectories for the WT control group had a median of 1030 seconds, with an interquartile range (IQR) of 608–1332 seconds. For the trophozoite incubated with the O55 group, the median was 324 seconds, with an IQR of 240–472 seconds. For the other groups, the median times were as follows: WT on micropillars without fibronectin-coating: 1060 seconds (IQR: 968–1116), WT + K12: 560 seconds (IQR: 360–1152), WT + dead O55: 640 seconds (IQR: 544–858), and WT + fibroblast: 740 seconds (IQR: 664–1160). There is no truncation in the data, as consecutive frames are recorded at 4-second intervals. Recording stops when any part of the cell moves out of the frame. Additionally, we ensured that the extracted cell trajectories did not involve any obvious cell contact or collision.

The motility parameters used in this study are listed below and were calculated in R4.4.0. The schematic diagram in (Fig 3F) illustrates the related definitions.

1. Average Speed: The total accumulated trajectory length divided by time:

(1)

where d_i represents the Euclidean distance between consecutive time points, and T is the total duration of the experiment. To calculate the Average Speed, the coordinates and trajectory of each cell are smoothed using loess with a span of 0.3.

2. Maximum Euclidean Displacement (MED) Speed: The maximum Euclidean displacement divided by the corresponding time interval:

(2)

where (x₀, y₀) is the starting position, (x_t, y_t) is the position corresponding to the MED, and (t - t₀) is the time interval.

3. Directionality Ratio: The ratio of Euclidean distance to accumulated trajectory length:

(3)

where the numerator represents the Euclidean distance from the start to the endpoint, and the denominator is the total accumulated trajectory length. A value closer to 1 indicates more directed movement.

4. Normalized Area Coverage Rate: The accumulated covered area normalized by both time and cell size:

(4)

where Covered Area represents the total accumulated area explored by the cell during its migration (with cell overlapping regions not counted multiple times), T is the total time, and Mean Cell Area is the average cell size. This metric reflects the cell’s efficiency in exploring space relative to its size and time.

Inhibitors treatment

To assess the effects of different inhibitors on E. histolytica traction force and migration, E. histolytica trophozoites were pre-incubated with E64d (10 µM for 16 hours), metronidazole (MNZ, 5 µM for 16 hours), or Cytochalasin D (CytD, 5 µM for 24 hours) at 37°C. Control trophozoites were treated with DMSO. A total of 1 × 10^5 trophozoites were collected via tapping and subsequent centrifugation. These trophozoites were then seeded onto a fibronectin-coated micropillar-array in a glass-bottom dish filled with serum-free Diamond’s TYI-S-33 medium, supplemented with the same concentration of inhibitors as used during pre-incubation, followed by live-imaging on a confocal microscope.

Phalloidin staining on ibidi µ-Slide 8 Well plates

A total of 2 × 10^5 E. histolytica trophozoites were collected via tapping and subsequent centrifugation. These trophozoites were then seeded onto an ibidi µ-Slide 8 Well plate (#80826, ibidi GmbH) pre-coated with or without fibronectin and incubated in serum-free Diamond’s TYI-S-33 medium at 37°C for one hour. Regarding E. histolytica trophozoites treated with bacteria, a concentration of 1 × 10^6 trophozoites/mL was incubated with E. coli O55 or E. coli K12 (1 × 10^9/mL) in serum-free Diamond’s TYI-S-33 medium at 37°C for 30 minutes while shaking. Afterwards, most bacteria were removed by washing the trophozoites with TYI medium through centrifugation at 726 × g for three cycles. The trophozoites were then resuspended in TYI medium, followed by seeding onto an ibidi plate and incubated at 37°C for one hour. Next, the plates were washed three times with PBS to remove residual TYI-S-33 medium. Trophozoites were then fixed with 4% PFA in PBS for 15 minutes at room temperature. Following PFA removal by PBS washing, cells were permeabilized with 0.2% Triton X-100 for 10 minutes at room temperature. Subsequently, the cells were stained with Alexa Fluor 488 Phalloidin (A12379, Invitrogen) with a dilution of 1:1000 for 15 minutes at room temperature. Later, cells were washed with PBS and mounted with SlowFade Gold Antifade Mountant. Imaging was performed using confocal microscope with a 63x objective and Airyscan processing.

Image segmentation and polarization categorization

For the planar confocal imaging for fixed cells, all images included in the same analysis were obtained under identical experimental conditions and scanning parameters. The fluorescence signal range and cell contours were delineated using the SAM model (Segment Anything Model, https://segment-anything.com/) to generate the corresponding masks. These masks were then analyzed in Fiji for area, intensity, and center of mass coordinates.

For the analysis of F-actin adhesion polarization on micropillars, the following criteria were applied: 1) If the adhesion region overlapped with the cell boundary, the cell was categorized as completely polarized. 2) For the remaining cells, if the cell’s center of mass was within the adhesion region, the cell was categorized as unpolarized. 3) If the cell’s center of mass was outside the adhesion region, the cell was categorized as partially polarized.

Statistical analysis

Statistical analyses were conducted utilizing both Prism 9 (Graphpad Software Inc.) and R version 4.4.0. Plots are generated by ggplot2 and ggpubr packages with R version 4.4.0. Unless otherwise specified, significance was assessed using one-way ANOVA for multiple comparisons (with Holm-Sidak method for post-hoc pairwise comparisons) and unpaired two-tailed t-test for two-group comparisons. For data that did not meet normality assumptions, corresponding nonparametric methods, including Kruskal-Wallis (with Dunn’s test for post-hoc pairwise comparisons) and Mann-Whitney U tests, were used. Spearman correlation analyses were performed using ggpubr package. Chi-square test was performed by gmodels and tableone packages.