Chronic circadian disruption modulates breast cancer stemness and immune microenvironment to drive metastasis in mice

Breast cancer is the most common type of cancer worldwide and one of the major causes of cancer death in women. Epidemiological studies have established a link between night-shift work and increased cancer risk, suggesting that circadian disruption may play a role in carcinogenesis. Here, we aim to shed light on the effect of chronic jetlag (JL) on mammary tumour development. To do this, we use a mouse model of spontaneous mammary tumourigenesis and subject it to chronic circadian disruption. We observe that circadian disruption significantly increases cancer-cell dissemination and lung metastasis. It also enhances the stemness and tumour-initiating potential of tumour cells and creates an immunosuppressive shift in the tumour microenvironment. Finally, our results suggest that the use of a CXCR2 inhibitor could correct the effect of JL on cancer-cell dissemination and metastasis. Altogether, our data provide a conceptual framework to better understand and manage the effects of chronic circadian disruption on breast cancer progression.

1) The methods do not detail the background of the MMTV-PyMT mice, nor do they describe how the tumor studies were completed. The primary issue that, due to the lack of detailed methods, needs to be resolved is why the frequency of metastasis in LD PyMT mice is so low. Numerous previous studies have shown that virtually 100% of PyMT mice on the FVB background develop pulmonary metastases, not the 30(ish)% they show.
2) Color blind readers simply cannot see figures that are red-green. Figure 2C is an excellent example of a figure that does not need to be red-green, but because it is, it cannot be interpreted. It literally appears as the same shade for both top and bottom. Black and grey would work. Or simply a bar that shows % with metastasis. Other problem figures include: Figure  Also, supplemental data -please just change everything that is red/green. Figure 2D should have dots representing the number of metastatic foci, not just bins of categories. Please redo statistical tests using all the data. In addition, representative images should be shown.

3)
The statement that PC1 separates tumors with / without metastasis is unfounded based on the data in Figure 3B. In addition, there are far too few data points to even try this. This portion of the figure should be cut. Figure 4A is not convincing for an increase of stem cells in the jet lag mice. P-values seem to be largely driven by a handful of outliers in the control population. Figure 4F is based on 4 tumors from LD and 4 from JL -but NO characterization of the primary tumor was provided. There are numerous studies illustrating the heterogeneity in histology from PyMT mice. This was not taken into account and so the authors may be comparing spindleoid myoepithelial tumors to lobular epithelial tumors, with obvious bias in the percentage of potential tumor initiating cells. CXCR2 inhibitors have previously been shown to inhibit metastasis. However, the experimental design in Figure 6 is lacking an essential control -no LD mice were tested +/-CDCX2 inhibitor to determine if the response for the JL and LD mice was altered. Indeed, given the work of Halpern et al (PMID 21601983), this control is essential. Also, it is essential to place the work into the context of this prior work.
Reviewer #2 (Remarks to the Author): Expertise on circadian rhythm and cancer General comments The biomedical applications of circadian clocks represent a critical challenge for medical progress, especially for cancer. More specifically, an increased risk of breast cancer has been shown in women undergoing prolonged shift work, and this environmental condition was acknowledged as a likely cause of cancer by the International Agency for Research on Cancer both in 2007 and in 2019. Several other reports show that circadian disruption also impacts on the outcomes of tumour bearing rodents, as well as large cohorts of cancer patients. Taken together, the existing literature emphasizes the need for a better understanding of the mechanisms linking circadian disruption and carcinogenesis. This manuscript provides highly interesting and innovative data in this regard.
In aggregate, Hadadi et al. highlight the impact of iterative daily schedule shifts ("chronic jet lag protocol") for the metastatic dissemination of breast cancer in an experimental model of spontaneous mammary carcinogenesis. They identify key mechanisms at work at the chemokine/cytokine network level, and within the tumour infiltrating immune cells. They show that a CXCR2 inhibitor could prevent the deleterious effect of iterative schedule shifts on breast cancer dissemination. The data presented are new and convincing, yet several issues deserve to be answered.
Main comments (1) As mentioned by the authors, the jet lag protocol that has been applied in their study has proven its ability to suppress circadian rhythms in rest-activity and core body temperature as well as several clock genes expressions, both in the SCN and in peripheral tissues of male B6D2F1 mice. Although it is highly likely that similar effects would be observed for the transgenic mice used in this study, it is needed to report the circadian phenotype of these mice when kept both in usual light-dark conditions (LD12:12), and on the "chronic jet lag" protocol.
(2) Because the vast majority of the parameters studied undergo large endogenous circadian variations in mice on LD12:12 (as well as in constant darkness), it is essential that the sampling times and dosing times are reported in relation wto the Light-Dark schedule, as Zeitgeber Time.
(3) Because the underlying assumption in the study is that the "chronic jet lag protocol" suppresses the circadian organisation, such timing reference is of a lesser importance if the assumption is proven in (1). However, it would be useful to know when the samples were taken in relation to the effective light or dark span the mice were exposed to. (4) Throughout the manuscript, there are some imprecisions, that are addressed in my specific comments (5) English should be improved.

Specific comments
Results P3, last sentence refers to "slight tendency to observe more malignant lesions in JL mice". Table  S1 and Fig S2 illustrate this sentence for 8 mice only (4 LD and 4 CJL). On which criteria were these 8 mice selected for pathology among the total of 46 that were on study? P4, lines 8-10: a bone lesion is by definition abnormal. If the bone lesions were histologically proven as being metastatic deposits, better to say it as such. P4, Line 11: rather say: "the proportion…increased from… to…" P4, line 8 before the end: The down regulation of the light perception and phototransduction genes in the bone marrow mononuclear cells in "chronic jet lagged" mice needs to be interpreted against time-qualified expressions in controls. Is the expression of these genes known to be rhythmic? How large is their amplitudes? P5, line 5: better say "statistically significant" rather than "slight" which implies a subjective interpretation of a non statistically significant difference, which is not the case according to Fig 4. P5, line 24: when it comes to clock genes expression, it is crucial at least to report the ZT sampling time in the LD12:12 mice, and whether the sampling time in the "chronically jet lagged" mice also occurred at a similar time in the light or dark span as in the controls. P5, second to last par.: You should speak of circadian clock disruption, but not of Per genes specifically, because there are no data in Per KO. P6: sampling times are needed throughout…. P6, first line. Can you provide a brief statement as to why the reduced P6, line 11: Can you discuss the evidence relating tumour CD4/CD8 to prognosis, and what it means in terms of the immunologic control of tumours. Isn't it enough to consider the CD8+ cells (suppressor/cytotoxic) that infiltrate the tumour? P6, last sentences: How often was the CXCR2 inhibitor injected, through which route, at which ZT time/clock hour? Regarding timing, also see: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4967945/ First of all, we thank the two reviewers for their helpful comments and suggestions to improve the global quality of the manuscript.
Reviewer #1 (Remarks to the Author): Expertise in breast cancer (in vivo) genomics and transcriptomics 1) The methods do not detail the background of the MMTV-PyMT mice, nor do they describe how the tumor studies were completed. The primary issue that, due to the lack of detailed methods, needs to be resolved is why the frequency of metastasis in LD PyMT mice is so low. Numerous previous studies have shown that virtually 100% of PyMT mice on the FVB background develop pulmonary metastases, not the 30(ish)% they show.
Indeed, this important point was missing from the manuscript. We provided details and complete description in the main text (P3), methods (P10) and in Fig. 1A. We did not use a pure PyMT FVB background because it was too aggressive and not compatible with our long-term chronic CRD protocol. We decided to use a mixed B6*FVB PyMT background. Using this background, we observed a delayed onset of tumour development and slower progression ( Fig. 1A) with low prevalence (c.a. 30%) of lung metastasis at the age of 16 weeks.
2) Color blind readers simply cannot see figures that are red-green. Figure 2C is an excellent example of a figure that does not need to be red-green, but because it is, it cannot be interpreted. It literally appears as the same shade for both top and bottom. Black and grey would work. Or simply a bar that shows % with metastasis. Other problem figures include: Figure  Also, supplemental data -please just change everything that is red/green. We thank the reviewer for drawing our attention to this point. We have changed the colours of the figures to make them interpretable for colour blind readers.
3) Figure 2D should have dots representing the number of metastatic foci, not just bins of categories. Please redo statistical tests using all the data. In addition, representative images should be shown.
We chose to represent our data in categories due to high variation. As you can see on the graphs below, two mice were detected with extremely high number of metastatic foci. Possible, these two mice reached the exponential late tumour growth phase. However, the tumour burden data suggest that only one of the mice reached the late phase. These samples were not excluded because: 1) From the same cohort/litter neither the 2 other JL nor the 3 LD mice developed lung metastasis 2) Tumour burden did not correlate with number of foci (highlighted on the image below) 3) None of the downstream analysis showed deviation from the median Please note the Figure 2D and its statistic have been done using all data. Fig 4B) together with the distribution of metastatic foci number between groups, including the graph shown here.

Representative images can be found in Supplementary data (Supplementary
The statement that PC1 separates tumors with / without metastasis is unfounded based on the data in Figure 3B. In addition, there are far too few data points to even try this. This portion of the figure should be cut. We modified the figure and the text. From the previous figure 3B, PC1 separates primary tumours with/without metastasis only from JL mice (in red) but not from LD mice (black). We agree that this graph is a bit misleading and that the number of samples is reduced and cannot allow general conclusions to be drawn. We removed the PCA from the main figure and modified the text accordingly. Figure 4A is not convincing for an increase of stem cells in the jet lag mice. P-values seem to be largely driven by a handful of outliers in the control population.
We would like to clarify that the stem population was identified based on the mouse mammary stem cell (MaSC) signature (Fig.4B) and not based on the individual expression of stemness markers. Importantly, we evaluated stemness by functional assays (mammosphere formation and tumour initiation). Both data confirmed our observation about enriched MaSC compartment in the JL tumours, which led us to the conclusion that CRD promotes stemness of primary tumour cells. We agree that our data sets show relatively high variability but this is considered normal in in vivo data sets. Important to note that  standard deviation (SD) /variability is not consistently different between LD vs JL group and LD SD is not consistently higher compared to JL SD  Data sets were tested to identify outliers (ROUT method Q=1%, GraphPad): CD24% data set: one outlier in LD group CD29% data set: one outlier in JL group No other outliers were detected which also confirms that the observed variability is not due to a specific group of mice. Elimination of the two outliers do not alter the statistical analysis results, therefore we decided to keep both outliers. Figure 4F is based on 4 tumors from LD and 4 from JL -but NO characterization of the primary tumor was provided. There are numerous studies illustrating the heterogeneity in histology from PyMT mice. This was not taken into account and so the authors may be comparing spindleoid myoepithelial tumors to lobular epithelial tumors, with obvious bias in the percentage of potential tumor initiating cells.
We agree with the reviewer that this is an important point. The histology of the primary tumours used for the tumour initiation study (6 primary tumours from LD and 6 primary tumours from JL mice) were not included in the first version of the manuscript. We did not see striking anatomical differences when we dissected the tumours to be used for the tumour initiation study but to confirm this empirical observation we also performed and included HES staining of the tumours we used in the revised manuscript. The tumours have been analysed by a veterinary pathologist (Isabelle Raymond Letron, professor in histology and pathology at the University of Toulouse and a recognized expert in veterinary pathology). The information is shown on Supplementary Table 2. Unsurprisingly, JL tumours are slightly more aggressive (Supplementary Fig.  3) but JL and LD tumours were quite homogenous and all arise from epithelial cells. We did not observed spindleoid myoepithelial tumours in these samples. CXCR2 inhibitors have previously been shown to inhibit metastasis. However, the experimental design in Figure 6 is lacking an essential control -no LD mice were tested +/-CDCX2 inhibitor to determine if the response for the JL and LD mice was altered. Indeed, given the work of Halpern et al (PMID 21601983), this control is essential. Also, it is essential to place the work into the context of this prior work.
We agree that we have to However, it is important to note that our flow cytometry analysis showed no difference in the proportion of CXCR2+ primary tumour cells between LD and JL mice ( Supplementary Fig. 9C). In addition, data from Luminex assay did not show altered LIX/CXCL5 level in plasma from JL mice compared to the control group. All together these data suggest that in our model CXCR2/CXCL5 axis does not drive directly the invasion and intravasation of tumour cells from the primary tumour site. Based on these observations we expected that CXCR2 inhibition primarily decreases metastasis through blocking myeloid cell recruitment and consequently improving anti-tumour immunity / reducing the development of pre-metastatic niche as it have been shown before ( We agree that the use of CXCR2 inhibition is not specifically counteracting the effects of JL but we propose it as a possible complementary therapy to control the speed-up of tumour progression in CRD conditions. This point was included in the discussion (P10).
Reviewer #2 (Remarks to the Author): Expertise on circadian rhythm and cancer General comments The biomedical applications of circadian clocks represent a critical challenge for medical progress, especially for cancer. More specifically, an increased risk of breast cancer has been shown in women undergoing prolonged shift work, and this environmental condition was acknowledged as a likely cause of cancer by the International Agency for Research on Cancer both in 2007 and in 2019. Several other reports show that circadian disruption also impacts on the outcomes of tumour bearing rodents, as well as large cohorts of cancer patients. Taken together, the existing literature emphasizes the need for a better understanding of the mechanisms linking circadian disruption and carcinogenesis. This manuscript provides highly interesting and innovative data in this regard.
In aggregate, Hadadi et al. highlight the impact of iterative daily schedule shifts ("chronic jet lag protocol") for the metastatic dissemination of breast cancer in an experimental model of spontaneous mammary carcinogenesis. They identify key mechanisms at work at the chemokine/cytokine network level, and within the tumour infiltrating immune cells. They show that a CXCR2 inhibitor could prevent the deleterious effect of iterative schedule shifts on breast cancer dissemination. The data presented are new and convincing, yet several issues deserve to be answered.
Main comments (1) As mentioned by the authors, the jet lag protocol that has been applied in their study has proven its ability to suppress circadian rhythms in rest-activity and core body temperature as well as several clock genes expressions, both in the SCN and in peripheral tissues of male B6D2F1 mice. Although it is highly likely that similar effects would be observed for the transgenic mice used in this study, it is needed to report the circadian phenotype of these mice when kept both in usual light-dark conditions (LD12:12), and on the "chronic jet lag" protocol. (2) Because the vast majority of the parameters studied undergo large endogenous circadian variations in mice on LD12:12 (as well as in constant darkness), it is essential that the sampling times and dosing times are reported in relation to the Light-Dark schedule, as Zeitgeber Time.
(3) Because the underlying assumption in the study is that the "chronic jet lag protocol" suppresses the circadian organisation, such timing reference is of a lesser importance if the assumption is proven in (1). However, it would be useful to know when the samples were taken in relation to the effective light or dark span the mice were exposed to.
We agree with the reviewer and as mentioned before, we validated our "chronic jet lag protocol " using locomotor activity and core body temperature measurement. Our data confirmed the circadian disruption effect of the applied jetlag protocol in our transgenic model. For LD mice, and in response to (2) and (3): we clarified this in the methods and added the respective ZT times.
(4) Throughout the manuscript, there are some imprecisions, that are addressed in my specific comments The imprecisions have been corrected. See our answers to specific comments (5) English should be improved. The manuscript was reread by a professional scientific writer.

Specific comments
Results P3, last sentence refers to "slight tendency to observe more malignant lesions in JL mice". Table S1 and Fig S2 illustrate this sentence for 8 mice only (4 LD and 4 CJL). On which criteria were these 8 mice selected for pathology among the total of 46 that were on study?
We selected the mice from random selection based on sample availability. Histology was performed in the middle of cohort collection, 8 samples were picked from 20 mice. We agree that this number was low and we analysed the histology of tumours from 9 additional mice and completed the previous table S1 (now Supplementary  Table 2). The new dataset confirms and strengths the increased aggressiveness of tumours from JL mice (see new graph on Supplementary Fig. 3) P4, lines 8-10: a bone lesion is by definition abnormal. If the bone lesions were histologically proven as being metastatic deposits, better to say it as such. Addressed and we modified the text accordingly.
P4, Line 11: rather say: "the proportion…increased from… to…" We corrected the text.
P4, line 8 before the end: The down regulation of the light perception and phototransduction genes in the bone marrow mononuclear cells in "chronic jet lagged" mice needs to be interpreted against time-qualified expressions in controls. Is the expression of these genes known to be rhythmic? How large is their amplitudes? This is an important point and we were really surprised to observe down regulation of phototransduction genes in internal tissues and cells. Indeed, we can not exclude that all these genes harbour the same circadian rhythmic expression peaking around ZT3-ZT4, when LD mice were sampled. In this case, it is logical that these genes appeared downregulated in JL mice (in relation with the amplitude of their rhythmic expression), where the core circadian clock is disrupted. This important point has to be confirmed, as little is known about the expression/function of these genes in peripheral non-visual tissues. In our mind it would be strange that all these genes express the same rhythmicity (if they are rhythmic) but we modified the discussion in relation to this point. Moreover another important question relies on the biological consequences of such downregulation.
P5, line 5: better say "statistically significant" rather than "slight" which implies a subjective interpretation of a non statistically significant difference, which is not the case according to P5, line 24: when it comes to clock genes expression, it is crucial at least to report the ZT sampling time in the LD12:12 mice, and whether the sampling time in the "chronically jet lagged" mice also occurred at a similar time in the light or dark span as in the controls. We corrected the text.
P5, second to last par.: You should speak of circadian clock disruption, but not of Per genes specifically, because there are no data in Per KO. Addressed and we modified the text accordingly.
P6: sampling times are needed throughout…. We added ZT, collection times.
P6, first line. Can you provide a brief statement as to why the reduced Most probably, in our experimental system the reduced number of TICs were due to the disrupted diurnal trafficking of leukocytes. As Zhao et al 2017 showed under circadian disruption leukocytes lost their rhythmic trafficking leading to a decrease in the daily total number of circulating leukocytes. Other mechanisms linked to increased tumour burden (e.g. hypoxia or necrosis) could also contribute to this reduction. We added a sentence in the discussion (P9). Number of TILs can be highly variable between individuals therefore additional use of CD4/CD8 ratio can provide cleaner results and further support the CD8+ TIL data (this is more obvious on the peripheral blood data Fig. 5F). We added a sentence in the results' section to clarify this point (P6).
P6, last sentences: How often was the CXCR2 inhibitor injected, through which route, at which ZT time/clock hour? Regarding timing, also see: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4967945/ The injection protocol is detailed in method section: through 8 weeks, on each week intra peritoneal injection was performed once daily 5 days in a raw followed by 2 days resting (adapted from Acharyya et al 2012 DOI: 10.1016/j.cell.2012.04.042) LD mice were injected at ZT6-ZT7, at the previously described peak of circulating CD45+ cells (Zhao et al 2017 DOI 10.1182/blood-2017-04-778779). Regarding the timing in JL mice: based on the same study from Zhao et al 2017, which showed no significant circadian oscillation of leukocyte trafficking in mice exposed to chronic jetlag, we did not apply specific ZT time for injection. However, mice were consistently injected at the same time of the day to keep daily dosing. We clarified this in the method section. This reviewer was asked to comment on the immunological aspects of the work. However, given that this work falls on my area of expertise, I also made comments on the rest of the manuscript. Major: Photon flux is acceptable to show tumor initiation, but it does not accurately reflect tumor growth: Luciferase is driven by MMTV in a constant manner, independently of whether those cells become transformed and contribute to tumor formation. In other words, many of those Luciferase-positive cells will not form tumors. Moreover, the luciferase signal becomes rapidly saturated in this model, and does not allow accurate quantification of tumor growth. Tumor volume should be used instead.
Minor: It is unclear why in the last panel (1F) the value is a % between tumor weight and mouse weight. It is customary to plot combined tumor weight per mouse when using transgenic mice.
The way Figure 1 is presented is confusing, and it gives the wrong impression that the jet-lag model was applied to the mixed-background, while the FvB mice were kept under regular darklight conditions. Please clarify.
The tumor growth kinetic difference between the two backgrounds is known and established, and the reference is enough to justify their election. It is my suggestion to eliminate this panel, to avid confusion on the utilized models.
It is not clear what "mixed background" refers to. If this is the F1 between the FvB and B6 mice, in which there is 50% of each background, this would be acceptable, but not if a random number of crosses was done between the 2 strains.      However, their gating strategy does not follow any of the ones described in the references. While there are many different ways to analyze the TIL populations, the main issue is the order of gating, using less cell type-specific markers first, and eliminating important populations from subsequent analysis with more accepted markers. For example, the first gating after identifying hematopoietic cells (CD45+) should be CD11B and CD3/TCRB to identify myeloid cells and T cells, respectively, and then continue gating on each individual population. A non-exhaustive list of other issues: *While the CD45+ CD11B+ Gr1+ cell population contains neutrophils, not all cells here are neutrophils, and therefore this is not how they should be named. In fact, this is how the community defines immature myeloid cells, also called myeloid-derived suppressor cells (MDSCs). *The sub classification of TAMs based on MHCII levels is usually done after discrimination of Ly6C (monocytic) cells, as properly depicted in Movahedi et al. *Monocytes are defined as CD45+ CD11B+ Ly6C+ cells. There is no Ly6C staining done in this analysis, and therefore monocytes cannot be properly identified. *NK cells are usually defined as NK1.1+ cells, and separated in NK or NKT based on their expression of the T cell receptor. Here, NK cells are identified as CD45+ CD11B+Gr1-CD11C+MHCII-CD64+cells…none of these markers are Nk-specific, and seem more of a random collection. *B cells are named, but there are no B cell specific markers (like CD20 or B220) Etc… While it is still possible that there are changes consistent with a more pro-tumorigenic immune microenvironment, the cell populations need to be better defined. Figure 5E-F It is well established that increased CD8T cell infiltration is a marker of better prognosis in breast (and other cancers). It is also well established that CD4+ Foxp3+ Treg cells are a marker for poor prognosis in breast (and other cancers). The correlation of CD4T cells without any additional marker is, however, very controversial. This is simply because CD4T cells are heterogenous: CD4+ Foxp3+ Treg cells are a marker of bad prognosis, TH2 CD4T cells induce alternative activation of macrophages, which is associated with poor prognosis, and TH1 CD4T cells are anti-tumorigenic, to name a few. Therefore, Treg cell frequencies and their relative ratio with respect to CD8T cells, would be a better correlate for outcome than generic CD4/CD8 ratios.
Supplementary Figure 5 With exception of IL-4, none of the other changes are significant. The comment of the elevation of CXCL12 levels in the JL mice should be removed from the text, because equally irrelevant observations in the opposite direction can be made from that table (For example increase in IL-12, which is an inducer of TH1 responses, antitumorigenic phenotypes). Furthermore, IL-4 is a critical cytokine inducing alternative activation of tumor-associated macrophages, which argues the JL tumors have a tumor-promoting immune phenotype. Unfortunately, this data not only does not help the main conclusion the authors are trying to make, but also argues against it. The authors would have been better off discarding this due to the high biological variability observed.

Figure 6
Major Expression data shown in Figure 6A (from RNASeq experiment) does not indicate which changes are statistically significant, nor what cut-off value was used for the analysis. Several of these transcripts do not seem to be significantly changed, like CXCL11 and CXCL9. Interestingly, those are bona fide and highly sensitive targets of IFNg, which the authors claim is significantly downregulated (p values?). Other transcripts with opposing functions look like they could possibly be significantly upregulated in both models. For example, IL1s (alpha and beta -pro-tumorigenic factors) and IL-10 (immunesuppressive factor) seem upregulated in JD mice. All of these data seem to be "selectively" interpreted to fit the proposed hypothesis. Perhaps a pathway analysis could provide more unbiased support to their hypothesis?
IN SUMMARY: the rationale for selecting CXCR2 for further studies is convoluted and lacks rigor. The authors should make the effort to much better justify the target.

Major
The fact that the CXCR2 inhibition works to the same degree in jet-lagged mice as in non jetlagged mice suggests that this pathway is not selective to circadian regulation of metastatic behavior in these mice. If it is not, then we are still lacking an explanation for the main observation of the paper. The discussion is not sufficient to explain this fact.

Reviewer #1 (Remarks to the Author):
Revised manuscript addresses all points satisfactorily.

Reviewer #2 (Remarks to the Author):
My comments and questions have been addressed properly well.

Reviewer #3 (Remarks to the Author):
This reviewer was asked to comment on the immunological aspects of the work. However, given that this work falls on my area of expertise, I also made comments on the rest of the manuscript. Major: Photon flux is acceptable to show tumor initiation, but it does not accurately reflect tumor growth: Luciferase is driven by MMTV in a constant manner, independently of whether those cells become transformed and contribute to tumor formation. In other words, many of those Luciferase-positive cells will not form tumors. Moreover, the luciferase signal becomes rapidly saturated in this model, and does not allow accurate quantification of tumor growth. Tumor volume should be used instead.
We do not have the data on tumour volume for these mice. We clearly stated in the main text that we used in vivo bioluminescence to monitor tumor growth (p.4). We agree that luciferase measurement does not allow accurate quantification of tumour growth and that it is possible that cells expressing luciferase are not all actively proliferating and contributing to tumour development.
However, we also observed gradual increase of signal with tumour growth supporting the idea that the majority of signals are coming from proliferating cancer cells (see the dedicated Figure below showing luciferase imaging in the same mice at different times of tumor development). Moreover, in our experimental design, we did not observe saturation of the luciferase signal mostly because we focused on early phases of tumor development. In addition measuring luciferase in vivo also allowed us to monitor tumor growth even when, as was sometimes the case, palpable tumors appeared late (14 weeks) and when we observed non-homogenous tumour growth in-between mammary fat pads.

Minor:
It is unclear why in the last panel (1F) the value is a % between tumor weight and mouse weight. It is customary to plot combined tumor weight per mouse when using transgenic mice.
We decided to represent total tumour weight as tumour burden to remove the potential biased effect of mouse weight. CRD was shown to result in increased weight gain/obesity (Van Dycke et Fig 1C).

Like Fig.1G, the figure below without body weight normalisation illustrates a significant increase of total tumour weight in JL mice compared to LD mice (unpaired t-test):
The way Figure 1 is presented is confusing, and it gives the wrong impression that the jet-lag model was applied to the mixed-background, while the FvB mice were kept under regular darklight conditions. Please clarify.
We totally agree and we simplified the graph on Figure 1A to avoid confusion.
The tumor growth kinetic difference between the two backgrounds is known and established, and the reference is enough to justify their election. It is my suggestion to eliminate this panel, to avid confusion on the utilized models.
We simplified the graph on Figure 1A to avoid confusion.
It is not clear what "mixed background" refers to. If this is the F1 between the FvB and B6 mice, in which there is 50% of each background, this would be acceptable, but not if a random number of crosses was done between the 2 strains.

Minor:
It is not clear what the point of showing bone marrow mets is. This data is not quantified, and therefore it does not contribute to the differential effect on metastasis the authors are trying to establish. Consider removing.

As the PyMT model is not a classical bone metastasis model we aimed to further confirm our flow cytometry and qPCR data on tumour cell homing and colonisation to the bone.
Supplementary Tables 3 and 4 are missing.
Supplementary Tables 3 and 4 are not in the supplementary information but are provided as supplementary data.

Figure 3
Major Very weak figure. There is no significant conclusion drawn, nor use of this data. Maybe supplementary? As is, it only disrupts the flow of the manuscript and does not contribute in any meaningful way.
We agree that the mRNA-seq study did not bring clear global information, due to the low number of DEGs between conditions. We simplified the figure and focused on phototransduction genes because this point seems interesting to us.  Previous studies linked circadian clock genes' function to stemness. Indeed, downregulation of Per2 in MCF10A breast cell line increased cells' stemness. However, this experiment does not make it possible to discriminate whether the phenotype results from an alteration of the circadian clock linked to the downregulation of Per2 or to a direct effect of Per2 on the stemness of breast epithelial cells. Our results showed that the phases of the functional circadian clock are intrinsically able to modulate the stemness of human mammary epithelial cells ( Figure 4E). These two aspects are complementary and consistent: we observed here that human breast cells present more stemness during the "night" phase when Per2 is low and Bmal1 is high and their stemness decreases during the "day" phase, when Per2 is high and Bmal1 is low. Figure 4F poses a problem: if JL treatment increases stemness, tumor initiation and growth, then it is not clear why the primary tumors in the PyMT mice are not significantly affected, but only metastasis.
Indeed, we observed more drastic changes in metastatic spread but primary tumours were also affected. Primary tumours from JL mice were significantly bigger ( Figure 1G, see also the total tumour weight graph provided above) and they were classified with significantly higher tumour grades (Supplementary Fig.3

Major
In identifying the references for their flow cytometrical analysis of tumor-infiltrating leukocytes, the authors correctly point to two landmark papers pioneering this classification in murine breast tumors, and more specifically PyMT tumors: Movahedi et al, Cancer Research, 2010;and Franklin et al, Science, 2014. However, their gating strategy does not follow any of the ones described in the references. While there are many different ways to analyze the TIL populations, the main issue is the order of gating, using less cell type-specific markers first, and eliminating important populations from subsequent analysis with more accepted markers. For example, the first gating after identifying hematopoietic cells (CD45+) should be CD11B and CD3/TCRB to identify myeloid cells and T cells, respectively, and then continue gating on each individual population.
A non-exhaustive list of other issues: *While the CD45+ CD11B+ Gr1+ cell population contains neutrophils, not all cells here are neutrophils, and therefore this is not how they should be named. In fact, this is how the community defines immature myeloid cells, also called myeloid-derived suppressor cells (MDSCs).
*The sub classification of TAMs based on MHCII levels is usually done after discrimination of Ly6C (monocytic) cells, as properly depicted in Movahedi et al.
*Monocytes are defined as CD45+ CD11B+ Ly6C+ cells. There is no Ly6C staining done in this analysis, and therefore monocytes cannot be properly identified.
*NK cells are usually defined as NK1.1+ cells, and separated in NK or NKT based on their expression of the T cell receptor. Here, NK cells are identified as CD45+ CD11B+Gr1-CD11C+MHCII-CD64+cells…none of these markers are Nk-specific, and seem more of a random collection. *B cells are named, but there are no B cell specific markers (like CD20 or B220) Etc… While it is still possible that there are changes consistent with a more pro-tumorigenic immune microenvironment, the cell populations need to be better defined.
Thank you, indeed this section was missing essential information.
First, we would like to clarify that we made a significant mistake as we missed out to cite the reference paper of our gating strategy in our previous submitted manuscript version. We corrected this. Our primary gating strategy of tumour infiltrating leukocytes is based on the publication of Yu et al. 2016 (doi: 10.1371/journal.pone.0150606). We chose this method as it provides one relatively simple panel to identify the main immune cell types, which was adjustable to our 8-colour flow cytometer and allowed to perform complete characterisation even in case of limited sample size. We agree the used method is less accurate for lymphoid populations and it is a rather unorthodox approach. However, it showed to work well to identify different myeloid cell populations in non-lymphoid tissues including PyMT mammary tumours (Ye et al. 2016). As based on this gating strategy we only concluded information for tumour associated macrophages and we are convinced that the strategy to define TAM populations is correct.
Indeed, we modified the main text and figures to be more precise on gate naming.

Regarding NK, T, B cells we did not use specific markers for main immune cell characterisation based on Yu et al gating strategy where the gated cells were confirmed by specific markers.
To identify T cell phenotypes we used a separate antibody panel based on classical markers. Gating has been added to Supplementary Figure 7B.
We also provided a detailed antibody list (supplementary Table 7).

Figure 5E-F
It is well established that increased CD8T cell infiltration is a marker of better prognosis in breast (and other cancers). It is also well established that CD4+ Foxp3+ Treg cells are a marker for poor prognosis in breast (and other cancers). The correlation of CD4T cells without any additional marker is, however, very controversial. This is simply because CD4T cells are heterogenous: CD4+ Foxp3+ Treg cells are a marker of bad prognosis, TH2 CD4T cells induce alternative activation of macrophages, which is associated with poor prognosis, and TH1 CD4T cells are anti-tumorigenic, to name a few. Therefore, Treg cell frequencies and their relative ratio with respect to CD8T cells, would be a better correlate for outcome than generic CD4/CD8 ratios.
Indeed, we agree that it is more informative to use CD4+FoxP3+ Treg to CD8T cell ratio and it has a stronger prognostic value. We did perform FoxP3 immunostaining on primary tumours and we added the panel to the revised version ( Figure 5F). Due to sample size limitation this staining was not performed on PBMCs and on tumours from CXCR2 inhibitor study. Gating strategy is presented in Supplementary Figure 7B.
Supplementary Figure 5 With exception of IL-4, none of the other changes are significant. The comment of the elevation of CXCL12 levels in the JL mice should be removed from the text, because equally irrelevant observations in the opposite direction can be made from that table (For example increase in IL-12, which is an inducer of TH1 responses, antitumorigenic phenotypes). Furthermore, IL-4 is a critical cytokine inducing alternative activation of tumor-associated macrophages, which argues the JL tumors have a tumor-promoting immune phenotype. Unfortunately, this data not only does not help the main conclusion the authors are trying to make, but also argues against it. The authors would have been better off discarding this due to the high biological variability observed.
We agree that globally, these results were disappointing and are only shown on a supplementary figure. Following the reviewer's suggestion we removed the comment about CXCL12 level as indeed it did not provide additional information to our results.
Regarding IL-4 data, we are aware that it is a critical cytokine to induce TH2 type immune responses but we would like to highlight several point: 1) Plasma cytokine/chemokine levels are not necessarily representing the tumour microenvironment.
2) Plasma level of several cytokines/chemokine including IL-4 shows circadian fluctuation (most recent publication: 10.1038/s41598-019-56951-5). Cortisol, which is able to increase IL-4 plasma level is also under circadian regulation. Altogether, these suggest that the observed decrease can be due to circadian disruption.
3) IL-4 plasma level showed only a moderate decrease in JL mice (0.8 fold), which might be well within the range of its circadian fluctuation. It is unreasonable to think that this level of decrease would completely impaired TH2 type immune responses.

Figure 6
Major Expression data shown in Figure 6A (from RNA-Seq experiment) does not indicate which changes are statistically significant, nor what cut-off value was used for the analysis. Several of these transcripts do not seem to be significantly changed, like CXCL11 and CXCL9. Interestingly, those are bona fide and highly sensitive targets of IFNg, which the authors claim is significantly downregulated (p values?).
Other transcripts with opposing functions look like they could possibly be significantly upregulated in both models. For example, IL1s (alpha and beta -pro-tumorigenic factors) and IL-10 (immune-suppressive factor) seem upregulated in JD mice.
All of these data seem to be "selectively" interpreted to fit the proposed hypothesis. Perhaps a pathway analysis could provide more unbiased support to their hypothesis?
IN SUMMARY: the rationale for selecting CXCR2 for further studies is convoluted and lacks rigor. The authors should make the effort to much better justify the target.
We addressed this comment by rewriting the cytokine-chemokine network result section (page 7). We also modified Figure 6A by adding cut-off values to the boxplot and we added a supplementary table (supplementary Table 6) Table 6).
We provide below a more logical and cohesive reasoning for the rationale behind our study flow and target selection.
In summary, CXCL5, which happened to be upregulated in both primary tumours and BM of JL mice ( Figure 6A and Altogether, this information led us to identify the CXCR2 axis as a potential underlying mechanistic target of CRD.

Major
The fact that the CXCR2 inhibition works to the same degree in jet-lagged mice as in non jetlagged mice suggests that this pathway is not selective to circadian regulation of metastatic behavior in these mice. If it is not, then we are still lacking an explanation for the main observation of the paper. The discussion is not sufficient to explain this fact.
We did not claim that CXCR2 driven metastatic processes are selective or specific to CRD induced alterations. This mechanism is well described and reported previously in normal circadian conditions (Steele et al  Here we showed that CRD leads to an increased metastatic spread associated with an upregulation of the CXCR2/CXCL5 axis and consequently CXCR2 driven metastatic burden. The use of CXCR2 inhibitor in JL mice helps to decrease cancer cell dissemination and metastasis formation and even though CXCR2 inhibition presents similar effects in LD mice, we propose that CXCR2 inhibition is important to counteract the adverse effects of JL and to slow down tumour progression. The use of CXCR2 inhibition could help when used in combination with conventional chemotherapy to improve therapy outcome in patients with circadian rhythm disrupted tumours. Major novelties of our study: we showed that CRD increases metastatic spread and we described potential underlying mechanisms: 1) increased stemness and tumour initiation capacity and 2) immunosuppressive shift in the tumour microenvironment driven at least by enhanced CXCR2 axis.
The majority of the issues raised by this reviewer were acceptably addressed. This includes comments on Figures 1-4, and Figures 5E-F.
However, some issues remain ( Figure 5 -flow data) and Figure 6. In addition, the authors present valuable arguments only in the rebuttal for some points raised by this reviewer, but these arguments have not been incorporated in the main text. The prime example of this is the prominent place the blood cytokine screen still has in the main text (although the data was sent to supplementary). The obvious explanation authors use in their rebuttal, about levels in the blood not been as informative as levels in the primary tumors, has not been used to justify looking at the gene expression levels (but it should be).
Immunophenotyping. IV. Monocytes should be CD64+ CD11B+ >here, the gating is covered by legends and it's difficult to appreciate, but the CD64 stain is extremely dim and the populations look very different than the same tumors gated in the proposed Ye et al reference.
Without a direct comparison between JL and LD conditions, the claim that inhibition of CXCR2 is responsible for the shift in the immune tumor microenvironment is difficult to make. If the treatment inhibits metastasis in both conditions to a similar degree, which it does, then it is not clear what is specific to CDR conditions. The argument that this treatment could provide "a novel therapeutic tool to thwart the effects of CRD on tumor progression" is flawed if there is no CRDspecificity. In other words, any LD-unrelated treatment that affects metastasis could have been used with the same results. Short of showing a more pronounced effect of the CXCR2 inhibition under JL conditions, I don't see how CXCR2 can be made responsible for CRD-dependent effects.
We sincerely thank the last reviewer for helping us to correct our flow cytometry data and thus to improve the quality and credibility of our research.

Reviewer #3 (Remarks to the Author):
The majority of the issues raised by this reviewer were acceptably addressed. This includes comments on Figures 1-4, and Figures 5E-F.
However, some issues remain ( Figure 5 -flow data) and Figure 6. In addition, the authors present valuable arguments only in the rebuttal for some points raised by this reviewer, but these arguments have not been incorporated in the main text. The prime example of this is the prominent place the blood cytokine screen still has in the main text (although the data was sent to supplementary). The obvious explanation authors use in their rebuttal, about levels in the blood not been as informative as levels in the primary tumors, has not been used to justify looking at the gene expression levels (but it should be).
To clarify this point, we included this sentence in the main text: "Since plasma cytokine/chemokine levels are not necessarily representing the tumour microenvironment, we used the data of our transcriptomic study and real-time PCR to assess the expression levels of cytokines/chemokines and their receptors in primary tumours.".
We decided to mention the Luminex assay in the main text despite the lack of statistical differences between conditions because we think that this data still has information value. And taken that there are no marked differences between LD and JL mice this data was presented in the supplementary information from the beginning. Immunophenotyping.
1. The paper used to define the gating strategy for the different immune populations is not an obvious one. Indeed, as the authors recognized, quite unorthodox. This was explained to the reviewer, yet it is missing from the manuscript. Please add this reference in lane #211 and make it very clear to the reader. Moreover, remove the references from Movahedi and Franklin (#30 and #31), because leaving them in that section is extremely misleading to the unaware reader. Moreover, ref#30 (Movahedi et al) is used to justify pro-and anti-tumor TAM definition, but it is incorrect, since TAMs are not defined in the same way in this work. Again misleading and inappropriate.
The gating strategy is now clearly stated in the main text and referenced throughout. 2. Some of the populations do not follow the gating strategy proposed in the Ye et al 2016 paper, critically, the TAM cell population. As per the paper of reference: I. TAMs should be defined as CD64+CD24+ or CD24-> here they are simply defined as CD24-, therefore including CD64-cells and excluding most of the CD24+ cells. From there on, the remaining derived populations need to be fixed (including MHCII high and low) We corrected our gating regarding CD64. According to this, we defined macrophages as CD64+CD24-cells. Gating on this population we identified anti-and pro-tumour macrophages as CD11b+MHCIIhi or CD11b+MHCIIlow TAMs respectively. All represented data was corrected (Fig. 5B-D, Supp.Fig 7 and 8).
II. B cells should be CD24+ > here all "B" cells are CD24 negative Indeed, we did not observe similar CD24 staining on the suspected B cell population as was reported by Yu et al. However, it is important to note that B cells are constitutively expressing MHCII, while murine T cells do not express this marker. Therefore, we believe that MHCII alone is sufficient to differentiate between these two populations.
III. Eosinophils are CD11B+ cells >here, the Eos gate includes both CD11B + and CD11B negative cells.
Gate was corrected and corresponding data was updated ( Figure 5B).
IV. Monocytes should be CD64+ CD11B+ >here, the gating is covered by legends and it's difficult to appreciate, but the CD64 stain is extremely dim and the populations look very different than the same tumors gated in the proposed Ye et al reference.
Legends were rearranged, the gating is clearly visible now. Indeed, in our hand these populations look different but we consistently observed this profile. Please find below a figure showing the mentioned plot compared to the isotype control. Here you can see that we have obvious positive staining (respective gated populations overlaid with isotype control). Potentially it might be that using different fluorochrome combinations results in different resolutions of the targeted populations. Observing a prominent third population ('CD11c+T cells) might be explained by the significant difference in sampling time and consequent tumour stage: in Yu et al study the analysed tumours were harvested at much later stage (around 6 months/24 weeks of age compared to 16 weeks of age in our study).

CXCR2 inhibition.
Without a direct comparison between JL and LD conditions, the claim that inhibition of CXCR2 is responsible for the shift in the immune tumor microenvironment is difficult to make. If the treatment inhibits metastasis in both conditions to a similar degree, which it does, then it is not clear what is specific to CDR conditions. The argument that this treatment could provide "a novel therapeutic tool to thwart the effects of CRD on tumor progression" is flawed if there is no CRD-specificity. In other words, any LD-unrelated treatment that affects metastasis could have been used with the same results. Short of showing a more pronounced effect of the CXCR2 inhibition under JL conditions, I don't see how CXCR2 can be made responsible for CRD-dependent effects.
We agree that the small number of mice used in our experimental design using the CXCR2 inhibitor does not allow us to observe clear differences between JL and LD mice. But we provided data showing enhanced accumulation of suppressor myeloid and T cells in primary tumours and increased dissemination under JL. Which suggests that the CXCR2 inhibition in combination with conventional therapy could be more beneficial (but not exclusively) to patients with CRD / CRD tumours where the CXCR2 driven mechanisms are accelerated.

Indeed we can only conclude that the shift in immune microenvironment (improved CD4/CD8 ratio and reduced myeloid infiltration) upon CXCR2 inhibition is present in both
We re-write our conclusion about CXCR2 inhibition in CRD tumours and made it clear that it's effect is not CRD specific but it could be especially beneficial for patients with CRD.