Hypoxia induces senescence of bone marrow mesenchymal stem cells via altered gut microbiota

Systemic chronic hypoxia is a feature of many diseases and may influence the communication between bone marrow (BM) and gut microbiota. Here we analyse patients with cyanotic congenital heart disease (CCHD) who are experiencing chronic hypoxia and characterize the association between bone marrow mesenchymal stem cells (BMSCs) and gut microbiome under systemic hypoxia. We observe premature senescence of BMSCs and abnormal d-galactose accumulation in patients with CCHD. The hypoxia that these patients experience results in an altered diversity of gut microbial communities, with a remarkable decrease in the number of Lactobacilli and a noticeable reduction in the amount of enzyme-degraded d-galactose. Replenishing chronic hypoxic rats with Lactobacillus reduced the accumulation of d-galactose and restored the deficient BMSCs. Together, our findings show that chronic hypoxia predisposes BMSCs to premature senescence, which may be due to gut dysbiosis and thus induced d-galactose accumulation.

function in their models. Major comments: 1-Did the CCHD patients have any lactose intolerance? Is this something that has been reported in the literature? Do they have gastro-intestinal issues if they have gut dystrophy? 2-Only 3 rats per group were used. Please show that this has enough statistical power. Based on our experience, a larger sample size would be required even in animal studies. Also what sex of rats were used? Both? 3-Because of reproducibility issues, in vitro studies should be ideally performed in at least triplicates in 3 independent experiments -the experiments reported seem to have been done in triplicates but only one time. 4-For the gut microbiome, please add what regions of 16S/18S/ITS genes were sequenced and the primers used. 5-For the statistical tests, it is not clear if multiple comparison adjustment was performed or not, so P used for significance should be smaller than the <0.05 described. This should have been done where there were more than 3 groups and for all metabolomics and gut microbiome data. 6-Is Figure 4 data from 16S or metagenomics? The reported 20,000 reads/sample in the rarefaction curve ( Fig 4A) is quite low compared to recent studies where a depth of 50,000-100,000 would be analysed. 7- Figure 4F says the analysis show biomarkers between the two groups. I think this is an overinterpretation of this data as it only shows an association. 8- Figure 6: there is no significance described. 9-With supplementation of Lactobacillus, is there any data on heart function? Ultimately if the authors want to propose it as a new therapy this needs to be performed and conclusions about it being beneficial cannot be drawn without it. Minor comments: 1-Check English grammar through the manuscript. Some sentences are also incomplete. Scientific writing also needs to be observed, the use of wording such as 'pretty lower' is not appropriate. 2-Some abbreviations are cited through the manuscript but only explained at the end in the methods. This is very confusing and difficult to follow. Please explain abbreviations at the first time they are shown in the text. 3-Real-time PCR should read as qPCR and not QPCR (page 23). 4-If paraffin-embedded samples were stained for H&E this is only histology and not immunohistochemistry. 5- Figures 1A, 3A, 5 and 6E: need to quantify each protein relative to b-actin, not only p16, and provide it in a graph with proper statistical analyses. 6- Figure 3: a two-way ANOVA with multiple comparison adjustment should have been performed, not t-test. 7-Firmicutes to Bacteroidetes ratio was not reported in the manuscript but it is referenced to in the discussion.
We appreciate the constructive comments and believe our manuscript has been improved significantly after addressing the comments. Please see below for our point-by-point responses to these comments.

Reviewers' comments:
Reviewer #1 (Remarks to the Author): The manuscript by Xing  What is the connection between the intestine and where Lactobacillus, presumably, is located, and the bone marrow? This is not clearly explained or discussed.
In our study, the amount of lactobacillus was markedly reduced in patients with CCHD, which correlated negatively with D-galactose accumulation in the peripheral blood. As a constituent of milk, lactose is the major nutrient for infants and young children. Lactose is degraded to glucose and galactose in the intestine and absorbed into the body. Most of the absorbed galactose is then degraded by the liver for energy production, while only a small amount is used for human early development 1 .
In the present study, the 16s results for patients with CCHD showed reduced Lactobacillus (one of the most important probiotic bacteria responsible for lactose and D-galactose metabolism in the intestinal tract), and furthermore, atrophied intestinal villus with impaired mucosal barrier fortification and reduced GALK1 expression in liver were observed in the rat hypoxia model. Therefore, we presume that the reduced Lactobacillus, atrophied intestinal villus and impaired liver function work together to contribute to the accumulation of D-galactose. D-galactose would flow through the arteries to penetrate the cortical bone and medullary canal and then into the BM metaphyseal region and arterial capillaries. Based on the specified physiological structures, the metabolites would enter the BM niche and interact with MSCs.
Several studies have shown that the metabolite distribution differs significantly but is correlated between bone marrow and peripheral blood, such as glucose 2 and lactate 3 . Our results also showed that the D-galactose concentration was significantly increased in BM compared with peripheral blood, in both the Cy group and the NCy group.
A series of enzymes that metabolize D-galactose are also present in BMSCs, but there were no significant changes between the NCy and Cy groups (Fig. S4B). Therefore, excessive D-galactose would trigger the production of ROS and ultimately induce premature senescence of MSCs (Fig. 3).
We have revised the discussion to clarify this issue per your suggestion.

Are there other reasons why there might be elevated D-gal concentrations in CCHD
patients? Is D-gal on apoptotic cells? Are there more apoptotic cells in these patients?
To our knowledge, the regulation of the D-gal level is a relatively complicated process that is fine-tuned by both the intestine and liver. In the rat model of cyanosis, a detailed inspection of the process was conducted in our study. Apart from the remarkable decrease in Lactobacillus in the gut, our results suggested that decreased expression of liver GALK1 could contribute jointly to the D-gal elevation in CCHD patients. Nonetheless, gut dysbiosis was likely the main reason for the accumulation of D-gal since supplementation with Lactobacillus dramatically restored the D-gal level in rat models.
Indeed, D-gal has been reported to harbour a cytotoxic effect on cells, either for expediting senescence or inducing apoptosis 4 . It is worth noting that both cytotoxic effects depend largely on ROS through the oxidative metabolism of D-gal as well as advanced glycation end products 5 . It is well known that cells enter a state of senescence when the ROS level is sub-lethal 6 or apoptosis after exposure to high levels of ROS 7 . Per your thoughtful suggestions, we have performed additional experiments to determine the level of apoptosis in the implicated BMSCs. Intriguingly, BMSCs from patients with CCHD exhibited much worse anti-apoptotic capability than those from non-cyanotic patients, indicative of the propensity for apoptosis in CCHD patients (Fig. S1C). Although a complete comprehension of the involved molecular mechanisms between cellular senescence and apoptosis remains to be achieved, it is believed that ROS levels play a pivotal role in balancing the two physiological processes 8 .
Moreover, ageing may predispose related cells to apoptosis through enhanced mitochondrial dysfunction despite some controversies 9,10 . Herein, the observed premature senescence of deficient BMSCs from CCHD patients appeared to be the primary toxic effect of D-gal in our study due to the relatively low level of accumulated D-gal in contrast to that used in other reports. With excessive exposure to continuously increasing ROS, some BMSCs or senescent BMSCs likely undergo a stage of disruption of cellular function and membrane integrity and, in turn, are induced to undergo apoptosis. The underlying mechanism responsible for this process clearly merits further characterization.
We have revised the Discussion to clarify this issue per your suggestion.
Figure S1 (C) BMSC apoptosis was quantified by flow cytometric analysis after staining with Annexin V and propidium iodide (PI). The Annexin V+/PI-cells were undergoing early apoptotic processes, and the Annexin V+/PI+ cells were undergoing late apoptotic processes. **P<0.01.

One disconnect is why BM MSCs are important in heart disease? Are there confounding impacts of CCHD on hematopoiesis, for example?
Thank you for the good suggestions. BMSCs have been widely reported to be a promising candidate cells for cell-based therapy for heart disease [11][12][13] . For children with CCHD, BMSCs are also a widely applied cell resource for the fabrication of tissue engineering patches or conduits 14 . In this regard, it appears to be extremely important to elucidate the functional integrity of BMSCs in CHD patients. The present study showed the deficiency of BMSCs in cyanotic CHD patients compared with non-cyanotic ones and further investigated the underlying mechanism, which might facilitate the development of future therapies for CCHD.
With regard to the impact of CCHD on haematopoiesis, previous studies have shown that hypoxia in CCHD likely stimulates the increase in haemoglobin through increases in EPO, which is generally considered to be a compensatory regulation. Interestingly, Randa et al. 15 recently found that the elevated immature platelet fraction in CCHD patients may be indicative of peripheral platelet destruction. Apparently, haematopoiesis has been perturbed to some extent in CCHD patients. Given that BMSCs and HSCs share the same microenvironment in bone marrow, it is conceivable that the accumulated D-gal in CCHD patients influences the function of HSCs, and even other progenitor cells residing in the bone marrow. We are interested in exploring the dynamic changes in HSCs and likely malfunction of haematopoiesis in CCHD patients.
We have described this issue as a limitation of study in the revised manuscript.

Legends in some figures are almost impossible to read due to the very small font size
( Figure 4, for example).
Your points are well taken. All figures have been re-edited in the revised manuscript.

Minor
1. The entire manuscript contained grammatical mistakes. Please review carefully.
We very much appreciate your suggestion. We have revised the manuscript carefully and corrected the grammatical mistakes.

The histograms are not labeled clearly in Supp. Figure 1.
We have relabelled the histograms in the revised manuscript.

The first part of the results section where patient enrollment is described is confusingparticularly the last statement which states "among them"-it is not clear who is "them".
Thank you for the suggestion. We have replaced "them" with "the 90 patients" in the revised version of the manuscript. Thank you for the thoughtful suggestions. There is no doubt that improved heart function is an important parameter for evaluating the efficacy of heart cell therapy. However, in the present study, we wanted to emphasize how to rescue the premature senescence of BMSCs in children with CCHD. In the rat hypoxia model, compared with the myocardial infarction model, impaired heart function could not be observed during the experimental period (3 weeks) (the data have been supplied in the Supplemental files). Therefore, we would like to perform a new study in the future to investigate the effect of transplantation of rescued or non-rescued BMSCs on heart function with a prolonged experimental period. In the clinical setting, in CCHD patients, the purpose of cell-based regenerative therapy is to utilize cell infusion to prevent the deterioration of heart function or cell seeding to fabricate a tissue engineered conduit. Our findings will also be helpful for the optimization of cell resources for tissue engineering for CCHD patients. factor, we reviewed the milk type used for the enrolled patients. We found that 4 patients in the NCy group were given formula milk, as compared to 5 patients in the Cy group (p = 0.725). In the rat hypoxia model, we also confirmed that hypoxia did not affect the intestinal lactase enzyme activity (Fig. 6D). Taken together, we did not observe correction between patients with CCHD and lactose intolerance. Therefore, we presumed that the presence of lactose intolerance did not affect the conclusion.
Our results revealed intestinal dystrophy in the hypoxic rat model. Cell lysates were analysed for ageing-related proteins by western blotting .(B) Rat liver was subjected to western blotting to analyse the protein levels of GALM, GALK1, GALK2, aldose reductase, GLB1 and β-actin.
All the experimental animals were young male rats. Because all the patients enrolled in the study were very young children, we believe that gender may not be a major factor in the intestinal microbiota disorders observed under hypoxia.

3-Because of reproducibility issues, in vitro studies should be ideally performed in at least triplicates in 3 independent experimentsthe experiments reported seem to have been done in triplicates but only one time.
Thank you for the thoughtful suggestions. In fact, all the experiments were performed in triplicate in 3 independent experiments, but we present only one of them. We have increased the sample size and repeated the animal study to confirm the conclusion.

4-For the gut microbiome, please add what regions of 16S/18S/ITS genes were sequenced and the primers used.
Your point is well taken. We have now provided the sequenced regions and primers in the Materials and Methods.

5-For the statistical tests, it is not clear if multiple comparison adjustment was performed or
not, so P used for significance should be smaller than the <0.05 described. This should have been done where there were more than 3 groups and for all metabolomics and gut microbiome data.
Thank you for the suggestions. All statistical analyses were repeated with multiple comparison adjustment as described in the revised manuscript. Now the adjustment P values were presented for more than 3 groups and metabolomics and gut microbiome data . Figure 4 data from 16S or metagenomics? The reported 20,000 reads/sample in the rarefaction curve (Fig 4A) is quite low compared to recent studies where a depth of 50,000-100,000 would be analysed.

6-Is
Yes, the data from figure 4 were based on the results of 16s and metagenomics analyses.
Since the study was launched in 2013, technical limits restricted it to 30,000 reads/sample. Per your requirement, we have re-sequenced the corresponding samples preserved in liquid nitrogen, and the results are presented in the new figure 4. We have also revised the corresponding methods section.

7-Figure 4F says the analysis show biomarkers between the two groups. I think this is an over-interpretation of this data as it only shows an association.
Thank you for the suggestion. We have changed "biomarkers" to "association" in the revised manuscript. Figure 6: there is no significance described. Per your suggestions, we have performed additional experiments to examine the heart function of the rats (Fig. S5C). We have also listed this issue as a study limitation and revised the conclusion per your suggestion. Minor comments:

8-
1-Check English grammar through the manuscript. Some sentences are also incomplete.
Scientific writing also needs to be observed, the use of wording such as "pretty lower" is not appropriate.
We very much appreciate your suggestion. We have asked a native English speaker to correct the grammatical mistakes and inappropriate language.

2-Some abbreviations are cited through the manuscript but only explained at the end in the methods. This is very confusing and difficult to follow. Please explain abbreviations at the first time they are shown in the text.
Thank you for the suggestions. We have defined the abbreviations the first time they mentioned in the revised manuscript.

3-Real-time PCR should read as qPCR and not QPCR (page 23).
Thank you for the suggestions. This has been corrected in the revised methods.

4-If paraffin-embedded samples were stained for H&E this is only histology and not immunohistochemistry.
We have corrected this in the revised methods.

5-Figures 1A, 3A, 5 and 6E: need to quantify each protein relative to b-actin, not only p16, and provide it in a graph with proper statistical analyses.
Your point is well taken, and the graph is now provided in the revised manuscript.

6-Figure 3: a two-way ANOVA with multiple comparison adjustment should have been performed, not t-test.
Your points are well taken, and we have performed the statistical analysis in the revised manuscript.

7-Firmicutes to Bacteroidetes ratio was not reported in the manuscript but it is referenced to in
the discussion.
We have added the Firmicutes-to-Bacteroidetes ratio in the Results section.

Reviewers' comments:
Reviewer #1 (Remarks to the Author): The revised version is much improved. Some axes are still problematic, however, such as Figure  1D.
The issues raised were addressed through editorial modifications.
Reviewer #2 (Remarks to the Author): The manuscript has been significantly improved. It would have been helpful to see a version of the manuscript with the changes marked. Where possible, most comments were addressed.
In the light of a recent paper published in Nature (https://www.nature.com/articles/nature24628) showing a link between Lactobacillus and blood pressure, did the authors observe any association in their cohort?
I don't work with hypoxia models so I cannot comment on the model itself. The hypothesis behind their manuscript was that CCDH caused hypoxia, which then caused changes in the gut microbiota, resulting in bone marrow senescence, which in turn exacerbated heart disease. Based on their results from echocardiography this did not seem to be the case as there was no change in heart function.
Reviewer#3 (Remarks to the Author): The work of Xing et al describes the association between hypoxia in cyanotic heart disease with perturbed gut microbiota, circulating metabolites and bone marrow mesenchymal stem cell (MSC) senescence. The work is interesting and in general well performed, but the link between MSC function and cardiovascular disease is poorly addressed and in fact not important for the point the manuscript explores, which is the link between hypoxia, gut dysgenesis and MSC senescence. To address this concern, comments follow below: 1. The animal model used in the study is not a CHD model, but a hypoxic model. Therefore, any conclusions about the role the microbiome/MSCs play in CHD are not appropriate and should be removed from the manuscript. In fact, CHD seems to be irrelevant for the described phenotype, even though it is the cause of cyanosis in patients.
On the same topic, the Echo data presented in the paper has no value for interpretation of the results, because 1. This model is not a CHD model, as previously stated; 2. There is no evidence that endogenous MSCs populate the heart of patients or animal models with CHD or any other cardiovascular disease, as a matter of fact. In fact, most murine studies show that cardiac fibroblasts are the major source of cells in diseased hearts. The authors should also be aware of the consensus report on cardiomyocyte regeneration (Circulation. 2017;136:680-686. DOI: 10.1161/CIRCULATIONAHA.117.029343), which addresses the contentious role of cardiac progenitor cells and MSCs in heart disease. Most studies seem to suggest that the role of MSCs is a paracrine one, instead of a regenerative one. The presentation of the Echo data could also be improved in case this dataset remains in the manuscript. The number of animals used in the study should be included, and the graph should show each individual datapoint in the bar, as heart function is subjected to strong biological variation.
We appreciate the constructive comments and believe our manuscript has been improved significantly after addressing the comments. Please see below for our point-by-point responses to these comments.

Reviewer #1 (Remarks to the Author):
The revised version is much improved. Some axes are still problematic, however, such as Figure 1D.
The issues raised were addressed through editorial modifications.
Thank you for the suggestion. All figures have been re-edited in the revised manuscript to meet the requirements of the journal.

Reviewer #2 (Remarks to the Author):
The manuscript has been significantly improved. It would have been helpful to see a version of the manuscript with the changes marked. Where possible, most comments were addressed.
In the light of a recent paper published in Nature (https://www.nature.com/articles/nature24628) showing a link between Lactobacillus and blood pressure, did the authors observe any association in their cohort?
Thank you for the thoughtful suggestion. Since the CCHD patients has relatively lower level of Lactobacillus, we reviewed our cohort and did not find altered blood pressure in CCHD group. Wilck and colleagues' data showed high the salt consumption reduced survival of Lactobacillus, increased Th17 cells and blood pressure. Their study concerned healthy adults with normal blood pressure and a high salt diet for a relatively short time.
As we know, blood pressure is regulated by many factors, such as kidneys, sympathetic nervous system, vasculature, immune system and so on. Our enrolled patients were children. Neither the gut Microbiota 1-3 nor the immune system 4 share the same between children and adult. Therefore, in our cohort, we didn't observe the altered blood pressure in the CCHD group.
We have listed this issue in the study limitation.
I don't work with hypoxia models so I cannot comment on the model itself. The hypothesis behind their manuscript was that CCDH caused hypoxia, which then caused changes in the gut microbiota, resulting in bone marrow senescence, which in turn exacerbated heart disease. Based on their results from echocardiography this did not seem to be the case as there was no change in heart function.
Thank you for the question.
It could be some misunderstandings towards this study because of our misguiding descriptions in certain respects.
The present study explored the down-stream effects of hypoxia driven by CCHD, and the established rat hypoxia model help to elucidate the underlying mechanism. We have reorganized some parts of writings to focus this work on the relationship between chronic hypoxia and BMSCs function, not only limited in the field of cardiovascular disease.
Indeed, in the clinical setting, the heart function of CCHD patient could maintained normal for a long time. Therefore, we did not observe any impaired heart function in either the hypoxic rat or the CCHD patients. The impaired heart function will occur in the end-stage of CCHD if the patients didn't receive the appropriate surgical repair.
Given this, it is proposed that the poor heart function of CCHD might not be imputable to BMSCs deficiency caused by hypoxia. Additionally, a naturally direct link of BMSCs to heart function could not be provided by literatures to date.

Reviewer#3 (Remarks to the Author):
1. The animal model used in the study is not a CHD model, but a hypoxic model. Thank you for your suggestion, we have changed the title in the revised manuscript.

The authors cannot exclude the fact that gut dysgenesis independent of the microbiome
is also a cause for the observed phenotype either in patients or in the animal model, especially considering the observed gut atrophy shown in Figure 6. Therefore, this hypothesis should be also carefully laid out in the discussion section.
Thank you for the suggestion, this part has been added in Discussion section of the revised manuscript. We have highlighted the changes as the follows: 'Lactose is also degraded by lactase, which is secreted by the intestinal gland around the brush border area. We did not observe impaired lactase activity in the intestine in the rat hypoxia model. Likewise, Intestinal mucosa atrophy was noted in our model, which has been previously observed both in hypoxia rat 5 and patients with chronic obstructive pulmonary disease 6 , indicating impaired mucosal barrier fortification and thus the entry of more bacterial metabolites into the circulation 5 . Given this case, we cannot exclude the fact that gut dysgenesis independent of microbiome is also a cause for the observed phenotype. We found that the replenishment of hypoxic rats with Lactobacillus could fully restore the deficient BMSCs, while the intestinal mucosa atrophy did not fully recover even after probiotic therapy. Therefore, the alterations of the gut microbial communities might be the main reason for the deficient BMSCs, and complete surgical repair of cardiac anomalies is warranted to relieve hypoxia and restore the integrity of the mucosal barrier. Thanks for your suggestions. This issue is very important for the clinical translation of probiotic therapy. We have performed the additional experiment as your suggestion.
As shown in the new Fig S7, we found that when the Lactobacillus was removed from the diet of hypoxic rats, the BMSCs showed premature senescence again. It means the continues supplementary would be necessary before the hypoxia status is corrected.