ZIKV infection induces robust Th1-like Tfh cell and long-term protective antibody responses in immunocompetent mice

Induction of long-lived antibody responses during infection or vaccination is often essential for subsequent protection, but the relative contributions of T follicular helper (Tfh) cells and T helper 1 (Th1) cells for induction of antigen specific antibody responses to viruses are unclear. Here, we establish an acute Zika virus (ZIKV) infection model in immunocompetent mice, and show that ZIKV infection elicits robust Th1-like Tfh cell and protective antibody responses. While these Th1-like Tfh cells share phenotypic and transcriptomic profiles with both Tfh and Th1 cells, they also have unique surface markers and gene expression characteristics, and are dependent on T-bet for their development. Th1-like Tfh cells, but not Th1 cells, are essential for class switching of ZIKV-specific IgG2c antibodies and maintenance of long-term neutralizing antibody responses. Our study suggests that specific modulation of the Th1-like Tfh cell response during infection or vaccination may augment the induction of antiviral antibody response to ZIKV and other viruses.

1. The authors hint at an underlying factor that could account for the difference in humoral response to ZIKV infection in IFNAR blocked vs control BALB/c mice. Viremic infection in the former likely have resulted in more antigens being expressed compared to the latter. Greater antigen load could hence have resulted in more antigen presentation by Tfh cells or activation of these cells in the lymph nodes, or both. Alternatively, the immune response could have been dependent on replicating ZIKV, where the pro-inflammatory factors expressed in infected cells could have stimulated antigen presenting cells and/or T cells in ways that cannot be reproduced even if an equivalent level of inactivated antigens were inoculated in these animals. The inclusion of an inactivated ZIKV control in the experiments could have clarified these questions. It would also inform on whether, besides the degree of cellular immunity, qualitatively different humoral response could be elicited by replicating compared to nonreplicating vaccines due to the differences in Th1-like Tfh stimulation. This possibility could have important implications on how the pipeline of candidate Zika vaccines should be prioritised for clinical development. Perhaps the authors would consider adding a discussion on this issue? 2. Line 441 indicated that 500-1000 pfu/ml virus was incubated with an equivalent volume of serially diluted serum before inoculation onto a monolayer of Vero cells in a 24-well plate. Unless the volume was very small, which could introduce inaccuracies, the number of pfu appears to be very large for a 24-well plate assay. Is this description correct? Please clarify. 3. Figure 1a suggests that there is some background ZIKV neutralization activity in the uninfected animal controls (a 50% neutralisation titer of nearly 2 logs at 14 days). This is rather unusual. Can the authors explain what is going on? 4. Line 155. Figure 2b should be labelled here as Figure 3b. 5. There are scattered typographical/grammatical errors in the manuscript.
Point-by-point response to reviewers' comments Editorial comments: In this case reviewers raised significant concerns regarding some aspects of the models you present and suggest (1) further in vivo work to further enhance your manuscript, to fully and more appropriately understand the role of Th1 like Tfh cells in this context, which we agree would be beneficial in this case. Reviewers additionally raised concerns regarding (2) the molecular mechanisms at play and we agree further elucidation in this context would be necessary here. That is not to say we find any point raised by reviewers as less important and strongly advise you to address each point of concern raised by each reviewer as fully as possible by the most appropriate means.

Response:
We interpret the above as that you agree with the reviewers on two of their major concerns and you suggest that we address these issues by (1) performing more in vivo work, and (2) elucidating relevant molecule mechanisms.
To address these important issues, we took the sensible suggestion from reviewer #2 of doing mixed bone marrow chimera experiments. We performed new experiments by giving Tcrb -/mice two different mixture of cells: an equal number of WT: Bcl6 fl/fl Cd4-Cre, and Tbx21 -/-: Bcl6 fl/fl Cd4-Cre, to construct mixed bone marrow chimeras. As expected, there are specific deficiency in Th1-like Tfh cells, but still capable of generating conventional Tfh cells, with about 40% reduction of pre-Tfh cells in these mixed bone marrow chimera (Reb. Figure 1a, 1b), which is consistent with that observed in Tbx21 -/mice (Fig. 5a). And IgG2c production is impaired with IgG1 production increased (Reb. Figure 1c, 1d). These new data support the conclusion that Th1-like Tfh (and -pre-Tfh), but not conventional Tfh cells, are responsible for IgG2c class switching in ZIKV infection. Reb. Fig.1a has been included in the revised paper as supplementary  IgG1 and IgG2c intracellular staining in   B220 low CD138 + IgD low cells (left panel); bar graphs summarized the numbers of IgG1 + and IgG2c + cells (right   panel). (d) The concentrations of ZIKV envelope specific IgG1 and IgG2c antibodies in chimera were measured by ELISA. The summary data were presented as mean ± SEM. Statistical differences were determined by Student's t test and p values were indicated by * (p<0.05), or ** (p<0.01), or ***(p<0.001).
To study molecular mechanisms in greater depth, we again performed new experiments by giving Rag1 -/mice an equal number of cells from CD45.1: Ifngr1 -/to make a new type of mixed bone marrow chimera. These chimeric mice allowed dissection of whether IFN-γ signaling is required and by what cell types. In these chimeras, IFN-γ receptor-deficient Tfh cells developed normally, with slight reduction of pre-Tfh (Reb. Figure 2a), and these cells could produce similar levels of IFN-γ, compared to CD45.1 + wild type Tfh cells in the same chimeras (Reb. Figure 2b), consistent with that in Ifngr1 -/mice ( Supplementary Fig. 8a). during ZIKV infection. However, in the chimera, IFN-γ receptor-deficient B cells produced much reduced level of IgG2c antibodies than CD45.1 + wild type B cells (Reb. Figure 2c), suggesting that IFN-γ signaling is required for B cell to mediate antibody class switching. Reb. Figure  bar graphs summarized the percentages (Right panel). The summary data were presented as mean ± SEM.
Statistical differences were determined by Student's t test and p values were indicated by * (p<0.05).
In addition to Th1-like Tfh cells, many other cell types such as NK, Th1 and CD8 + T cells can produce IFN-γ. To address the question of whether IFN-γ produced from Th1-like Tfh cells induces IgG2 switching, we created an Ifng -/-: Bcl6 fl/fl Cd4-Cre mixed bone marrow chimera, in which Tfh cells all come from Ifng -/-CD4 + T cells. As expected, these mixed bone marrow chimera could generate Tfh and pre-Tfh cells normally (Reb. Figure 3a), but could not generate IFN-γ + Th1-like Tfh cells, distinctly different from the WT: Bcl6 fl/fl Cd4-Cre mixed bone marrow chimera (Reb. (d) The concentrations of ZIKV envelope specific IgG2c antibody in chimeras were measured by ELISA. The summary data were presented as mean ± SEM. Statistical differences were determined by Student's t test and p values were indicated by * (p<0.05), or ** (p<0.01), or ***(p<0.001).

Comments from Reviewer #1
(1) This study described "Th1-like" Tfh cells in the context of Zika virus infection. The study is well done on technical side, using various genetically modified models to demonstrate that IFNg-produced cells constitute a large portion of Tfh cells, that generation of these cells depend on Bcl6 and T-bet, and the IFNg produced is important for IgG2 class switching. However, on conceptual side, there is little new. It has been already known that Tfh cells produce IFNg, which requires Tbet, and generation of Tfh cells depends on Bcl6, although the portion of IFNg-Tfh cells might be higher in frequency in ZIKV infection compared with LCMV infection.

Response:
We thank the Reviewer #1 for acknowledging the technical strength of our paper, but would also like to highlight the conceptual advance our paper has been, albeit the degree of advancement may be judged subjectively. The novelty of our results may be highlighted from the following three aspects.
It is true that Tfh cells produce IFN-γ, for which T-bet is required, and it is known that the generation of Tfh cells depends on Bcl6. However, neither of which has been established in a Zika virus infection model, not to mention in immunocompetent mice that are hard to infect by Zika virus. Moreover, most other studies of Tfh have not included parallel investigation of pre-Tfh and GC B responses. Together, the three population of cells give a more comprehensive overview of the complex interactions among different cell subsets.
Secondly, some previous papers have indicated the existence of IFN-γ-producing Th1-like Tfh cells, however, quantitatively, such a large expansion of Th1-like Tfh cells (about 70% of the total Tfh cells) has never been observed before. It would not be unreasonable to emphasize that quantity is often as important as quality in a biological process. The specificity of such a phenomenon, i.e., a greater expansion of Th1-like Tfh cells in ZIKV infection than that in LCMV (supplementary Fig.4b), Influenza and Dengue infections (unpublished data) also added another level of novelty.
Functionally, we showed that Th1-like Tfh cells are responsible for IgG2c class switching, thus providing a convincing piece of evidence in support of the idea that IFN-γ being critically important for class-switch recombination [1].
Based on the above, we think this study may have some degree of novelty. (2) The authors may have different views of Tfh differentiation process. Generally, PD1-CXCR5+ cells are committed Tfh cells, not pre-Tfh cells, PD1+CXCR5+ cells are GC Tfh cells, and the interconversion between these two subsets can be quite dynamic. The data actually showed that both subsets can produce IFN-γ.

Response:
We agree with the reviewer that there are divergent views on the differentiation process of Tfh cells and how these cells are characterized phenotypically. As far as we can find in published literature, there are several studies describing pre-Tfh cells as being residing in the T-B border where they meet B cells and form stable T-B conjugates that migrate into the germinal center [1][2][3][4][5][6]. These pre-Tfh cells are characterized as being CXCR5 medium PD-1 medium on flow cytometry analyses [1,2,3]. Similarly, there are other studies using CXCR5 high PD-1 high as the phenotypic marker for Tfh cells within the CD4 + T cell population [7][8][9][10][11][12]. Therefore, in our study, we marked pre-Tfh as CXCR5 medium PD-1 medium , and Tfh as CXCR5 high PD-1 high .
(3) The authors focused on a specific role of IgG2c class-switch. Based on the data using CD4Cre+Tbet fl/fl and Ifngr1 KO mice, both IgG2b and 2c were affected. The germline deletion of Tbet had a more specific effect on IgG2c. These need to be reconciled.

Response:
The reviewer has rightly picked out an issue that puzzled us as well. Before obtaining direct experimental evidence, we reasoned that class-switch recombination (CSR) for IgG2c, and IgG2b may be mediated by different cytokines, and that the germline deletion of T-bet affected some, but not other cytokines. This view is more clearly presented in a recent review article by McHeyzer-Williams and colleagues in that CSR of IgG2c (or IgG2a) is facilitated by IFN-γ, whereas IgG2b by TGF-β [1]. Upon more careful literature review, we found several previous paper reporting that both IFN-γ and TGF-β signaling could modulate IgG2b class switching [2,3,4,5].
To more directly investigate the molecular events in relation to this question, we performed new experiments in a cleaner model of mixed bone marrow chimera by giving Rag1 -/mice an equal number cells from of WT: Tbx21 -/-. These chimeras will allow dissection of whether B cells or Th1-like Tfh cells function as mediators of IgG2b production. As expected, there is no Th1-like Tfh cells in Tbx21 -/-(CD45.2 + ) mice (Reb. Figure 4a). Functionally, there is no difference in IgG2b production between WT B cells and Tbx21 -/-B cells (Reb. Figure 4b). Our new data and previous studies support the idea that B cell-intrinsic T-bet is not responsible for IgG2b class switching [6,7]. Therefore, the differential effects on IgG2b and IgG2c observed previously in mice with germline deletion of T-bet may be resulted from factors produced by in non-T and non-B Tbx21 -/cells, such as Tbx21-expressing NK cells, Tbx21-expressing dendritic cells and Tbx21-expressing ILC1. A definitive answer to this question will require further research.

Our response:
We thank the reviewer for acknowledging our study showed something important. We believe the reviewer also asked us to address 3 issues, to which we respond as the following: Response to Q1 (However, the study could be strengthened by the use of cleaner models to pinpoint the role of Th1-like Tfh cells): As the reviewer has suggested, we first re-analyze our previous data and found no decrease in the numbers of non-Th1-like Tfh cell and non-Th1-like pre-Tfh cells in Tbx21 -/and Tbx21 f /f Cd4-Cre mice (Reb. Figure 5a, 5b).  Figure 5c, 5d; Reb. Figure 1a, 1b), IgG2c production is impaired, whereas IgG1 production is increased (Reb. Figure 5e, 5f; Reb. Figure 1c, 1d). These new data, together with our previous data, more definitively showed that Th1-like Tfh cells (and Th1-like pre-Tfh cells), but not non-Th1-like Tfh cells, are responsible for IgG2c class switching. Reb. Figure 5c has been included in the revised paper as supplementary Figure 7c; Reb. Figure 5d, 5e, 5f have been included as revised

Response to Q2 (whether IFNγ signaling is required)
We first reanalyze our previous data and found that Th1-like Tfh cells have normal differentiation in Ifngr1 -/-(Reb. Figure 6a), indicating that IFN-γ signaling has no effect on Th1-like Tfh cell differentiation.
To elucidate mechanistic details, we performed new experiments using a cleaner model mixed bone marrow chimera by giving Rag1 -/mice equal numbers of cells from ly5.1 and Ifngr1 -/-. The resultant chimeric mice were then infected by Zika virus. In these chimeras, IFN-γ receptor-deficient Tfh cells developed normally, with slight reduction of pre-Tfh (Reb. Figure 6b; Reb. Figure 2a). Consistently, Th1-like Tfh cells have not been affected in Ifngr1 -/-(CD45.2), and their numbers are similar as that in WT (CD45.1) (Reb. Figure 6c; Reb. Figure 2b), but IgG2c production is decreased (Reb. Figure 6d; Reb. Figure 2c). These data support the conclusion that IFN-γ signaling is required for B cell to mediate antibody class switching. Reb. Figure 6a We want to emphasize that our paper is different from previous publications and that our "Th1-like Tfh" cells are from "ex-Tfh1" cells for the following reasons:   Figure 7a).
We also compared the RNA-seq data from tableS2 in ref14 (Fang et al., JEM, 2018) with that of ours. In Th1-like Tfh cells, there are 104 unique up-regulated genes and 289 unique down-regulated genes, whereas ex-Tfh1 cells have 177 unique up-regulated genes and 32 unique down-regulated genes, the numbers do not match. Unsurprisingly, however, Th1-like Tfh cells and ex-Tfh1 cells share only 19 common up-regulated genes, and only 1 common down-regulated gene compared to conventional tfh cells. (Reb. Figure 7b). To acquire more information, we downloaded the raw RNAseq data of ex-tfh1, after normalization with our RNAseq data, we found 831 up-regulated genes (such as crtam and ly6i) and 846 down-regulated genes (such as il2ra and Reb. Table 1 Summary of difference between Th1-like Tfh cells and ex-Tfh1 cells.

Response:
We thank the reviewer for pinpoint a strength of our experimental system.

Q5:
The experiments depleting Th1-like Tfh cells using total or T cell-specific Tbx21 deficiency are the main weakness of this paper. While the authors intend to use these mice to specifically deplete Th1-like Tfh cells, they find that Tbx21-/-mice have decreased numbers of total pre-Tfh cells (Fig. 5a) and Tbx21f/fCD4-Cre mice have decreased numbers of both pre-Tfh cells and Tfh cells (Fig. S7a). Therefore, it is unclear whether the decreased IgG2c antibody titers are due to specific loss of Th1-like Tfh cells or just overall reduction of Tfh cells. The decrease in pre-Tfh and Tfh cell counts is potentially not an issue if the loss of Th1-like Tfh cells is solely responsible for the decreased counts -the authors could see if non-Th1-like Tfh cell counts remain the same in Tbx21-deficient mice. However, if both Th1-like and non-Th1-like Tfh cell counts are reduced, the authors would need to use a cleaner experimental model (e.g. Bcl6f/fCD4-Cre:Tbx21-/-mixed bone marrow chimera mice). The authors could also demonstrate that IFNγ is the operative signal from Th1-like Tfh cells that induces IgG2 switching with the use of Bcl6f/fCD4-Cre:Ifng-/-mixed bone marrow chimera mice.

Response:
We agree with the reviewer that these above underlined issues (added by the authors) are indeed important for understanding the specific molecular mechanisms of our reported observations. As the reviewer has suggested, we re-analyze our data and found no decrease in non-Th1-like Tfh cell and non-Th1-like pre-Tfh cell counts from Tbx21 -/and Tbx21 f /f Cd4-Cre mice (Reb. Figure 5a, 5b), suggesting the loss of Th1-like Tfh (and Th1-like pre-tfh) cells is solely responsible for the decreased counts in Tbx21 -/and Tbx21 f /f Cd4-Cre mice.
To further study this issue from the molecular mechanism perspective, we took this reviewer's insightful suggestion Q6: Additionally, Figure 6 argues that IFN-γ signaling is required for class switching to IgG2, but does not pinpoint which cell type requires such signaling due to the use of total knockout Ifngr1-/mice. The authors do demonstrate unimpaired pre-Tfh and Tfh cell induction in Ifngr1-/-mice, but do not demonstrate their cytokine profile -does Ifngr1 deficiency affect the ability of Tfh cells to produce IFNγ? Or is Ifngr1 deficiency in B cells solely responsible for decreased IgG2 production?
The authors could add clarity to this part of the manuscript by limiting Ifngr1 deficiency to specific cell types.

Response:
We thank Reviewer #2 for the insightful and critical comment. These questions have been asked and addressed earlier. To recap, we analyze our data and found that Th1-like Tfh cells have normal differentiation in Ifngr1 -/-(Reb. Figure 6a), indicating IFN-γ signaling has no effect in Th1-like Tfh cells differentiation.
To further address this question, we used a cleaner model by giving Rag1 -/equal numbers of ly5.1 and Ifngr1 -/mixed bone marrow chimera to perform Zika virus infection experiment. Results of which can be found above in "Response to Reviewer #2 (Q2)" and Reb. Figure 6.
Minor comments: 1. The manuscript contains a number of typos throughout.

Response:
We have carefully rechecked our revised manuscript and corrected the typos.
2. The initial characterization of Tfh cell response and cytokine profile is performed in BALB/c mice ( Fig. 1-2). However, later studies are done in Bcl6f/fCD4-Cre mice on C57BL/6 background. Do the authors have parallel characterizations in both strains? Obviously these two strains have very different T cell differentiation proclivity.

Response:
We thank Reviewer #2 for pointing out this important issue. In fact, during preliminary studies, we have done the parallel characterization in both C57BL/6 and BALB/c in T cell differentiation.
Because it is easier to establish an experiment model with BALB/c mice that are easier to handle, we opted to use it during the model set-up stage. For mechanistic studies, most KO mice are in C57BL/6 Basically, the two mouse strains showed similar but not identical results. Specifically, there is a stronger Tfh response but a similar pre-Tfh response in BALB/c compared with C57BL/6 mice (Reb. Figure 8a). As for responses from other T helper subsets, their Th1 responses are similar (Reb. Figure 8b), but their IFN-γ production are slightly different. These minor differences between the mouse strains do not change the conclusions derived from our experimental data.
Reb. Figure 8. Comparison of T cell responses between C57BL6 and BALB/c mice in ZIKA infection. 3. Statistics are missing in most of the subfigures within Fig. 1.

Response:
We have added statistics to Fig.1 in the revised manuscript.
4. In Figure 2b, please indicate on which day Tfh cells were assessed for IFNγ production.

Response:
We have now added the time-points in our revised figure legend of Fig 2b. 5. Line 213: Gata3 is not a conventional Tfh-associated gene.

Response:
We thank the reviewer for pointing this out. We have deleted gata3 in our revised manuscript.

Reviewer #3
This manuscript by Liang and colleagues report a role for Th1-like Tfh in shaping the humoral response to Zika virus (ZIKV) infection in a mouse model. The authors used BALB/c mice with IFNAR inhibition to enable productive ZIKV infection as evidenced by detectable viremia to identify how this subset of T cells is important in inducing robust IgG response, especially in the generation of IgG2 antibodies. The authors suggest that their findings provide insights into the targeting of Th1-like Tfh cells for manipulation to induce robust humoral response to vaccination.
The findings are indeed interesting and the authors appear to have done extensive amount of work to pinpoint an important role for this subset of Tfh cells. I appreciated the attention to detail in ruling out a possible effect of IFNAR blocking antibodies in shaping the humoral response. Likewise, the use of relevant knockout mouse models supported the line of thinking clearly. I have a few comments for the authors to consider.

Response:
We thank Reviewer #3 for the encouraging comment.
1. The authors hint at an underlying factor that could account for the difference in humoral response to ZIKV infection in IFNAR blocked vs control BALB/c mice. Viremic infection in the former likely have resulted in more antigens being expressed compared to the latter. Greater antigen load could hence have resulted in more antigen presentation by Tfh cells or activation of these cells in the lymph nodes, or both. Alternatively, the immune response could have been dependent on replicating ZIKV, where the pro-inflammatory factors expressed in infected cells could have stimulated antigen presenting cells and/or T cells in ways that cannot be reproduced even if an equivalent level of inactivated antigens were inoculated in these animals. The inclusion of an inactivated ZIKV control in the experiments could have clarified these questions. It would also inform on whether, besides the degree of cellular immunity, qualitatively different humoral response could be elicited by replicating compared to non-replicating vaccines due to the differences in Th1-like Tfh stimulation. This possibility could have important implications on how the pipeline of candidate Zika vaccines should be prioritised for clinical development. Perhaps the authors would consider adding a discussion on this issue.

Response:
We think the reviewer's major concern here is how antigen dose will affect the observed immune response. The following are pertinent results. We have also elaborated our thought on this point in the discussion section of our revised manuscript.
In fact, we have done an experiment with inactivated Zika virus as control. In the presence of anti-IFNAR1 antibody pre-treatment, inactivated Zika virus induced relatively fewer Tfh and Th1-like Tfh cells, or other helper T cell subsets, and much less antibody response compared with the same experiment done with live Zika virus (Reb. Figure 9).
Reb. Figure 9. 2. Line 441 indicated that 500-1000 pfu/ml virus was incubated with an equivalent volume of serially diluted serum before inoculation onto a monolayer of Vero cells in a 24-well plate. Unless the volume was very small, which could introduce inaccuracies, the number of pfu appears to be very large for a 24-well plate assay. Is this description correct? Please clarify.

Response:
We thank Reviewer #3 for mentioning this issue. For plaque reduction neutralization test (PRNT), Zika virus (50-100 PFU/100ul/well) was mixed with serial diluted inactivated serum at a volume ratio of 1:1. The actual volume is indeed small, each well is added with add 200 µl of Virus+Serum mixture that contains 0-100 PFU of the virus. We have also added detailed description of the assay in the method section of revised manuscript.
3. Figure 1a suggests that there is some background ZIKV neutralization activity in the uninfected animal controls (a 50% neutralization titer of nearly 2 logs at 14 days). This is rather unusual. Can the authors explain what is going on?

Response:
We thank Reviewer #3 for pointing out this issue. We observed that the PRNT50 values in uninfected mice at day14 is a little higher in one of the three mice, whereas results in other two mice are similar to normal control levels (Original Figure 1b, which is now Reb. Figure 10a). This apparent outlier has swayed the data off a bit.
To further verify the observed higher background neutralization in the uninfected mice is indeed spurious, we performed new experiments using both normal uninfected C57BL/6 and BALB/c mice, which also showed generally low PRNT50 levels, except for that at high concentration (Reb. Figure 10b, 10c). So, it is likely that the higher background neutralization activity at 14 dpi in one of the mice was due to individual variation among mice, but not a systemic problem with the PRNT assay.

Response:
We check the original submission for Figure 2b at line 155, which looks like a correct label. In contrast, Figure 3b shows

Response:
We have rechecked our manuscript carefully and corrected all the typos in the revised version.
this route, we would like to ask for an extension for resubmission until we heard an affirmative reply from JEM.
An alternative approach, which we think can also satisfactorily address the reviewer's concern. We reason that if we can demonstrate the NKG2D antibody can stain positively on cells that are expected to express high levels of this protein, such as NK cells, and then a negative or weakly positive staining on Th1-like Tfh with the same antibody is unlikely to be a false negative result.
In the current revision, we provided new data in the form of supplementary Fig.6d to show that NKG2D antibody we used can indeed stain NK cells nicely. We revised corresponding text section to reflect the new results.
In the interests of returning a revised version of our paper within two weeks, we submitted the revised paper with the second approach.

Response:
We have revised the mistake accordingly.

Response:
We agree with these suggestions and made all changes accordingly. 4. We recommend that the titles for Fig 7 & Supp Fig 10 be revised as follows to more accurately reflect the data shown: The IFN-γ pathway is required for IgG2c antibody class-switching.

Response:
We agree and made changes accordingly. 5. We recommend that the title for Fig 8 be revised as follows to more accurately reflect the data shown: IFN-γ produced by Th1-like Tfh cells is required in a B cell-intrinsic manner for IgG2c antibody class-switching.

Response:
We agree and made changes accordingly.