Fasting alters the gut microbiome with sustained blood pressure and body weight reduction in metabolic syndrome patients

Periods of fasting and refeeding may reduce cardiometabolic risk elevated by Western diet. We show that in hypertensive metabolic syndrome (MetS) patients (n=35), a 5-day fast followed by a modified DASH diet (Dietary Approach to Stop Hypertension) reduced systolic blood pressure (SBP), antihypertensive medication need, and body-mass index (BMI) at three months post intervention compared to a modified DASH diet alone (n=36). Fasting altered the gut microbiome, impacting bacterial taxa and gene modules associated with short-chain fatty acid production. Cross-system analyses revealed a positive correlation of circulating mucosa-associated invariant T (MAIT) cells, non-classical monocytes and CD4+ effector T cells with SBP. Furthermore, regulatory T cells (Tregs) positively correlated with BMI and weight. Machine learning could predict sustained SBP-responsiveness within the fasting group from baseline immunome data, identifying CD8+ effector T cells, Th17 cells and Tregs as important contributors to the model. The high-resolution multi-omics data highlights fasting as a promising non-pharmacological intervention in MetS.

patients with MetS using a multi-omics approach. The 'Western diet' is known to induce 1 metabolic inflammation, accelerating CMD 4 . The gut microbiota is a delicate ecosystem that 2 plays a pivotal role in health and disease. Dysbiosis has been observed as a characteristic of 3 several inflammatory, cardiovascular, and metabolic disorders (e.g. obesity) 5 , including 4 hypertension 6, 7 . The 'healthy' gut microbiome is relatively stable, although various factors such 5 as antibiotics, intestinal infections, profound dietary or lifestyle changes, such as moving on or 6 off a 'Western diet' can induce transient or persistent changes to this ecosystem. Traditionally, 7 fasting plays an important role in different cultural and religious practices. Dramatic caloric 8 restriction not only affects host health and physiology, but likely also has an impact on the 9 microbiome. However, the effect of fasting on the gut microbiome and the consequences of 10 refeeding after such an intervention have not yet been described. 11 12

Study design and patients 14
Subjects diagnosed with MetS and elevated systolic blood pressure (BP) were randomly 15 allocated into two intervention groups. The fasting arm underwent five days of fasting 16 according to 'Buchinger´s protocol' 8 followed by three months of a modified DASH diet (also 17 referred to as the refeeding period) ( Figure 1A, Table 1

1
Fasting affects the gut microbiome and immunome 2 As we have previously reported a major influence of common MetS drugs on the microbiota 9 , 3 we accounted for any changes in medication regime or dosage in our statistical tests, alongside 4 controlling for important demographic features such as age and sex. Neither fasting nor 5 refeeding significantly (Mann-Whitney U (MWU) P>0.05) altered microbiome species 6 richness/alpha diversity (Shannon, Figure 1B, Supplementary Figure S1), or intersample gut 7 taxonomic variability/beta diversity (intersample Bray-Curtis distance, Figure 1C; similar 8 results achieved on 16S data (results not shown)). However, there were substantial and 9 significant (PERMANOVA P=0.001) differences in microbial composition within individuals 10 during fasting, reflecting a characteristic intervention-induced shift, which later partially 11 reverted upon a three-month refeeding on a DASH diet ( Figure 1D, Supplementary Table S1 12 and Figure S2A). This was echoed by analogous significant (PERMANOVA P=0.001) changes 13 in host immune cell composition during the intervention, revealing a fasting-specific signature, 14 which likewise largely reversed during refeeding ( Figure 1E, Supplementary, Table S1). Fasting resulted in a reduction of CD3 + , CD4 + T cells and CD19 + B cells, while the frequency 18 of CD8 + T cells was unaltered. In contrast, fasting increased the abundance of monocytes 19 (CD14 + CD11c + CD19 -CD3 -) and TCR + T cells. However, these changes were reversed upon 20 refeeding ( Figure 1H, Supplementary Table S2). Of note, frequency of CD123 + CD14 -CD16 -21 HLA-DR + plasmacytoid dendritic cells also increased upon fasting and was still enriched after 22 refeeding ( Figure 1H, Supplementary Table S2). When looking closer into monocyte subsets, 23 fasting increased (and refeeding reduced) the frequency of classical CD14 high CD16 -, non-24 classical CD14 low CD16 ++ , and intermediate CD14 high CD16 + monocytes ( Figure 1I, 25 Table S2), which was confirmed by unbiased FlowSOM analyses 26 (Supplementary Figure S4A-D). Fasting also affected the relative abundance of differentially 1 activated T cells. Upon fasting, CD8 + T cells showed a higher percentage of terminally 2 differentiated cells (Teff, CD45RO -CD62L -) and a lower percentage of the naïve phenotype (Tn,  3 CD45RO -CD62L + ), while memory T cells were not affected ( Figure 1I, Supplementary Figure  4 S3, Table S1). A similar pattern was observed in CD4 + Teff ( Figure 1I, Supplementary Table  5 S1). Further, fasting decreased the frequency of pro-inflammatory Th17 (CD27 bright 6 CD161 + CCR6 + CXCR3 -CD25 -CD4 + ), as well as TNF-and IFN-producing Th1 cells (Figure  7 1I, Supplementary Table S1). These changes were partially reverted upon refeeding ( Figure 1I). 8 Neither fasting nor refeeding changed the overall frequency of CD161 + V7.2 + CD3 + mucosa-9 associated invariant cells (MAIT, Figure 1H, Supplementary Figure S3). However, frequency 10 of pro-inflammatory MAITs producing TNF and IFN significantly decreased upon fasting 11 and were minimally affected by refeeding ( Figure 1I, Supplementary Table S2). 12 Next, we tested all gut microbial taxa and gene functional (KEGG 10 , GMM 11 ) modules for 13 abundance shifts during fasting or refeeding, as well as persistent shifts across the three-month 14 study period, controlling for age, sex and any changes in medication ( Figure 1F-H, 15 Supplementary Table S1). Fasting stimulated shifts in abundance of several core commensals, 16 which was reversed upon refeeding ( Figure 1F, Supplementary Table S1). Many Clostridial 17 Firmicutes shifted significantly in abundance, with an initial decrease in butyrate producers 18 such as F. prausnitzii, E. rectale and C. comes, which had also reverted after three months. 19 Interestingly, C. comes abundance change was predominantly driven by changes in BMI. 20 Bacteroidaceae showed the opposite pattern and an OTU identified as an Odoribacter species 21 likewise bloomed during fasting. At the end of the refeeding period, a persistent depletion could 22 be seen in Enterobacteriaceae, especially Escherichia coli. Accompanying the shift in 23 composition are vast changes in microbial metabolic capacity ( Figure 1G, Supplementary Table 24 S1). Fasting enriches for propionate production capacity, mucin degradation gene modules and 25 for diverse nutrient utilization pathways. 26 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
(which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint Reanalyzing previously published data, we compared the microbiome signatures of metformin 1 use and MetS to those seen in our novel dataset 9, 12 . For ease of comparability, we proceeded 2 with only human gut specific functional modules (GMM) assessed from shotgun sequencing 3 data available for the fasting arm. Certain fasting-or refeeding-associated functional gene 4 modules from our data were found to overlap with signatures of metformin usage or MetS, but 5 there was little concordance on a taxonomic composition level, in line with previously described 6 higher functional than taxonomic concordance between microbiomes. Of note, when comparing 7 the metformin signal to the MetS signal, it is clear that these two effects are functionally distinct 8 and often oppose one another. Where there were overlapping functional features with results 9 from our novel cohort, the fasting effect appeared to be metformin-like, and during recovery 10 metabolic changes were again reverted (Supplementary Figure S5). That is, certain gut 11 metabolic capacities are promoted both by fasting and by metformin treatment. 12 13 Fasting reduces long-term systolic blood pressure and body weight in MetS patients 14 Assessing the clinical relevance of our intervention, we inspected clinical outcomes in the two 15 study arms. While DASH reduced office SBP after three months ( Figure 2H), it did not 16 significantly (MWU P=0.27) affect 24h ambulatory SBP, the gold standard of clinical BP 17 measurements ( Figure 2A) 3 . In contrast, fasting followed by a modified DASH diet led to a 18 sustained reduction both in 24h ambulatory SBP and mean arterial pressure (MAP) (MWU 19 P<0.05, Figure 2A). Further, subjects undergoing fasting could significantly ( 2 P=0.035) 20 reduce their intake of antihypertensive medication in 43% of cases, compared to only 17% of 21 the cases on DASH alone, while their BP remained under control ( Figure 2B, Supplementary 22 Table S2). Because the BP response to fasting was heterogenous in our cohort (Figure 2A-B), 23 we applied a decision tree model to stratify patients based on their ambulatory BP response, 24 adjusted for antihypertensive medication (Supplementary Figure S6, Table S3). The responder 25 group had a median SBP decrease of 8.0 mmHg including a high reduction of medication, while 26 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
(which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint the decrease in the non-responder group was significantly lower (0.3 mmHg; Figure 2C). In the 1 DASH only arm, 17 patients were classified as responders with a median SBP decrease of 8.0 2 mmHg, while the non-responders (n=14) showed no decrease in median SBP (0.5 mmHg, 3 Figure 2C). Fasting followed by a modified DASH diet, unlike a modified DASH diet alone, 4 significantly (drug-adjusted post hoc P<0.05) reduced BMI and body weight even three months 5 post-fasting ( Figure 2D-E). Although all participants showed a reduction in body weight, this 6 reduction alone could not explain the long-term ambulatory SBP and MAP changes exclusive 7 to the fasting arm ( Figure 2F-G), nor the microbiome or immunome changes accompanying it 8 (95% of significant findings retain significance when BMI is added as a predictor to the nested 9 models for longitudinal data, see Supplementary Table S4). Only a minority of significant 10 effects (and only two microbiome effects) observed in the fasting-DASH arm could be 11 replicated in the equally powered DASH-only arm ( Figure 2I). 12 13 BP responder-specific changes in the gut microbiome and immunome 14 Because the BP responsiveness was heterogenous in the fasting-DASH arm (Figure 2A-C), we 15 hypothesized that unique characteristics involving the immunome or microbiome of these 16 patients may contribute to their BP response. We compared the impact of fasting and refeeding 17 in the complete fasting arm, in the BP-responders of the fasting arm, and in the DASH-only 18 arm ( Figure 3A-B, Supplementary Table S1, 5, 6). Even at reduced statistical power, we were 19 able to capture changes in abundance of many gut microbial taxa that were uniquely 20 characteristic of successful fasting treatment, most of which displayed the fasting-refeeding 21 reversal dynamic described above ( Figure 3A, Supplementary Table S5, 6). Responders were 22 especially characterized by an initial and sustained enrichment of an unclassified Odoribacter 23 species with concomitant Actinomyces depletion. Responders experienced an initial depletion, 24 and later a strong enrichment, of the butyrate producer F. prausnitzii. Accordingly, in BP-25 responders, fasting resulted in an initial bloom and later depletion of Bacteroides ( Figure 3A). 26 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
(which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint Virtually no overlap with effects seen in the equally powered DASH arm were found, indicating 1 that fasting is needed on top of a BP-reducing diet for these changes to occur (Supplementary 2 Figure S7). 3 In profiling the microbial metabolic potential in BP-responders, we focused on gene modules 4 curated for relevance to metabolism in the human gut (GMM) 11 . On a functional level, 5 responder-characteristic changes resemble those in the fasting arm at large, but with even more 6 pronounced relative enrichment for propionate production (MF0126, MF0121) modules 7 ( Figure 3B). Some modules were significantly altered in abundance only in this stratified 8 subgroup, indicating these changes strongly characterize responders compared to non-9 responders (Supplementary Table S5). For example, pyruvate:formate lyase (MF0085) is 10 depleted during recovery only in responders. 11 Changes to the immunome of responders are similar to those seen in the unstratified fasting 12 group and differ from those in the DASH arm ( Figure 3C, Supplementary Table S1, 5, 6). In 13 the fasting arm, several immune features related to pathogen-sensing and mucosal immunity 14 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not peer-reviewed)
The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint of the same subjects at all three time points (see Methods, Supplementary Table S7). Figure 4A  1 shows a chord diagram constructed from these data, where the colored outer rings are lined with 2 components from one of our three tested system spaces during fasting, refeeding and over the 3 full duration of the study, and where the color of the connectors between factors indicate a 4 positive or negative association (Spearman´s rho). We identified a cluster of circulating 5 cytokine-producing MAIT cells (absolute number and fraction of CD3 + T cells), which 6 positively correlated with 24h ambulatory SBP ( Figure Table S9). 18 The association between MAIT cells and BMI is still a matter of debate 13 . We found in our 19 study that the abundance of MAIT cells did not correlate with either BMI, weight, waist 20 circumference, waist-hip ratio or body fat percentage (Supplementary Figure S8, Table S7). 21 However, we found that these weight-related parameters did correlate with the abundance of a 22 subset of circulating Treg-like cells (CD62L -CD45RO -CD25 + CD4 + ), a cell type previously 23 linked to morbid obesity in human subjects 14 . 24 A recent publication showed non-classical monocyte enrichment in hypertensive patients 15 . 25 Interestingly in our study, circulating non-classical monocytes were enriched upon fasting and 26 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not peer-reviewed)
The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint during refeeding may have stabilized a less hypertensive state through downstream 1 mechanisms. In addition, the difference in abundance of Parabacteroides sp. in responders and 2 non-responders had equalized by time of follow-up. 3 As mentioned above, immunome composition also differed already at baseline in responders 4 and non-responders. To further elucidate this phenomenon, we applied machine learning 5 algorithms and empirically show that we can make effective predictions from the immunome 6 data. From 494 total immune variables, step-wise forward regression identified the top ten 7 discriminators of responders from non-responders at baseline. A prediction accuracy of 71% 8 (sensitivity 75%, specificity 70%, and F1 score 77%) was achieved using a leave-subject-out 9 cross-validation for whether or not a future patient would respond favorably to fasting with 10 regards to BP ( Figure 5B). Within this multivariate analysis, the driving immune features of 11 this classifier highlighted a lower CXCR3 + CD25 -CD4 + /CD25 high CD4 + (most likely Th1/Treg 12 ratio), alongside lower abundances of CD24 + memory CD8 + T cells and IL-17 + TNF + MAIT 13 cells in responders relative to non-responders ( Figure 5C, Supplementary Figure S9E). 14 Regarding the top ten features derived as indicative for successful patient classification, 15 responders seem to have less of a pro-inflammatory immune signature at baseline ( Figure 5C). 16 Notably, we could increase the prediction performance of the classifier up to 78% by using 17 changes of immune cell abundances between baseline and three-month follow-up visit as a basis 18 for prediction of BP response at the single-patient level (Supplementary Figure S9D, F). In 19 contrast, for subjects on a DASH diet only, corresponding classifiers were unable to predict BP 20 response above chance level. 21 22

23
In summary, we have carried out the first high-resolution multi-omics characterization of 24 fasting in patients with metabolic syndrome. The major novel finding is that periodic fasting 25 followed by 3 months of modified DASH diet induces concerted and distinct microbiome and 26 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not peer-reviewed)
The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint immunome changes that are specific to fasting itself leading to a sustained BP benefit (Fig. 2I), 1 since it cannot be reduced by DASH diet alone. Fasting followed by modified DASH led to a 2 significant long-term reduction in body weight. However, neither the change in blood pressure 3 nor global changes to the microbial composition or immunome appear mediated by this BMI 4 decrease (95% of findings retain significance when deconfounding for BMI change, see Table  5 S4, and body weight reduction is not more pervasive in treatment responders than 6 nonresponders, Figure 2G). This suggests that further sustained weight loss beyond what is 7 observed here may synergize with the fasting benefits to reduce hypertension even further. 8 Fasting induced a profound change in circulating immune populations; e.g. depleted Th1 cells 9 and permanently enriched dendritic cells, which both have been shown previously to play a role 10 in the pathogenesis of experimental hypertension 16, 17 . Further, we discovered significant 11 correlations between circulating MAIT cells and 24h ambulatory BP and MAP. 12 A growing body of evidence suggests that the abundance of certain microbes is associated with 13 cardiovascular health. Previous reports on hypertensive patients show taxonomic and functional 14 gut microbiome shifts 6, 7 . For example, Firmicutes have been shown to be more abundant in 15 healthy controls compared to pre-hypertensive and hypertensive patients 7 . Upon fasting, several 16 Clostridial Firmicutes shifted significantly in abundance, with an initial decrease in butyrate 17 producers such as F. prausnitzii, E. rectale and C. comes, which were also reverted after three 18 months upon refeeding; with the latter taxon one of few likely to be indirect effects of the 19 observed weight reduction (Table S4). Bacteroidaceae showed the opposite pattern, and an 20 unclassified Odoribacter species likewise bloomed during fasting. Of note, Odoribacter sp. has 21 been negatively associated with both SBP 18 and vascular stiffness 19 in obese women. 22 Escherichia coli was shown to be associated with endothelial dysfunction and MetS. In our 23 study, fasting permanently depleted Escherichia coli, suggesting that fasting could promote 24 cardiovascular health through the gut microbiome. Further, functional microbial metabolism in 25 fasting patients at baseline share some similarities to the previously profiled hypertensive 26 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not peer-reviewed)
The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint microbiome 7 . In the fasting arm, the functional shift during refeeding enriches for functional 1 modules also enriched in non-hypertensive controls, i.e. for potentially BP-protective factors. 2 Clinical studies represent a highly heterogeneous situation with multifactorial disease features 3 and strongly variable microbial and lived environments. To account for this heterogeneity, we 4 compared the data from our longitudinal study (post-fasting and 3-month) to the respective 5 baseline values of the study subjects. This intraindividual analysis allowed us to identify BP 6 responder-specific changes despite even reduced power in such a sub-stratified analysis. 7 Notably, despite the non-homogenous starting conditions in our MetS patients, our five-day 8 fasting followed by a DASH intervention improved BMI in all participants and ambulatory SBP 9 in two thirds, even three months post-intervention. The responder-specific microbiome changes 10 in our fasting arm post-intervention (enrichment of an Odoribacter species, Bacteroides and 11 Firmicutes, depletion of Actinomyces and F. prausnitzii) are likely beneficial to the host. A 12 recent study profiling the hypertensive microbiome showed that during disease, patients 13 experienced an enrichment of Actinomyces, and a depletion of F. prausnitzii, Bacteroides and 14 Firmicutes 7 . In addition, some functional gut specific gene modules in our dataset were enriched 15 only in BP-responders, for example the pyruvate:formate lyase module, MF0085. Our data are 16 congruent with a study showing enrichment of the same enzyme in symptomatic atherosclerosis 17 patients compared to healthy controls 20 . 18 The fact must also be addressed as to how, despite the fact that many of the immunome and 19 microbiome shifts post-fasting are transient, there exists a sustained improvement of 20 cardiovascular health in our patients. We hypothesize that during a specific time window during 21 early refeeding, certain fasting-depleted taxa (including core butyrate producers) and associated 22 immune cells may be enriched beyond both baseline or follow-up at 3 month, leading to a 23 temporary state of high butyrate availability, especially on a DASH diet. Therefore, we 24 speculate that anti-inflammatory effects of this SCFA during regrowth, as well as similar effects 25 of propionate during fasting itself, may be important mechanisms behind the sustained 26 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not peer-reviewed)
The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint improvement in blood pressure that we observe. While we cannot directly test it in the present 1 cohort, it is a scenario consistent both with expectations from the literature and with our 2 observations of a consistent depletion-then-regrowth pattern. Thus, future work will involve 3 studying a fasting/refeeding cohort also at intermediate time intervals. It is likewise a limitation 4 that we cannot infer direction or causality of concomitant immunome-microbiome changes 5 from the present cohort study alone. Accordingly, our next steps will also involve direct 6 manipulation of microbiota, as well as direct infusion of their potentially immunomodulating 7 metabolites, in animal models for hypertension, similar to some of our recent proof-of-principle 8 studies 21, 22 . 9 Stratification of the cohort to BP-responsiveness showed that also immune changes present in 10 the fasting arm are more pronounced in responders than in non-responders, and are 11 fundamentally different from the changes observed in the DASH-only arm. The DASH-only 12 arm was associated with the decrease of CD8 + Tem cells, previously reported to play a role in 13 hypertension 17, 23 . Responders and non-responders not only reacted differentially to fasting, but 14 also differed at baseline, including by depletion of Bacteroides sp., Parabacteroides sp., and 15 propionate synthesis genes pre-intervention. These features were then normalized during 16 fasting. Notably, recent experimental work suggested an anti-hypertensive effect of propionate 17 treatment in mice. 22 Furthermore, Parabacteroides are more commonly found in the 18 microbiome of hypertensive subjects than in healthy controls 6 . Our findings indicate responders 19 and non-responders to our intervention differ with regards to several gut microbiome features 20 relevant to hypertension, with fasting-induced normalization of these differences seen during a 21 successful fasting intervention. 22 Through network analysis of the immunome, microbiome and clinical data, we identified 23 significant correlations between circulating MAIT cells and 24h ambulatory BP and MAP. 24 MAIT cells represent up to 10% of peripheral blood T cells, but in contrast to other classical T 25 cells 17 , have not yet been linked to the regulation of BP. They differ in many aspects from 26 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not peer-reviewed)
The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint conventional T cells by expressing a semi-invariant TCR -chain V7.2-J33. MAITs can 1 produce various cytokines mimicking an effector/memory-like phenotype and yet they behave 2 rather like innate cells. During aging 18 and CMD 13, 24 , absolute circulating MAIT number and 3 frequencies decrease, while certain subsets of cytokine producing and adipose tissue MAITs 4 were found to be enriched in obese type 2 diabetic patients 13 . In contrast to a previous report 5 showing significant enrichment of IL-17-producing MAIT cells (after re-stimulation) in 6 severely obese patients (average BMI=54 13 ), our cohort (average BMI=34) had very low IL-17 7 producing MAITs, and they were only quantifiable in a small group of patients (data not 8 shown). 9 Using machine learning, we were able to utilize deep immunophenotyping data to predict at 10 baseline which subjects were likely to decrease their BP during fasting despite the small number 11 of subjects. In addition, accuracy of the prediction was enhanced further taking the dynamics 12 of immune populations along the course of the study into account. No corresponding prediction 13 of favorable response to a DASH-only intervention was possible. Thus, we demonstrate the 14 practical utility of a machine learning analysis pipeline for predicting BP benefit of fasting in 15 MetS patients with hypertension. Personalized nutrition, informed by -omics measurements, is 16 therefore a possible translational benefit of our study. 17 18

19
It is important to recognize that our study represents patients with hypertension and MetS solely 20 from a Caucasian-European background. This selection criterion introduces a selection bias in 21 our study design. Additional research is necessary to elucidate whether the results presented 22 here could be generalizable to a more heterogenous patient population. Further, our recruitment 23 procedure could already have introduced a selection bias toward patients who were interested 24 in fasting/dietary studies and therefore are sensitive about their cardiometabolic health. Since 25 the participants were especially interested in the fasting procedure, the allocated DASH 26 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not peer-reviewed)
The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint participants were offered after successful completion of the study a cost-free fasting cycle. 1 However, we cannot exclude that this led to an increased long-term motivation compared to the 2 participants, who started with the fasting protocol. In addition, the relatively low patient number 3 could be regarded as a limitation. Although our present study is large enough to allow inference 4 of significance for most strong contributors to the observed effect, results are likely not 5 complete, and follow-up in additional and larger studies will be needed for a comprehensive 6 view of subtle fasting-associated host and microbiome features. Our study design did not allow 7 to blind participants regarding their intervention. To maximally reduce the bias, the scientific 8 staff were blinded during the process of processing and analyses of collected samples and 9 measurements. Further, the present study cannot infer how frequently fasting cycles should be 10 repeated to control BP in at-risk patients, nor whether it is as effective without a concomitant 11 DASH intervention. Future clinical studies are needed to test these hypotheses. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not peer-reviewed)
The copyright holder for this preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not peer-reviewed)
The copyright holder for this preprint  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not peer-reviewed)
The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not peer-reviewed)
The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
(which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
(which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
(which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.02.23.20027029 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.