Gonococcal invasion into epithelial cells depends on both cell polarity and ezrin

Neisseria gonorrhoeae (GC) establishes infection in women from the cervix, lined with heterogeneous epithelial cells from non-polarized stratified at the ectocervix to polarized columnar at the endocervix. We have previously shown that GC differentially colonize and transmigrate across the ecto and endocervical epithelia. However, whether and how GC invade into heterogeneous cervical epithelial cells is unknown. This study examined GC entry of epithelial cells with various properties, using human cervical tissue explant and non-polarized/polarized epithelial cell line models. While adhering to non-polarized and polarized epithelial cells at similar levels, GC invaded into non-polarized more efficiently than polarized epithelial cells. The enhanced GC invasion in non-polarized epithelial cells was associated with increased ezrin phosphorylation, F-actin and ezrin recruitment to GC adherent sites, and the elongation of GC-associated microvilli. Inhibition of ezrin phosphorylation inhibited F-actin and ezrin recruitment and microvilli elongation, leading to a reduction in GC invasion. The reduced GC invasion in polarized epithelial cells was associated with non-muscle myosin II-mediated F-actin disassembly and microvilli denudation at GC adherence sites. Surprisingly, intraepithelial GC were only detected inside epithelial cells shedding from the cervix by immunofluorescence microscopy, but not significantly in the ectocervical and the endocervical regions. We observed similar ezrin and F-actin recruitment in exfoliated cervical epithelial cells but not in those that remained in the ectocervical epithelium, as the luminal layer of ectocervical epithelial cells expressed ten-fold lower levels of ezrin than those beneath. However, GC inoculation induced F-actin reduction and myosin recruitment in the endocervix, similar to what was seen in polarized epithelial cells. Collectively, our results suggest that while GC invade non-polarized epithelial cells through ezrin-driven microvilli elongation, the apical polarization of ezrin and F-actin inhibits GC entry into polarized epithelial cells.


Point-to-point responses to comments PPATHOGENS-D-21-00895 Gonococcal invasion into epithelial cells depends on both cell polarity and ezrin
We would like to thank the Editors and the Reviewers for your detailed comments and suggestions. These comments are very relevant and helpful. The revision based on the comments has greatly improve the quality of our manuscript. The followings describe how we have addressed comments in the revised manuscript. Modifications have been highlighted in the manuscript.

1) analyzing more than one time point in binding and invasion
Based on this suggestion, we have compared gonococcal (WT MS11) adherence and invasion in non-polarized and polarized T84 cells on transwells parallelly at three different time points, 3, 6, and 12 h using the gentamicin-resistant assay. We found that at 3 and 6 h, the numbers of gentamicin-resistant bacteria were higher in non-polarized epithelial cells than in polarized epithelial cells, while the total numbers of epithelial-associated bacteria in non-polarized and polarized T84 were similar. By 12 h, the numbers of both epithelial-associated and gentamicinresistant bacteria in non-polarized and polarized epithelial cells were decreased compared to those at 3 and 6 h, due to bacterial overgrowth, even though unassociated bacteria were washed away at 6 h. This suggests that the effect of the epithelial cell polarity on GC invasion does not depend on time. Therefore, we chose the 6-h for the rest of our study. The new data is included in Fig. 1A-C.
2) using GC constitutively expressing a single Opa in the isogenic delta Opa background for Opa-dependent effects This is an excellent suggestion. In the revised manuscript, we compared the adherence, invasion of WT MS11 Neisseria gonorrhoeae (GC) strain that express phase variable Opa (Opa+) and Opa-deletion strain (∆Opa) with ∆Opa strain expressing OpaH that binds to Carcinoembryonic antigen-related cell adhesion molecules (CEACAMs) and cannot phase vary (OpaCEA) in nonpolarized and polarized T84 cells (Fig. 1D-F). We also compared their impacts on F-actin ( Fig.  2A and 2B) and ezrin ( Fig. 3A and 3B) distribution. Our results show that OpaCEA GC behave similarly as Opa+ GC, displaying higher numbers of epithelial-associated and gentamicinresistant bacteria and higher levels of F-actin and ezrin accumulation at GC adherent sites in both non-polarized and polarized epithelial cells than ∆Opa GC. These new results suggest that Opa can enhance GC adherence and invasion and GC-induced F-actin and ezrin accumulation through binding to CEACAMs.
3) considering that exfoliation may occur as a consequence, not cause of invasion This is an excellent point! In the revised manuscript, we have discussed the possibilities that GC invasion into cervical epithelial cells induces epithelial shedding, and that GC invasion into cervical epithelial cells after epithelial shedding in the discussion section (Line 602-613).

4) rewriting to frame in light of prior literature on this topic
We have extensively revised the manuscript, particularly the introduction and the discussion sections, to put our data in the context of the extensive early studies on GC invasion into epithelial cells and the underlying mechanism. 5) addressing the discrepancy between gentamicin and TEM data, and addressing the biological significance of the small but statistically significant differences reported We used both gentamicin resistance and transmission electron microscopy TEM to confirm the effects of cell polarity and Opa expression on GC invasion. Each of the two methods has its advantages and disadvantages. The gentamicin resistance assay is more quantitative and determines the total number of GC surviving the gentamicin treatment in each transwell containing ~2x10 5 epithelial cells before inoculation with 2x10 6 bacteria. However, this method may overestimate GC invasion levels, as gentamicin may not reach bacteria inside microcolonies, particularly Opa+ GC that form big and tight microcolonies [1]. In contrast, TEM allows distinguishing intra-and extra-cellular bacteria visually. However, TEM can only examine a limited number of epithelial cells and GC. Each TEM image only contains two epithelial cells and 10~30 bacteria at a magnification that can distinguish intra-and extra-cellular bacteria rather than all the epithelial cells and bacteria in a transwell. Thus, we have to quantify TEM results using a different method. We apologize for leaving some details in TEM quantification out, and these details have been included in the results (Line 206-219) and methods sections of the revised manuscript. Overall, the results of TEM analysis support the results of the gentamicin resistance assay that epithelial cell polarization inhibits and Opa expression facilitates GC invasion.
We agree with the reviewer's comment. Indeed, the differences in the F-actin FIR between Opa+ GC and ∆Opa GC are small and not obviously visible, even though the differences are significant when a number of GC microcolonies from three independent experiments are quantified. These small differences suggest that Opa proteins can facilitate but are not essential for GC to induce actin reorganization in non-polarized and polarized T84 cells. This finding is consistent with previous reports that Opa proteins are not essential [2,3] but can facilitate GC invasion into epithelial cells [4][5][6][7]. Notably, the differences in the F-actin FIR between GC inoculated nonpolarized and polarized T84 cells are much bigger than between Opa+ and ∆Opa GC, suggesting a more substantial role of epithelial cell polarity in regulating GC-induced actin reorganization. 6) using less strong language to describe the results and their interpretation We greatly appreciate the editors' advise. We have extensively modified the manuscript to describe and interpret our results appropriately in the context of accumulated research in the area.

Reviewers' comments
Reviewer #1: This work examines the differences in the capacity of GC to invade polarized and nonpolarized epithelial cells using a polarizable cell line (T84 cells, intestinal) and cervical explants, which contain the different epithelial cell types within the cervix. Piliated, Opa-positive and piliated Opa-negative gonococci are compared throughout the study. Strengths of the manuscript include i.) the state-of-the art microscopy; ii.) the use of a highly relevant cell culture system (cervical explants) that reproduces the heterogeneity of the histology of the cervix, and iii.) investigation of the mechanism of invasion from the viewpoint of the host cell, specifically whether actin polymerization, ezrin recuritment and phosphorylation differ among these cell types and in response to GC. The mechanistic studies were nicely hypothesis-driven.
Areas where this work could be improved by i) more discussion of why nonpolarized T84 cells are likely irrelevant system for studying GC invasion, based on what was found with ectocervical cells in cervical explants (which don't have polarity) Thank you for the suggestion, and we have added a paragraph (Line 615-630) in the discussion section regarding the shortcomings of cell line models, particularly non-polarized epithelia from cell lines, for studying GC invasion in the ectocervix.
ii) adding experimentation with nonpolarized cervical epithelial cells used by many other laboratories to determine whether these are also likely not representing events that occur in whole model systems (i.e. explants) (i.e. it would be very interesting to know whether actin polymerization, ezrin recruitment, etc. occurs in response to GC in ME180 cells in a way that is similar to that of nonpolarized T84 cells).
The revised manuscript provides new data on GC adherence and invasion in ME180 cells cultured on transwells for two days (new supplementary S3 Fig and S5 Fig). Gentamicin resistance assay showed that GC invaded into ME180 cells, and Opa expression facilitated GC invasion (S3 Fig). Immunofluorescence microscopic analysis showed that GC interaction with ME-180 cells induced accumulation of F-actin and ezrin at adherent sites, and Opa expression enhanced the accumulation (S5 Fig). These results are consistent with what we observed in nonpolarized T84 cells.
iii) the lack of staining polarized and nonpolarized cells for CEACAMs, which seems important in light of occasional differences between Opa-positive versus Opa-negative bacteria, and the lack of genetic complementation for in experiments in which a difference between Opa-positive and Opa-negative GC was found.
The revised manuscript provides new data on comparing MS11 strain that expresses a non-phase variable, CEACAM-binding OpaH (OpaCEA) with WT MS11 (Opa+) and its Opa-deletion mutant ΔOpa in new Fig. 1D-F and Fig. 2. Our data show that the expression of OpaCEA enhances GC invasion into non-polarized and polarized T84 cells (Fig. 1D-F) and increases Factin and ezrin at GC adherence sites (Fig. 2), similar to Opa+ GC that express phase-variable Opa. These results confirm the role of OpaCEA in GC invasion by enhancing actin reorganization. This is an excellent point! In the revised manuscript, we discussed the possibilities of GC invading into cervical epithelial cells, inducing epithelial shedding, and GC invading into cervical epithelial cells after epithelial shedding in the discussion section (Line 602-613).
It would also be useful to the reader to place these results, as well as the rationale of the study, in the context of what was previously reported by these authors using the cervical explant system (Yu et al, PLoSPath 2013) so that a larger view of GC invasion through and between cells, with and without CEACAMs could be made. A figure that depicts a model that ties these two pieces of work together would be very useful in summarizing what the cervical explant system has shown. This is an excellent suggestion. We have included a paragraph (Line 632-661) in the discussion section and a new Fig. 9 to describe our working model on cellular mechanisms by which GC infect the cervical epithelium.

Part II
1. Experiments should be done with nonpolarized cervical cells to determine whether the high rate of invasion of nonpolarized T84 cells and changes in actin, ezrin and phosphorylation are unique to the T84 intestinal cell line, and to confirm this important and interesting observation in a frequently used cervical cell line. 2. A complemented opa-negative mutant should be used for all experiments in which a difference was found for Pil+,Opa+ versus Pil+, Opa-GC. Introduction of a single opa gene should be sufficient for the complemented strain. CEACAM staining should also be incorporated into the immunofluorescence studies. This is an excellent suggestion. In the revised manuscript, we compared the adherence, invasion of WT MS11 Neisseria gonorrhoeae (GC) strain that express phase variable Opa (Opa+) and Opa-deletion strain (∆Opa) with ∆Opa strain expressing OpaH that binds to Carcinoembryonic antigen-related cell adhesion molecules (CEACAMs) and cannot phase vary (OpaCEA) in nonpolarized and polarized T84 cells ( Fig. 1D-F). We also compared their impacts on F-actin ( Fig.  2A and 2B) and ezrin ( Fig. 3A and 3B) distribution. Our results show that OpaCEA GC behave similarly as Opa+ GC, displaying higher numbers of epithelial-associated and gentamicinresistant bacteria and higher levels of F-actin and ezrin accumulation at GC adherent sites in both non-polarized and polarized epithelial cells than ∆Opa GC. These new results suggest that Opa can enhance GC adherence and invasion and GC-induced F-actin and ezrin accumulation through binding to CEACAMs. We have previously confirmed that OpaCEA expression induces CEACAM recruitment at the surface of T84 cells [8], ectocervical and endocervical epithelial cells [9], by immunofluorescence microscopy.

Part III
The lack of line numbers makes reviewing this manuscript difficult.
Line numbers have been provided in the revised manuscript.
Experimental: Data pertaining to Fig. 1: A brief description of the details of this experiment with respect to time points should be presented before discussing the data. Adherence was measured at 3 hrs; gentamicin-protected GC were measured at 6 hrs. When was the gentamicin added? While this information is in the methods, it would help the reader understand the data better if the assay is described better here. Similarly, to describe the T84 cells used as cells incubated for 2 days or 10 days in many figure legends is not as informative as describing them as nonpolarized and polarized as was done in the Figure 3 legend.
We have provided a brief description of the experiment for Fig. 1 in the result section (Line 167-172) and the figure legend of the revised manuscript.
Differences in the results when based on the Gm protection assay: While the greater sensitivity of the Gm protection assay compared to microscopy methods is discussed, it is also possible that the Gm protection assay if incomplete exposure of extracellular GC to the Gm due to microcrevices/folding in the cells or inadequate washing. This finding is also potentially useful to others who use this well-established assay. The recovery of greater than 1,000 Gm-protected CFU, for example, in Figure 1 corresponds to a value of 16 % or 7% intracellular GC (number of GC/100 epithelial cells). Is this difference in the two assays consistent for all experiments done in this paper?
We agree with the reviewer's comment. We used both gentamicin resistance and transmission electron microscopy TEM to confirm the effects of cell polarity and Opa expression on GC invasion. Each of the two methods has its advantages and disadvantages. The gentamicin resistance assay is more quantitative and determines the total number of GC surviving the gentamicin treatment in each transwell containing ~2x10 5 epithelial cells before inoculation with 2x10 6 bacteria. However, this method may overestimate GC invasion levels, as gentamicin may not reach bacteria inside microcolonies, particularly Opa+ GC that form big and tight microcolonies [1]. In contrast, TEM allows distinguishing intra-and extra-cellular bacteria visually. However, TEM can only examine a limited number of epithelial cells and GC. Each TEM image only contains two epithelial cells and 10~30 bacteria at a magnification that can distinguish intra-and extra-cellular bacteria rather than all the epithelial cells and bacteria in a transwell. Thus, we have to quantify TEM results using a different method. We apologize for leaving some details in TEM quantification out, and these details have been included in the results (Line 207-219) and methods sections of the revised manuscript. Overall, the results of TEM analysis support the results of the gentamicin resistance assay that epithelial cell polarization inhibits and Opa expression facilitates GC invasion.
Occasionally there is a difference when using Opa-positive versus Opa-negative Gc, for example in the enrichment of ezrin in infected nonpolarized cells and the detection of intracellular GC in endocervical and ectocervical cells (Fig. 5A). A complemented Opa-negative strain (i.e. with one opa gene) should be used to confirm these differences are due to lack of Opa expression. This is an excellent suggestion. In the revised manuscript, we compared the adherence, invasion of WT MS11 Neisseria gonorrhoeae (GC) strain that express phase variable Opa (Opa+) and Opa-deletion strain (∆Opa) with ∆Opa strain expressing OpaH that binds to Carcinoembryonic antigen-related cell adhesion molecules (CEACAMs) and cannot phase vary (OpaCEA) in nonpolarized and polarized T84 cells (Fig. 1D-F). We also compared their impacts on F-actin ( Fig.  2) and ezrin ( Fig. 3A and 3B) distribution. Our results show that OpaCEA GC behave similarly as Opa+ GC, displaying higher numbers of epithelial-associated and gentamicin-resistant bacteria and higher levels of F-actin and ezrin accumulation at GC adherent sites in both non-polarized and polarized epithelial cells than ∆Opa GC. These new results suggest that Opa can enhance GC adherence and invasion and GC-induced F-actin and ezrin accumulation through binding to CEACAMs.
With respect to differences in results when Opa-positive GC are used, CEACAM staining should be performed on polarized and nonpolarized T84 cells to confirm that Opa interactions with these receptors is occurring (or not), and to see if the CEACAMs are on the apical side of polarized cells or localized (i.e. recruited) differently when GC are present.
As discussed above, the new data provided in the revised manuscript ( Fig. 1D-G, Fig. 2, and Fig.  3A and 3B) showed that MS11 GC expressing phase variable Opa proteins (Opa+) behave similarly to MS11 GC expressing non-phase variable, CEACAM-binding OpaH (OpaCEA) in regulating GC adherence and invasion, as well as actin reorganization, in non-polarized and polarized T84 cells. These data suggest that most Opa proteins expressing on Opa+ GC are CEACAM-binding. Our previous studies (Yu et al. 2019. PLoS Path) [8,9] have shown that OpaCEA GC colonization recruits CEACAMs to adherent sites by immunofluorescence.
There are many studies on GC invasion of nonpolarized ME180 cells and primary cervical cells. This should be discussed with respect to the potential lack of polarity in these specific cell systems and how that may influence actin and ezrin recruitment and localization. Could it be that GC is not as invasive as is thought, based on these nonpolarized cell models? Also, no mention is made of the well-defined invasion pathway discovered by Jennifer Edwards, which leaves the discussion incomplete for the reader who is trying to put this all together.
We have modified the discussion section extensively in the revised manuscript to put our findings in the context of the accumulated studies in the last four decades. We have included the studies from Dr. Jennifer Edwards' group in the discussion and suggested that the binding of pili to CR3 is a potential way to activate ezrin and actin reorganization for GC invasion. Even though we still lack direct evidence, our data from human cervical tissue explants suggest lower invasiveness of GC (GC entry into epithelial cells) in vivo than in vitro during the first 24 h window. Low ezrin expression and epithelial cell polarity are likely two of the factors that contribute to this low GC invasiveness, even though CR3 is expressed. The cervical explant system used by the authors is a unique and powerful system in its relevance and capacity to compare differences in different anatomical sites within the cervix. It is puzzling to this reviewer why they do not build on their previous work (PLosPath 2015) with this system to better explain what they are testing and why (in the introduction) and to provide the reviewer with a larger and more comprehensive summary of what has been learned with this system in the discussion. This is an excellent suggestion. We have included a paragraph (Line 632-661) in the discussion section and a new Fig. 9 to describe our working model on cellular mechanisms by which GC infect the cervical epithelium.
Shedding should be discussed in the context of C. Hauck's work in this area. This is an excellent point! In the revised manuscript, we discussed both possibilities of GC invading into cervical epithelial cells, inducing epithelial shedding, and GC invading into cervical epithelial cells after epithelial shedding in the discussion section (Line 602-613) Minor: Last sentence of top paragraph, page 10, is incomplete or missing some words.
The opening line of the abstract states that GC causes symptomatic infections in the cervix but in the introduction, more emphasis is placed on the high percentage of asymptomatic infections in women. This is confusing and because symptomology doesn't have much to do with this paper, this could be removed from the abstract.
Thanks for the suggestion. We have made the correction accordingly.

Part I
In this study, Yu et al. employ two in vitro models (immortalized T84 cells, and cervical explants) to analyze the mechanisms of gonococcal invasion in cells with various morphologies. Building on prior work from their group and others, they present descriptive results on cell morphology during infection, as well as perturbation experiments which suggest roles for actin, the actin-membrane linker ezrin, and signaling events including ezrin phosphorylation and Ca 2+ signaling in host cell responses and bacterial invasion.
In general the experiments are well-controlled and described and the experimentation is extensive. The major experimental limitations (see below) are the use of pharmacological inhibitors that may have off-target effects, and a focus on analyses conducted at only a single time point (in most cases, 6 h post-infection). Most of the microscopy is beautiful.
We thank the reviewer for appreciating our microscopic work and pointing out the limitation of our pharmacological approach and single time point examination. In the revised manuscript, we have discussed the possible off-target effects of inhibitors. Because the molecules we focused on are essential for various functions of epithelial cells, the gene knockdown and knockout approach would have broad impacts on T84 cells, particularly viability and polarity. To minimize the offtarget effects of inhibitors used, we selected minimal effective concentrations. We also tested the effects of inhibitors on GC growth and adherence (for the ezrin inhibitor NSC668394, please see S6 Fig in the revised manuscript and for the non-muscle myosin II light chain kinase inhibitor ML-7 and the Ca 2+ flux inhibitor 2APB, please see our previous publications [9,10]).
Based on reviewer's suggestions, we have compared gonococcal (WT MS11) adherence and invasion in non-polarized and polarized T84 cells on transwells parallelly at three different time points, 3, 6, and 12 h using the gentamicin-resistant assay. We found that at 3 and 6 h, the numbers of gentamicin-resistant bacteria were higher in non-polarized epithelial cells than in polarized epithelial cells, while the total numbers of epithelial-associated bacteria in nonpolarized and polarized T84 were similar. By 12 h, the numbers of both epithelial-associated and gentamicin-resistant bacteria in non-polarized and polarized epithelial cells were decreased compared to those at 3 and 6 h, due to bacterial overgrowth, even though unassociated bacteria were washed away at 6 h. This suggests that the effect of the epithelial cell polarity on GC invasion does not depend on time. Therefore, we chose the 6-h for the rest of our study. The new data is included in Fig. 1A-C.
The manuscript itself raises more significant concerns. As mentioned above, the experiments build on older work. Key papers, sometimes with nearly identical experiments, are not cited as prior art, and in some cases the data presented appear at odds with older results. It will be essential in any resubmission to address these issues. In particular, divergence between the present manuscript and other published data should be highlighted so that inconsistencies can be addressed (and, one hopes, resolved) in future work. This is an excellent point! We have extensively revised the manuscript to put this study in the contest of accumulated research on GC invasion into epithelial cells over the four decades. In the introduction section, we expand our literature review to include critical findings of the previous literature. In the discussion section, we expand the discussion on the similarity and differences between previous studies and this study and future directions.

Part II
Most of the following issues do NOT require additional experimentation, but are still important. It might be possible, with extensive rewriting, or presentation of control data collected but not presented, to address the following concerns without additional experimentation. We agree with the reviewer that non-polarized and polarized T84 cells have been used for studying GC invasion since 1996. The reviewer established these models. That is one of the reasons for us to choose T84 cells as our model. Our data that Opa expression enhances GC invasion in both non-polarized and polarized T84 cells are consistent with the early findings, suggesting the T84 cell model work appropriately. We have extensively revised the manuscript to put our data in the contest of previous studies, including the references suggested by the reviewer. Different from previous studies, we set up both non-polarized and polarized T84 on transwells with the same number of T84 cells per transwells and parallelly compared GC adherence, invasion, and induced actin reorganization in non-polarized and polarized epithelial cells. This enables us to reveal the distinctive mechanisms by which GC interact and invade into non-polarized and polarized epithelial cells. We have clarified this point in the revised manuscript.
1a. Similarly, on p. 22, the authors write: "Our findings that Opa expression promotes GC invasion into T84 epithelial cells, no matter if they are non-polarized and polarized using gentamicin resistant assay, are consistent with previous reports [11] [45]." But that paper (Grassmé) does not employ T84 or any other polarized cell line. It uses Chang cells, now known to be a HeLa derivative.
We apologize for our wrong wording. We intended to state that our results are consistent all previous reports on the enhancing role of Opa on GC invasion into various epithelial cells, including T84. In the revised manuscript, we have modified the sentence to correctly reflect our points.
1b. In Fig. 1, why are the invasion indices (ratio of invasion : cell association) so much higher when assessed by TEM vs. Gm protection assay?
TEM can only examine a limited number of epithelial cells and GC. Each TEM image only contains two epithelial cells and 10~30 bacteria and often cannot cover an entire bacterial microcolony at a magnification that can distinguish intra-and extra-cellular bacteria. Thus, we have to quantify TEM results using a different method from the gentamicin-resistant assay. Eight images were randomly acquired from each condition and each of three independent experiments. We found that ~25% of the images from GC-inoculated non-polarized T84 cells but none of the eight images from GC-inoculated polarized T84 cells showed intracellular GC. We quantified the percentage of intracellular GC over the total number of GC directly contacting epithelial cells in individual images using the images showing intracellular GC. We observed ~16% Opa+ and ~7% ΔOpa GC inside non-polarized T84 cells (Fig. 1H). In the new Fig. 1G, we add big arrows to indicate intracellular GC and small arrows to indicate GC directly contacting the plasma membrane of epithelial cells. We apologize for leaving some details in TEM quantification out previously, and these details have been included in the results (Line 207-219) and methods sections of the revised manuscript. Overall, the results of TEM analysis support the results of the gentamicin resistance assay that epithelial cell polarization inhibits and Opa expression facilitates GC invasion.
2. The authors say (P. 12, 1st paragraph) that Ca 2+ signaling does not occur with non-polarized cells (Fig. 3). That claim appears to be in direct contradiction to prior results (again, key work by other authors is not cited -e.g., Källström  2a. The prior art needs to be cited and discrepancies between those studies and the present work need to at least be mentioned and preferably discussed. We agree with the reviewer's comment. The revised manuscript has provided additional discussion on the early Ca 2+ flux minutes post GC inoculation in epithelial cells shown by included the suggested references and the elevated cytoplasmic level of Ca 2+ at 4 h in GCinoculated polarized T84 (Line 567-573).

2b. Given previously reported results, what might have happened during the previous 4h?
This is an interesting question that requires further investigation. We have expanded the discussion section to discuss this question in revised manuscript (Line 567-573). Early studies have shown that GC interaction can trigger a transient Ca 2+ flux in non-polarized epithelial cells minutes post-inoculation [12][13][14]. The relationship between GC-induced Ca 2+ flux at the early time and the elevated cytoplasmic Ca 2+ level hours later is unclear. If the early Ca 2+ influx does happen in the polarized T84 cells, the persistent Ca 2+ influx could be one possibility for the elevated cytoplasmic Ca 2+ level 4-h post GC inoculation. Another possibility is that the later Ca 2+ elevation is induced by a different mechanism. We have previously shown that the elevated cytoplasmic Ca 2+ in polarized T84 cells observed at later times is sensitive to 2APB, an inhibitor inhibiting Ca 2+ release from the intracellular storage, suggesting an involvement of GC-induced signaling. However, the relationship between the early and late Ca 2+ flux induced by GC is beyond the scope of this manuscript.
2c. Minor point: the authors say that Fluo-4 is used to detect Ca 2+ , but Fluo-4 is cell-impermeant. If an AM ester was used that should be specified and the product number noted. Correct the name.
We have made the correction in the revised manuscript.
3. The descriptions of the roles of microvilli are very interesting, but could be considerably clarified and sharpened. For example, it's pretty clear that it's microvillus *remodeling or dynamics*, not the presence or absence of microvilli per se, that controls the ability of the pathogen to invade cells. That's clearly stated in the Discussion (top of p. 20), but much less clear elsewhere in the manuscript.
We have modified wording accordingly to clarify this point. 4. The general conclusion that polarity, per se, (and especially cytoskeleton polarity) is controlling GC entry, is found throughout the manuscript, but is not particularly well-supported. The development of polarity is accompanied by other changes that occur during differentiation, such as the sequestration of receptors on the basolateral surface or differences in expression of relevant receptor or downstream signaling molecules. Modulating polarity by simply changing the time after plating the cells onto transwell filters does not control for these potential differences. This limitation needs to be discussed.
We agree with the reviewer's comments and discussed the limitation of the cell line model in the revised manuscript (Line 615-630). Our previous studies and this study have partially characterized polarized T84 cells cultured on transwells. We have shown that T84 cells form the tight junction and the adherens junction at the apical surface (S1 Fig, S2 Fig) [10] and polarize the distribution of ezrin-actin networks (Fig. 2, Fig. 3, and S2 Fig), actomyosin [10], and the Opa host receptors CEACAMs apically [8], and epidermal growth factor receptor basolaterally [15].
However, T84 cells cultured on transwells may not have the same transcriptional and protein expression profiles and polarized distributing molecules exactly as endocervical epithelial cells.
5. The small molecule inhibitors used, particularly the ezrin inhibitor, are not particularly wellcharacterized. 2-APB is known to have fairly complex effects and to differentially target different classes of Ca2+ channels, even acting as an *agonsist* of Trpv channels at concentrations similar to those used here (Gao, PMID: 26876731) Again, the experimental limitation should at least be mentioned, with appropriate pointers to literature.
The revised manuscript has discussed the limitation of the inhibitor approaches and included references regarding their off-target effects.
6. The experiments with cervical explants are particularly useful, and represent a real step forward. However, there are again some unstated limitations.
6a. First and foremost, there is the curious result that little or no invasion was detected in endocervical or ectocervical cells. Instead, bacteria were noted in shed cells, leading to the authors' conclusion that GC may prefer to enter shedding cells. The alternative conclusion, of course, is that GC entry *causes* the cells to delaminate. Both alternative hypotheses should be mentioned. This is an excellent point! In the revised manuscript, we discussed the possibilities of GC invading into cervical epithelial cells, inducing epithelial shedding, and GC invading into cervical epithelial cells after epithelial shedding in the discussion section (Line 602-613).
6b. The lack of invasion noted in explants could conceivably reflect not just terminal differentiation or cornification, but cell death. Are the cells that are not invaded alive? Was viability examined?
Cornification of epidermal keratinocytes is a terminal differentiation process and a programmed cell death, which leads to the formation of the outermost skin barrier [16]. It is unknown whether the ectocervical epithelium undergoes a similar process, but it is one of the possibilities. Unfortunately, it is not technically possible yet to determine the viability of epithelial cells in cervical tissues. However, it is definitely one of our future interests. 7. As mentioned above, the study's biggest limitation, to my mind, is the examination of single time points rather than trajectories. The choices of time points need to be explained, and the limitations of using single time points (we can't see what we don't look at) need to be forthrightly discussed.
Based on this suggestion, we have compared gonococcal (WT MS11) adherence and invasion in non-polarized and polarized T84 cells on transwells parallelly at three different time points, 3, 6, and 12 h using the gentamicin-resistant assay. We found that at 3 and 6 h, the numbers of gentamicin-resistant bacteria were higher in non-polarized epithelial cells than in polarized epithelial cells, while the total numbers of epithelial-associated bacteria in non-polarized and polarized T84 were similar. By 12 h, the numbers of both epithelial-associated and gentamicinresistant bacteria in non-polarized and polarized epithelial cells were decreased compared to those at 3 and 6 h, due to bacterial overgrowth, even though unassociated bacteria were washed away at 6 h. This suggests that the effect of the epithelial cell polarity on GC invasion does not depend on time. Therefore, we chose the 6-h for the rest of our study. The new data is included in Fig. 1A-C.

Part I
Yu et al. examine use microscopy to examine the roles of cell polarization, ezrin, and F-actin in gonococcal interactions with human cells. An invasion method is described that occurred in nonpolarized T84 colonic cells and in cervical cells found in the medium of cervical explant cultures, though the mechanisms leading to cell infection and exfoliation were not determined. By contrast, invasion was said not to occur in polarized T84 cells or in cervical explant tissue, raising questions about the significance of this invasion phenomenon.

Part II
1. The biological importance of the differences in FIR reported throughout the paper are difficult to discern, particularly when the differences are so small. For instance, in Fig. 1F the authors report a statistically-significant difference in F-actin FIR between Opa+ GC infections and deltaopa GC infections of non-polarized cells, and yet the averages appear to be 1.4 and 1.2, respectively, and the points in the delta-opa infection condition appear to completely overlap with those in the Opa+ condition, with only eight points in the Opa+ condition outside this range. Are these results actually different? Certainly the micrographs provided ( Fig 1E) do not show any discernable difference between the Opa+ and delta-opa infections.
We agree with the reviewer's comment. Indeed, the differences in the F-actin FIR between Opa+ GC and ∆Opa GC are small and not obviously visible, even though the differences are significant when 54~81 GC microcolonies from three independent experiments are quantified. These small differences suggest that Opa proteins can facilitate but are not essential for GC to induce actin reorganization in non-polarized and polarized T84 cells. This finding is consistent with previous reports that Opa proteins are not essential [2,3] but can facilitate GC invasion into epithelial cells [4][5][6][7]. Notably, the differences in the F-actin FIR between GC inoculated non-polarized and polarized T84 cells are much bigger than between Opa+ and ∆Opa GC, suggesting a more substantial role of epithelial cell polarity in regulating GC-induced actin reorganization.
2. Fig. 1 claims to report both invasion (1B) and intracellular GC (1D), yet these two experiments yield different results. The authors conclude from 1D that no intracellular GC are present in polarized cells, while the standard invasion assay in 1B shows maybe 30% as many GC in the polarized cells as in the non-polarized.
We used both gentamicin resistance and transmission electron microscopy TEM to confirm the effects of cell polarity and Opa expression on GC invasion. Each of the two methods has its advantages and disadvantages. The gentamicin resistance assay is more quantitative and determines the total number of GC surviving the gentamicin treatment in each transwell containing ~2x10 5 epithelial cells before inoculation with 2x10 6 bacteria. However, this method may overestimate GC invasion levels, as gentamicin may not reach bacteria inside microcolonies, particularly Opa+ GC that form big and tight microcolonies [1]. In contrast, TEM allows distinguishing intra-and extra-cellular bacteria visually. However, TEM can only examine a limited number of epithelial cells and GC. Each TEM image only contains two epithelial cells and 10~30 bacteria at a magnification that can distinguish intra-and extra-cellular bacteria rather than all the epithelial cells and bacteria in a transwell. Thus, we have to quantify TEM results using a different method. We apologize for leaving some details in TEM quantification out, and these details have been included in the results (Line 207-219) and methods sections of the revised manuscript. Overall, the results of TEM analysis support the results of the gentamicin resistance assay that epithelial cell polarization inhibits and Opa expression facilitates GC invasion. Fig 1, the results for Opa+ and delta-opa in Fig 2B and 2G appear very small. The error bars presented do not appear to represent the large spread in the data as represented by the data points shown.

As in
The differences in the ezrin FIR between Opa+ and DOpa GC are consistent with the differences in the F-actin FIR between Opa+ and DOpa GC. As detailed above, these results suggest that Opa proteins can facilitate but are not essential for GC-induced actin reorganization in nonpolarized and polarized T84 cultured on transwells, as discussed above. Data points represent individual microcolonies. The statistical analysis is based on 54~81 individual GC microcolonies from three independent experiments, and error bars are based on the standard error of the mean (SEM) for a large number of data points. 4. It is unclear if the microscopy experiments were repeated, i.e., does the "n=3" found in the figure legends mean that the experiment was done on three different days with three different cultures, or does it mean that it was done once with three technical replicates?
We performed the experiment on three different days with three different culture. We have changed n=3 to three independent experiments in the revised manuscript. 5. One of the most interesting results of the study is the observation of sloughed cervical cells with internalized gonococci. The authors claim both that quantification of this process is impossible and that the number of gonococci that invaded cervical cells is less than the number that remain on the surface. Isn't it possible that gonococci invaded significant numbers of cervical cells that were then sloughed from the epithelium? They should collect sloughed cells from the medium and plate the bacteria, add more earlier time points to the microscopy studies to try to see the invasion and sloughing process, and quantify cell sloughing as has been done by Muenzner (Science 329:1197).
We thank the reviewer for appreciating our results and for the excellent suggestion. Indeed, we can estimate the relative number of epithelial cells shed from the cervical tissue (see our previous publications [9,10]), similar to Dr. Hauck's group did with a mouse model [17]. To quantify the number of exfoliated epithelial cells with intracellular GC, we did try to harvest exfoliated epithelial cells from culture supernatants before our initial submission. However, we got very low numbers of epithelial cells, suggesting that most exfoliated epithelial cells were not in the culture supernatant. We suspect that most exfoliated epithelial cells may remain loose association with the cervical tissue and may be removed during washing, and some of them may attach to the surface of the culture well surface [18], which made quantification almost impossible. Importantly, even if we could collect epithelial cells from all these factions, we still cannot distinguish whether GC enter epithelial cells before or after epithelial cells shed off. Limited availability of human cervical tissues prevents us from analyzing multiple time points. To address these issues, we are currently developing live imaging of cervical tissue using the multiphoton intravital microscopic technique, but this is beyond the scope of this manuscript.
While we currently cannot experimentally address this question, we revised the discussion section to include the possibility of GC invasion before epithelial exfoliation, causing epithelial shedding.
6. The authors' previous cervical tissue studies identified the transition zone as the region of the cervix where gonococci invade tissue (Yu et al. doi.org/10.1371/journal.ppat.1008136). Since the authors in that study reported non-polarized cells in the transition zone, it seems quite possible that the invasion phenomenon studied in the current manuscript could be going on in those cells. Why was the transition zone not examined in this study? Thus the conclusions in the last paragraph of the discussion that the cervix is protected from GC invasion may be wrong.
We did not include transformational zone because epithelial cells in this location are extremely heterogeneous, gradually increasing cell polarity from the ectocervical side into the endocervical side. The area of the transformational zone also different from one cervix to the other. We do not have a cell line model to mimic such a polarity gradient. In addition, transformational zone epithelial cells express a low level of CEACAMs. While GC invasion in the transformational zone is very interesting and important, its complexity requires sole focus that is beyond the scope of the current manuscript. We agree with the reviewer's comments and revised this part of the discussion accordingly. The low expression of ezrin and the epithelial cell polarization may protect the ectocervix and the endocervix from GC invasion, respectively. The revised manuscript has extensively revised the introduction and the discussion section to put our study in the context of the research accumulated in the last four decades.
The data presented in this manuscript is largely consistent with the studies reviewed and published by Merz and So [6,19,20] on GC invasion and Opa's roles in GC invasion. GC entering epithelial cells from the apical surface can exit from the basolateral surface, leading to transcytosis. We have recently shown that GC transmigration across polarized epithelial cells, including endocervical epithelial cells, is associated with their ability to disrupt the apical junction that seals the paracellular space [9,10,21]. The lower level of intraepithelial GC than subepithelial GC after 24 h incubation supports that GC can transmigrate across the epithelium by breaking the apical junction, but does not exclude the transcytosis pathway, such as GC entering and exiting epithelial cells rapidly or GC entery and transcytosis of the epithelium after 24 h. Because GC transmigration across the epithelium is not the focus of this manuscript, we have decided not to discuss this point. --------------------