Thirty-seven AD patients and 21 healthy subjects were enrolled in this study. Alcohol dependence was diagnosed based on the criteria of the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV) (American Psychiatric Association, 1994), when subjects were admitted to the Guizhou 2^nd Provincial People’s Hospital from July 2019 to September 2019. At admission, the subjects’ medical history, demographic data, and relevant clinical data were collected.

The inclusion criteria for the AD group were the following: (1) diagnosis of AD according to the DSM-IV and (2) Han Chinese males aged 18-60 years. The exclusion criteria for the AD group were the following: (1) metabolic disorders, such as diabetes and obesity, (2) severe physical diseases, such as cardiovascular disease and infectious disease, (3) DSM-IV Axis I disorders other than alcohol and nicotine dependence, (4) night shift workers, (5) use of antibiotics, steroids, or other microbiota-modulating medications within 2 months before enrollment, and (6) abstinence longer than 1 week. The inclusion criterion for the healthy control (HC) group was Han male aged 18-60 years. The exclusion criteria for the HC group were identical to the AD group except DSM-IV Axis I disorders other than alcohol and nicotine dependence, where AD was excluded.

All enrolled subjects were in the hospital the entire time before the completion of four fecal sample collections and had the same meals provided by the hospital. The details of medications that were prescribed during fecal sampling are provided in Supplementary Table S1 .

The Pennsylvania Alcohol Craving Scale (PACS) (Flannery et al., 1999), Revised Clinic Institute Alcohol Withdrawal Syndrome Assessment Scale (CIWA-Ar) (Sullivan et al., 1989), Fagerstrom Test of Nicotine Dependence (Heatherton et al., 1991), and self-reported alcohol-related questionnaires were used to measure the level of alcohol craving, severity of withdrawal, AD duration, and daily alcohol consumption.

Based on the disease duration, daily consumption, withdrawal severity according to CIWA-Ar, and craving levels according to PACS, AD patients were subdivided into a series of subgroups. With regard to daily alcohol consumption, AD patients were divided into the high standard drinks (HSD) group (≥ 15 standard drinks; 16 subjects) and the low standard drinks (LSD) group (< 15 standard drinks; 16 subjects) (Whelan, 2002). With regard to the duration of AD, AD patients were divided into the long courses (LC) group (AD duration ≥ 10 years; 18 subjects) and the short courses (SC) group (AD duration < 10 years; 14 subjects) (Schuckit and Smith, 2011). With regard to the severity of withdrawal symptoms, AD patients were divided into the high CIWA-Ar (HCW) group (CIWA-Ar ≥ 16; 18 subjects) and the low CIWA-Ar (LCW) group (CIWA-Ar < 16; 14 subjects) (Zhuo et al., 2009). With regard to craving levels, based on the distribution of PACS scores in our study, AD patients were divided into the high PACS (HPA) group (PACS ≥ 6; 16 subjects) and the low PACS (LPA) group (PACS < 6; 16 subjects).

Four fecal samples were collected from each subject at 7:00 AM, 11:00 AM, 3:00 PM, and 7:00 PM within 4 days of enrollment. The four fecal samples collected from 32 AD patients and 20 HC subjects were used for the analysis, the fecal samples at a different time point rather than 7:00 AM, 11:00 AM, 3:00 PM, and 7:00 PM were not collected for the analysis and the details are presented in Supplementary Figure S1 and Table S2 .

According to the 12th edition of the Manual of Procedures for the Human Microbiome Project (Mcinnes and Cutting, 2010), fecal samples were collected in a sterile tube, frozen in a liquid nitrogen flash-freezer, and stored at -80°C within 30 min of collection. For 16S rDNA sequencing, polymerase chain reaction (PCR) amplification was performed, spanning the V4 region of the 16S rRNA gene. The PCR products were sequenced using 2 × 250 bp paired-end sequencing (Illumina HiSeq). Further details of the PCR methods are provided in the Supplementary Information.

The statistical analyses of sociodemographic and clinical data were performed using SPSS 23.0 software (IBM Corp). Education years and BMI from all the included subjects met normality of the data and equal variance, determined by the Kolmogorov–Smirnov and Levene tests, respectively, and independent t-tests were used to compare the AD and HC groups; FTND scores and ages from all the subjects did not meet normality of the data and equal variance, and the Kruskal-Wallis test was performed. For marital status and occupation, the χ² test was performed. To test differences in the gut microbiota between the AD and HC groups, the Wilcoxon test was performed for the α-diversity, β-diversity, relative abundances of bacteria and KEGG pathways. To determine diurnal oscillations, because of the limited number of time points, Jonckheere-Terpstra-Kendall analysis and cosine analysis were not feasible. Therefore, to determine the diurnal oscillation, based on the fecal samples at the four different time points, Kruskal-Wallis test and Dunn’s multiple-comparison test by Prism 8.00 software (GraphPad, La Jolla, CA, USA) and R 3.1.1. software were used for the OTUs, β-diversity, relative abundances of bacteria at different taxonomic levels and KEGG pathways, and Principal coordinate analysis (PCoA) was performed by QIIME (version 1.8.0) (Caporaso et al., 2010) based on weighted UniFrac distance. Microbiota-related data are expressed as the mean ± SEM.

The study protocol was approved by the Ethics Committee of Peking University Sixth Hospital on April 17, 2019. All enrolled subjects signed an informed consent form.

We statistically analyzed clinical baselines between the AD and HC groups. On the aspect of age, occupation, education years, marital status, BMI and FTND scores, no significant differences were found between AD patients and healthy subjects. However, the sleep quality of the AD group was statistically worse than the HC group ( Table 1 ). The duration of AD, daily consumption of alcohol, craving levels and withdrawal severity were also elaborated in Table 1 .

This trial collected 52 × 4 fecal samples at the four different time points following enrollment ( Figure 1A ). The number of clean reads per sample is 71090*2~75123*2. To test for alterations of the gut microbiota in AD patients reported previously in human studies (Mutlu et al., 2012; Leclercq et al., 2014; Dubinkina et al., 2017), fecal samples taken at 7:00 AM (i.e., when most subjects normally defecated spontaneously) were analyzed. α-Diversity of the gut microbiota was significantly different between the AD and HC groups (p < 0.0001, Wilcoxon test; Figure 1B ). Based on weighted UniFrac distance, the gut microbiota in the AD group differed from the HC group (p < 0.001, Wilcoxon test; Figure 1C ). Further analyses showed that at the bacterial phylum level, the AD group exhibited an increase in the relative abundance of Fusobacteria and Tenericutes (all p < 0.05, Wilcoxon test) and a decrease in the relative abundance of Bacteroidetes, Lentisphaerae, Synergistetes, and unclassified bacteria compared with the HC group (all p < 0.05, Wilcoxon test; Figure 1D , Supplementary Figure S2A-C ). At the class level, Bacteroidia, Lentisphaeria, and Synergistia were less abundant (all p < 0.05, Wilcoxon test), whereas Fusobacteriia, Actinobacteria, and Mollicutes were more abundant in the AD group (all p < 0.05, Wilcoxon test; Supplementary Figure S2D ). At the order level, Fusobacteriales, Rickettsiales, and Actinomycetales were enriched (all p < 0.05, Wilcoxon test), whereas Bacteroidales, Synergistales, and Victivallales were depleted in the AD group (all p < 0.05, Wilcoxon test; Supplementary Figure S2E ). At the family level, the AD group had higher levels of Fusobacteriaceae, Veillonellaceae, Actinomycetaceae, Eubacteriaceae, Porphyromonadaceae, and Pseudomonadaceae (all p < 0.05, Wilcoxon test) and lower levels of Ruminococcaceae, Lachnospiraceae, Paraprevotellaceae, Peptococcaceae, Barnesiellaceae, Mogibacteriaceae, Christensenellaceae, Dehalobacteriaceae, Bacteroidaceae, Clostridiaceae, Dethiosulfovibrionaceae, and Victivallaceae (all p < 0.05, Wilcoxon test; Supplementary Figure S2F ). The AD group exhibited increases in the relative abundance of 15 genera (all p < 0.05, Wilcoxon test; Figure 1E , Supplementary Table S3 ) and 16 species (all p < 0.05, Wilcoxon test; Supplementary Figure S2G , Table S4 ). In contrast, the AD group exhibited decreases in the relative abundance of 15 genera (all p < 0.05, Wilcoxon test; Figure 1E , Supplementary Table S3 ) and 14 species (all p < 0.05, Wilcoxon test; Supplementary Figure S2G , Table S4 ).

Among the detected KEGG pathways in the AD group, the relative abundance of KEGG pathways involved in xenobiotics and carbohydrate metabolism was higher ( Figure 1F ), whereas the relative abundance of KEGG pathways involved in replication and repair, transcription, translation, nucleotide metabolism, cell motility, cell growth and death, environmental adaptation, and amino acid metabolism were lower ( Figure 1G ) compared with the HC group ( Supplementary Figure S2H ).

To characterize diurnal oscillation of the gut microbiota, taxonomic analysis of the fecal microbiota at the four different time points was performed. In the AD group, based on weighted UniFrac distance, β-diversity among the four time points exhibited robust oscillation (p = 0.01, Kruskal-Wallis test; Figure 2A ). In the HC group, the oscillation of β-diversity was completely absent ( Figure 2B ). Approximately 1.68% of OTUs in the HC group and 0.66% of OTUs in the AD group exhibited robust oscillation ( Figure 2C ). Additionally, the abundance of bacteria showed different oscillation patterns between the AD and HC groups.

In the AD group at the phylum level, Cyanobacteria exhibited diurnal fluctuations among the four different time points (p < 0.05, Kruskal-Wallis test; Figure 2D , Supplementary Figure S3A ). At the class and order levels, none exhibited diurnal fluctuations. At the family, genus, and species levels, Pseudomonadaceae, Pseudomonas, Clostridium sordellii, and Prevotella melaninogenica exhibited diurnal fluctuations (all p < 0.05, Kruskal-Wallis test; Figure 2E, F , Supplementary Figure S3B ).

In the HC group at the phylum level, no significant diurnal fluctuations were found among the four different time points ( Figure 2G ). However, Firmicutes showed a trend toward a significant fluctuation (p = 0.06, Kruskal-Wallis test). At the class level, Flavobacteriia exhibited diurnal fluctuations (p < 0.01, Kruskal-Wallis test; Supplementary Figure S3C ). At the order level, Rickettsiales, Flavobacteriales, and Neisseriales exhibited diurnal fluctuations (all p < 0.05, Kruskal-Wallis test; Supplementary Figure S3D ). At the family, genus, and species levels, Flavobacteriaceae, Neisseriaceae, Streptococcaceae, Porphyromonas, Capnocytophaga, Moryella, Tannerella, Atopobium, Prevotella melaninogenica, Streptococcus infantis, Porphyromonas endodontalis, Prevotella pallens, and Prevotella nanceiensis exhibited diurnal fluctuations (all p < 0.05, Kruskal-Wallis test; Figure 2H , Supplementary Figure S3E, F ).

Among the detected KEGG pathways in the AD group, the relative abundance of “calcium signaling pathway” and “photosynthesis-antenna protein” oscillated among the four time points ( Figure 2I , Supplementary Figure S3G ), but none oscillated in the HC group ( Supplementary Figure S3H ).

In the HSD group, β-diversity of the gut microbiota among the four different time points exhibited robust oscillation (p < 0.05, Kruskal-Wallis test; Figure 3A ), and no OTUs exhibited diurnal fluctuation. In the LSD group, no oscillation of β-diversity was observed ( Figure 3B ), and approximately 0.63% of OTUs displayed diurnal fluctuation ( Figure 3C ). At all taxonomic levels, no bacteria oscillated diurnally in either group ( Figure 3D, E ). Furthermore, among the detected KEGG pathways, the relative abundance of “general function prediction only” exhibited diurnal oscillation in the HSD group, and the relative abundance of “calcium signaling pathway” exhibited diurnal oscillation in the LSD group ( Figure 3F ).

In the LC group, β-diversity of the gut microbiota exhibited no diurnal oscillation ( Figure 4A ), and β-diversity displayed diurnal oscillation in the SC group (p < 0.05, Kruskal-Wallis test; Figure 4B ). Approximately 0.50% of OTUs displayed diurnal oscillation in the LC group and approximately 0.26% of OTUs exhibited diurnal oscillation in the SC group ( Figure 4C ). In both groups at the phylum, class, order, family, and species levels, the gut microbiota exhibited no statistically significant diurnal oscillation ( Figure 4D, E ). At the genus level, SMB53 exhibited diurnal oscillation in the LC group ( Figure 4F ). Among the detected KEGG pathways, none displayed diurnal oscillation in either group.

In both the HCW and LCW groups, β-diversity of the gut microbiota displayed diurnal oscillation (both p < 0.001, Kruskal-Wallis test; Figure 5A, B ). Approximately 0.75% of OTUs in the HCW group, and 0.39% of OTUs in the LCW group displayed diurnal oscillation ( Figure 5C ). At the phylum, class, order, family, and genus levels, the gut microbiota did not exhibit diurnal oscillation in either group ( Figure 5D, E ). At the species level, Clostridium sordellii exhibited diurnal oscillation in the HCW group, and Actinomyces hyovaginalis exhibited diurnal oscillation in the LCW group ( Figure 5F ). Among the detected KEGG pathways, none displayed diurnal oscillation in either group.

In the HPA group, β-diversity of the gut microbiota displayed diurnal oscillation (p < 0.05, Kruskal-Wallis test; Figure 6A ), whereas the fluctuation of β-diversity was lost in the LPA group ( Figure 6B ). Approximately 0.62% of OTUs in the HPA group and 1.04% of OTUs in the LPA group exhibited diurnal oscillation ( Figure 6C ). In the HPA group at the phylum level, TM7 exhibited diurnal oscillation ( Figure 6D , Supplementary Figure S4A ). At the class level, TM7_3 exhibited diurnal oscillation ( Supplementary Figure S4B ). No order, family, or species showed diurnal oscillation. At the genus level, three genera exhibited diurnal oscillation (Sarcina, Anaerococcus, and Succinivibrio; Figure 6E , Supplementary Figure S4C, D ). In the LPA group at the phylum level, Cyanobacteria exhibited diurnal oscillation ( Figure 6F , Supplementary Figure S4E ). At the class level, none exhibited diurnal oscillation. At the order level, the relative abundance of Pseudomonadales and Rickettsiales exhibited diurnal oscillation ( Supplementary Figure S4F, G ). At the family, genus, and species levels, Pseudomonadaceae, Pseudomonas, and Prevotella melaninogenica exhibited diurnal oscillation ( Figure 6G, H , Supplementary Figure S4H, I ). Among the detected KEGG pathways, the relative abundance of “photosynthesis-antenna protein,” “calcium signaling pathway,” and “pentose phosphate pathway” exhibited diurnal oscillation in the LPA group ( Figure 6I ), whereas no KEGG pathways exhibited diurnal oscillation in the HPA group.