Healthy human gut phageome

Significance

Humans need a stable, balanced gut microbiome (GM) to be healthy. The GM is influenced by bacteriophages that infect bacterial hosts. In this work, bacteriophages associated with the GM of healthy individuals were analyzed, and a healthy gut phageome (HGP) was discovered. The HGP is composed of core and common bacteriophages common to healthy adult individuals and is likely globally distributed. We posit that the HGP plays a critical role in maintaining the proper function of a healthy GM. As expected, we found that the HGP is significantly decreased in individuals with gastrointestinal disease (ulcerative colitis and Crohn’s disease). Together, these results reveal a large community of human gut bacteriophages that likely contribute to maintaining human health.

Abstract

The role of bacteriophages in influencing the structure and function of the healthy human gut microbiome is unknown. With few exceptions, previous studies have found a high level of heterogeneity in bacteriophages from healthy individuals. To better estimate and identify the shared phageome of humans, we analyzed a deep DNA sequence dataset of active bacteriophages and available metagenomic datasets of the gut bacteriophage community from healthy individuals. We found 23 shared bacteriophages in more than one-half of 64 healthy individuals from around the world. These shared bacteriophages were found in a significantly smaller percentage of individuals with gastrointestinal/irritable bowel disease. A network analysis identified 44 bacteriophage groups of which 9 (20%) were shared in more than one-half of all 64 individuals. These results provide strong evidence of a healthy gut phageome (HGP) in humans. The bacteriophage community in the human gut is a mixture of three classes: a set of core bacteriophages shared among more than one-half of all people, a common set of bacteriophages found in 20–50% of individuals, and a set of bacteriophages that are either rarely shared or unique to a person. We propose that the core and common bacteriophage communities are globally distributed and comprise the HGP, which plays an important role in maintaining gut microbiome structure/function and thereby contributes significantly to human health.

The human body supports a diverse community of microorganisms, the majority of which are thought to be bacteria residing in the lower gastrointestinal tract (1–4). A “healthy gut microbiome (GM)” concept, where similar, but not the same microbes provide common functions that promote human health is supported by metagenomic sequencing of microbes and comparative analyses of gene content (3). The healthy-GM structure is conserved among individuals at higher taxonomic levels (especially phylum) (2, 4), and a core GM at the species level shared by more than 90% of the individuals. Shared bacterial species have been proposed to form a “core bacterial community” that provides beneficial functions (3). This concept, however, is based almost entirely on bacterial sequence data and, with few exceptions (5–8), has not considered the role of bacteriophages.

Gut bacteria host a diverse bacteriophage community (phageome) that potentially helps determine microbial colonization and contributes to shaping GM structure and function. The active phageome of a healthy human (i.e., actively replicating as opposed to nonreplicating, integrated prophage) has been estimated to comprise 35–2,800 viruses (6, 8), the vast majority of which have no significant sequence homology to known bacteriophages. Although the phageome was found to be relatively stable within a person (6–8), comparative phageome analysis at a global scale has received little attention.

To address large-scale phageome overlap, we assembled metagenomic reads from a novel, deep bacteriophage dataset from stool samples of two healthy individuals and analyzed currently available data from 62 healthy people around the world. This analysis identified 23 bacteriophages that were present in more than one-half of all individuals. We then applied a sequence-based viral community network tool (9) to identify closely related viruses at higher taxonomic levels. This analysis identified 44 bacteriophage groups, of which 9 (20%) were shared across more than one-half of individuals around the globe. We propose shared bacteriophages as a healthy gut phageome (HGP) and that, together with their core bacterial hosts, play an essential role in maintaining GM function in the healthy human gut.

Results

Previous studies that identified a core GM in healthy people showed that increasing the depth of sequencing was critical to detect shared bacterial species (3). Thus, we conducted an ultradeep baseline study of the active human gut phageome of two healthy, unrelated adults at two sampling intervals (15 or 3 mo). We focused our efforts on DNA bacteriophages because of their relevance to host bacteria and because most (>95%) RNA viruses in the human gut are likely plant viruses (10). Total DNA was extracted from stool samples as well as purification of virus-like particles (VLPs). TEM analysis of VLPs showed a high concentration of head-tailed bacteriophages consistent with their taxonomic classification in the Caudovirales order (Podoviridae, Siphoviridae, and Myoviridae). Microviridae bacteriophages were also identified, and no eukaryotic virus sequences were found. 16S rDNA-based metagenomics (11) of total stool DNA confirmed that the bacterial community of both individuals was typical of healthy human adults, dominated by Bacteriodetes and Firmicutes. Also, as expected, the GM structure was stable between time points. Unlike previous studies, DNA extracted from VLPs was sequenced without prior amplification to avoid potential amplification bias. Sequencing yielded a total of 8.1 Gb of data, representing one of the most in-depth human gut phageome studies to date, even after deduplication of reads (Table S1). Rank abundance analysis indicated that sufficient sequencing depth was achieved per sample to detect dominant bacteriophages (Figs. S1 and S2).

Table S1.

Sequencing depth compared between datasets used in this study

Dataset	No. of reads	No. of unique reads	% Unique reads	Length (bp)*	No. of total bp	No. of unique bp	n	bp per individual	Unique bp per individual	Unique sequencing depth comparison
Reyes	1,318,248	1,229,707	93	251†	331,454,378	312,560,713	12	27,621,198	26,046,726	82.23
Minot	907,909	904,562	100	544†	493,947,707	492,539,860	6	82,324,618	82,089,977	26.09
Norman	153,591,468	80,230,006	52	250	38,397,867,000	19,057,800,725	62	619,320,435	307,383,883	6.97
Manrique 1.1	5,813,020	3,588,782	62	300	1,743,906,000	974,237,582	1	4,031,007,600	2,278,580,409	0.94
Manrique 1.2	7,623,672	4,945,215	65	300	2,287,101,600	1,304,342,827	1
Manrique 2.1	6,631,762	3,621,775	55	300	1,989,528,600	957,459,794	1	4,126,410,600	2,005,029,078	1.07
Manrique 2.2	7,122,940	3,895,252	55	300	2,136,882,000	1,047,569,284	1

n = number of individuals in study. Unique sequencing depth was calculated by dividing total of unique base pairs per individual in dataset by total unique base pairs per individual in Manrique dataset.

^*Length of untrimmed deduplicated reads.

^†Average read length. Original reads length was heterogeneous.

Fig. S1.

Bacteriophage community abundance distribution (A and B, individual 1; C and D, individual 2). (Left) Relative abundance of the most abundant bacteriophages (10–0.1% range) in the gut of study individuals. (Right) Percentage of normalized reads that are recruited to percentage of contigs shown in different abundance ranges of the community. On average, 7% of the bacteriophage contigs in each community account for 91% of the normalized reads, showing that a minority of the bacteriophage species dominates the bacteriophage community.

Fig. S2.

Percentage of bacteriophages present in more than 20% of healthy individuals. A total of 62 healthy individuals from three cohorts around the world were screened for all of the bacteriophages (complete and partial) assembled in this study (4,301). In total, 160 bacteriophages were present in more than 20% of the individuals, and the percentages of individuals that harbor them in each cohort are represented as a heat map. Each line of the heat map represents a virus.

Four individual active phageome datasets were assembled together resulting in 4,301 contigs (ALL). A total of 70 complete circular and 2 nearly-full length linear phage genomes, based on end sequence comparison were identified (Table S2; referred to as “complete” genomes). The remaining contigs were potentially complete or partial phage genomes. On average, 73% (range, 66–79%) of sequencing reads from each sample assembled into the ALL contigs, demonstrating a relatively even contribution from both individuals. Consistent with previous studies (6, 8), the diversity of the phage community was low (Shannon diversity index = 4.25; SD = 0.95), and the phage distribution was similar in each person, being dominated by a few phages, with no more than 115 genomes accounting for 75% of the normalized reads (Fig. S1). A total of 1,703 of the 4,301 contigs (40%), and 63 of the 72 (88%) complete genomes contained at least one ORF with homology to a phage protein demonstrating their viral origins. However, only 314 (7%) of ALL contigs contained a family taxon-specific phage orthologous group (mostly to Caudovirales). This indicates that the majority (60%) of active bacteriophages in the GM are novel, with only a limited subset that can be taxonomically classified. Taken together, this approach was successful in assembling complete phage genomes and showed the expected community structure.

Table S2.

Information about complete genomes

Phage group in network	Phage	Genome length	Healthy individuals with phage (Norman dataset), n = 62	Percentage of healthy individuals with phage	Taxonomy	Type of genome
Not in network	Phage 375	5,820	59	95	Not classified	Circular
13	Phage 31	96,711	57	92	Not classified	Circular
Not in network	Phage 468	6,401	53	85	Microviridae	Circular
37	Phage 48	56,540	46	74	Not classified	Circular
12	Phage 345	5,835	45	73	Microviridae	Circular
41	Phage 25	84,871	39	63	Caudovirales	Circular
7	Phage 58	61,291	38	61	Caudovirales	Circular
12	Phage 536	5,769	38	61	Not classified	Linear
37	Phage 14	96,114	37	60	Caudovirales	Circular
41	Phage 81	85,111	32	52	Caudovirales	Circular
13	Phage 26	96,024	30	48	Not classified	Circular
23	Phage 23	94,753	28	45	dsDNA viruses	Circular
23	Phage 18	96,620	26	42	dsDNA viruses	Circular
15	Phage 11	156,560	26	42	Not classified	Circular
Not in network	Phage 148	36,364	25	40	dsDNA viruses	Circular
7	Phage 93	45,306	24	39	Caudovirales	Circular
Not in network	Phage 20	102,927	22	35	Not classified	Circular
29	Phage 51	43,881	21	34	Caudovirales	Circular
37	Phage 115	39,855	20	32	Caudovirales	Circular
16	Phage 68	57,271	20	32	Caudovirales	Circular
11	Phage 29	96,982	20	32	Not classified	Circular
6	Phage 85	45,769	20	32	Caudovirales	Circular
11	Phage 33	96,365	19	31	Not classified	Circular
37	Phage 97	43,545	18	29	dsDNA viruses	Circular
15	Phage 13	152,970	18	29	Not classified	Circular
15	Phage 12	155,773	18	29	Not classified	Circular
6	Phage 67	56,277	17	27	Caudovirales	Circular
29	Phage 82	45,634	16	26	Caudovirales	Circular
Not in network	Phage 101	43,360	16	26	Caudovirales	Circular
6	Phage 15	140,123	15	24	Caudovirales	Circular
Not in network	Phage 10	177,409	15	24	Not classified	Circular
Not in network	Phage 449	5,755	13	21	Not classified	Circular
Not in network	Phage 516	5,504	13	21	Not classified	Circular
Not in network	Phage 24	99,239	12	19	Not classified	Circular
24	Phage 76	51,439	11	18	Not classified	Circular
Not in network	Phage 43	77,090	10	16	Caudovirales	Circular
24	Phage 77	51,074	9	15	Not classified	Circular
13	Phage 36	87,718	8	13	Not classified	Circular
6	Phage 141	37,732	7	11	Caudovirales	Circular
13	Phage 35	90,345	7	11	Not classified	Circular
7	Phage 61	41,548	6	10	dsDNA viruses	Circular
7	Phage 106	41,925	6	10	dsDNA viruses	Circular
Not in network	Phage 405	7,603	5	8	Not classified	Circular
Not in network	Phage 98	43,537	5	8	Caudovirales	Linear
16	Phage 132	35,573	4	6	Caudovirales	Circular
29	Phage 198	23,531	4	6	Caudovirales	Circular
29	Phage 136	37,523	4	6	Caudovirales	Circular
8	Phage 94	44,194	4	6	dsDNA viruses	Circular
Not in network	Phage 108	40,910	4	6	Caudovirales	Circular
7	Phage 110	41,293	3	5	dsDNA viruses	Circular
Not in network	Phage 73	44,832	3	5	Caudovirales	Circular
Not in network	Phage 158	34,522	3	5	Caudovirales	Circular
Not in network	Phage 70	50,286	3	5	Caudovirales	Circular
7	Phage 46	77,094	2	3	Caudovirales	Circular
Not in network	Phage 127	38,570	2	3	dsDNA viruses	Circular
Not in network	Phage 72	54,603	2	3	Myoviridae	Circular
Not in network	Phage 498	6,047	2	3	Not classified	Circular
Not in network	Phage 2126	1,077	2	3	Not classified	Circular
6	Phage 142	37,127	1	2	Caudovirales	Circular
35	Phage 38	84,151	1	2	dsDNA viruses	Circular
41	Phage 47	75,845	1	2	Caudovirales	Circular
Not in network	Phage 71	54,702	1	2	dsDNA viruses	Circular
Not in network	Phage 171	31,919	1	2	Caudovirales	Circular
Not in network	Phage 145	37,040	1	2	Caudovirales	Circular
4	Phage 622	3,811	0	0	Not classified	Circular
6	Phage 107	40,223	0	0	dsDNA viruses	Circular
10	Phage 66	44,776	0	0	Podoviridae	Circular
20	Phage 32	96,526	0	0	Not classified	Circular
Not in network	Phage 55	61,931	0	0	dsDNA viruses	Circular
Not in network	Phage 545	5,360	0	0	Microviridae	Circular
Not in network	Phage 2955	1,002	0	0	Not classified	Circular
Not in network	Phage 1675	1,732	0	0	Not classified	Circular

To maximize the identification of shared phages between our two test subjects, reads from the four samples were mapped to the ALL contigs. Recruitment of a single read was used as evidence of phage presence in a particular sample, allowing the identification of both high- and low-abundance phages that were not originally assembled due to low coverage. As has been shown previously (6, 8), phageome structure was more similar within individuals over time [Bray–Curtis (BC) distance = 0.47; SD = 0.24] than between individuals (BC distance = 0.96; SD = 0.01). However, 17% of ALL contigs were present in both individuals in at least one time point, and 7% were present in both individuals at all time points. Of the 72 complete genomes, 56 (78%) were found in both individuals. These results reveal that there is a smaller but significant fraction of phage sequences shared between these two unrelated test subjects.

Given the high level of interindividual phageome variability previously reported, the identification of significant phageome overlap between our two test subjects was surprising, but consistent with the existence of a core microbial community identified by others (3, 12). This result led us to investigate the presence of ALL contigs in other human gut phageome datasets. The dataset by Norman et al. (7) was selected for comparison based on the large number of healthy (n = 62) and diseased subjects (n = 102), their geographic distribution, and the sequencing depth achieved (Table S1). The dataset was deduplicated, and 11% of the total reads were mapped to ALL contigs (Table S3). In total, 1,787 of the 4,301 ALL contigs (42%) were present in at least one person from the Norman dataset; 1,679 (39%) were present in 2–19% of people (termed “low overlap” bacteriophages), 132 (3%) were detected in 20–50% (termed “common” bacteriophages), and 23 (0.5%) were present in >50% (termed “core” bacteriophages) (Fig. S2). Of the 72 complete genomes from our two subjects, 10 (14%) were core and 23 (32%) were common bacteriophages (Fig. 1). One of the core bacteriophages is the recently identified “crAssphage” (13). Our independent identification of crAssphage supports the robustness of our approach and highlights the discovery of a significantly expanded set of core GM bacteriophages. Given their prevalence, we refer to both core and common bacteriophages as the “healthy gut phageome,” or HGP. It is important to note that the 155 bacteriophages comprising the HGP represented only 4% of the estimated total bacteriophage community, and we presume that a larger set of HGP bacteriophages will be discovered with increasing sequencing depth of additional individuals. Also, the observation that some HGP members were detected by a single sequencing read suggests that they are either rare members of the GM or that the sequencing depth of the Norman dataset was too low to accurately estimate their true presence (Fig. S3). Regardless, these results support the presence of a broadly distributed HGP in healthy human adults.

Table S3.

Number of reads from Reyes, Minot, and Norman datasets recruited to ALL contigs dataset

Cohort	Individual	Total reads	Total reads used	% Reads used	Total unique reads*	% Unique reads*	Unique reads used	% Unique reads used
Minot	Healthy†	907,909	112,535	11	904,562	100	NA	NA
Reyes	Healthy†	1,318,248	348,747	28	1,229,707	93	NA	NA
Boston	Healthy 1	1,855,868	151,886	8.18	751,022	40	30,162	4
Boston	Healthy 2	3,077,606	919,913	29.89	1,459,460	47	322,958	22
Boston	Healthy 3	2,363,534	431,989	18.28	1,415,938	60	235,284	17
Boston	Healthy 4	2,545,426	174,344	6.85	1,138,171	45	75,142	7
Boston	Healthy 5	3,594,714	1,987,264	55.28	1,829,231	51	617,223	34
Boston	Healthy 6	3,032,490	869,133	28.66	1,184,156	39	144,158	12
Boston	Healthy 7	2,754,790	830,673	30.15	1,943,469	71	395,908	20
Boston	Healthy 8	2,736,014	6,236	0.23	1,208,815	44	2,166	0
Boston	Healthy 9	2,677,362	1,165,711	43.54	2,031,749	76	575,048	28
Boston	Healthy 10	1,217,356	37,264	3.06	704,349	58	28,478	4
Boston	Healthy 11	3,963,300	165,613	4.18	2,243,045	57	68,805	3
Boston	Healthy 12	3,410,232	3,095	0.09	1,516,125	44	1,582	0
Boston	Healthy 13	3,212,322	1,384,980	43.11	1,764,791	55	326,411	18
Boston	Healthy 14	2,035,696	1,366,499	67.12	1,197,715	59	530,039	44
Boston	Healthy 15	1,435,100	357,568	24.91	891,091	62	192,196	22
Boston	Healthy 16	5,688,510	837,832	14.73	3,207,930	56	240,724	8
Boston	Healthy 17	4,192,464	647,346	15.44	2,366,494	56	67,784	3
Boston	Healthy 18	2,612,060	56,094	2.15	1,430,980	55	17,392	1
Boston	Healthy 19	2,774,234	276,275	9.96	1,282,998	46	206,015	16
Boston	Healthy 20	3,741,796	860,840	23.00	1,823,341	49	184,955	10
Cambridge	Healthy 21	813,036	40,290	4.95	362,723	45	30,453	8
Cambridge	Healthy 22	2,191,674	53,945	2.46	798,150	36	12,521	2
Cambridge	Healthy 23	3,987,976	1,259,661	31.59	1,487,481	37	212,936	14
Cambridge	Healthy 24	744,306	96,954	13.02	520,367	70	53,508	10
Cambridge	Healthy 25	1,412,874	9,063	0.64	429,886	30	6,229	1
Cambridge	Healthy 26	2,148,478	20,278	0.94	1,200,855	56	14,646	1
Cambridge	Healthy 27	2,167,272	695,799	32.10	1,261,182	58	364,978	29
Cambridge	Healthy 28	1,982,540	145,999	7.36	734,000	37	15,439	2
Cambridge	Healthy 29	2,456,352	1,699,167	69.17	1,325,059	54	562,147	42
Cambridge	Healthy 30	2,205,292	343,671	15.58	1,183,795	54	144,100	12
Cambridge	Healthy 31	2,242,066	6,184	0.27	979,536	44	4,666	0
Cambridge	Healthy 32	1,666,460	618,605	37.12	1,013,793	61	266,076	26
Cambridge	Healthy 33	873,858	2,584	0.29	391,024	45	1,836	0
Cambridge	Healthy 34	889,734	34,762	3.90	399,021	45	14,981	4
Cambridge	Healthy 35	2,477,270	82,617	3.33	864,305	35	49,443	6
Cambridge	Healthy 36	2,755,938	1,066,225	38.69	1,458,486	53	251,497	17
Cambridge	Healthy 37	2,747,478	160,672	5.85	1,227,408	45	64,346	5
Cambridge	Healthy 38	2,964,822	600,366	20.25	1,263,739	43	143,055	11
Cambridge	Healthy 39	3,169,706	511,639	16.14	1,242,348	39	128,583	10
Cambridge	Healthy 40	3,101,720	225,445	7.27	953,805	31	47,335	5
Cambridge	Healthy 41	3,488,468	170,214	4.88	2,002,130	57	129,009	6
Chicago	Healthy 42	2,403,326	14,722	0.61	1,391,663	58	5,223	0
Chicago	Healthy 43	2,410,296	509,131	21.12	1,345,791	56	95,852	7
Chicago	Healthy 44	2,368,680	580,817	24.52	1,321,717	56	133,899	10
Chicago	Healthy 45	2,213,436	530,977	23.99	1,395,454	63	289,514	21
Chicago	Healthy 46	3,332,532	72,812	2.18	1,983,779	60	22,311	1
Chicago	Healthy 47	1,886,752	1,195,555	63.36	601,889	32	92,555	15
Chicago	Healthy 48	2,414,120	307,535	12.74	1,347,958	56	193,963	14
Chicago	Healthy 49	1,892,214	4,268	0.22	511,371	27	3,383	1
Chicago	Healthy 50	2,731,028	180,879	6.62	1,702,024	62	58,545	3
Chicago	Healthy 51	1,968,640	185,476	9.42	1,191,892	61	37,410	3
Chicago	Healthy 52	1,981,600	19,512	0.98	1,374,198	69	8,691	1
Chicago	Healthy 53	2,704,256	101,438	3.75	2,047,476	76	53,442	3
Chicago	Healthy 54	2,699,190	205,790	7.62	1,470,581	54	38,420	3
Chicago	Healthy 55	2,733,304	72,812	2.66	1,624,789	59	56,216	3
Chicago	Healthy 56	2,574,044	581,983	22.61	1,301,623	51	123,267	9
Chicago	Healthy 57	2,658,816	664,954	25.01	1,496,994	56	295,962	20
Chicago	Healthy 58	2,182,842	112,919	5.17	1,481,366	68	90,999	6
Chicago	Healthy 59	2,426,646	444,779	18.33	1,598,184	66	332,880	21
Chicago	Healthy 60	1,590,130	468,032	29.43	742,336	47	79,208	11
Chicago	Healthy 61	1,570,368	477,413	30.40	1,108,109	71	268,227	24
Chicago	Healthy 62	1,441,054	2,383	0.16	700,849	49	1,598	0
Boston	Disease 1	2,388,256	24,222	1.01	NA	NA	22,654	NA
Boston	Disease 2	2,363,664	1,398,078	59.15	NA	NA	574,102	NA
Boston	Disease 3	2,575,262	1,589,148	61.71	NA	NA	730,253	NA
Boston	Disease 4	2,134,480	94,343	4.42	NA	NA	62,443	NA
Boston	Disease 5	2,453,782	54,306	2.21	NA	NA	43,918	NA
Boston	Disease 6	1,416,910	58,142	4.10	NA	NA	51,045	NA
Boston	Disease 7	2,183,796	100,330	4.59	NA	NA	44,764	NA
Boston	Disease 8	2,992,536	803,376	26.85	NA	NA	315,793	NA
Boston	Disease 9	3,079,596	109,046	3.54	NA	NA	86,546	NA
Boston	Disease 10	1,209,042	153,107	12.66	NA	NA	49,222	NA
Boston	Disease 11	2,651,402	929,362	35.05	NA	NA	541,354	NA
Boston	Disease 12	2,112,430	14,213	0.67	NA	NA	10,424	NA
Boston	Disease 13	2,506,052	3,550	0.14	NA	NA	2,768	NA
Boston	Disease 14	835,412	143,716	17.20	NA	NA	91,479	NA
Boston	Disease 15	614,422	104,165	16.95	NA	NA	47,673	NA
Boston	Disease 16	2,773,122	107,708	3.88	NA	NA	55,308	NA
Boston	Disease 17	2,349,140	416,753	17.74	NA	NA	289,689	NA
Boston	Disease 18	4,892,418	118,074	2.41	NA	NA	47,234	NA
Boston	Disease 19	3,930,552	6,221	0.16	NA	NA	5,446	NA
Boston	Disease 20	3,374,786	194,894	5.78	NA	NA	106,648	NA
Boston	Disease 21	1,162,722	4,247	0.37	NA	NA	4,154	NA
Boston	Disease 22	2,123,106	1,384,249	65.20	NA	NA	595,006	NA
Boston	Disease 23	2,969,610	1,806,143	60.82	NA	NA	971,425	NA
Boston	Disease 24	1,213,940	1,195	0.10	NA	NA	1,099	NA
Boston	Disease 25	2,182,894	229,344	10.51	NA	NA	76,412	NA
Cambridge	Disease 26	785,338	39,670	5.05	NA	NA	32,284	NA
Cambridge	Disease 27	1,840,304	304,711	16.56	NA	NA	226,167	NA
Cambridge	Disease 28	5,050,694	1,315,276	26.04	NA	NA	748,814	NA
Cambridge	Disease 29	1,014,488	85,600	8.44	NA	NA	41,023	NA
Cambridge	Disease 30	1,205,032	23,593	1.96	NA	NA	22,036	NA
Cambridge	Disease 31	2,494,434	279,913	11.22	NA	NA	134,044	NA
Cambridge	Disease 32	807,868	9,754	1.21	NA	NA	8,814	NA
Cambridge	Disease 33	1,978,690	930,080	47.00	NA	NA	500,668	NA
Cambridge	Disease 34	2,577,566	46,319	1.80	NA	NA	42,910	NA
Cambridge	Disease 35	2,102,706	106,816	5.08	NA	NA	94,586	NA
Cambridge	Disease 36	2,815,246	438	0.02	NA	NA	434	NA
Cambridge	Disease 37	2,199,754	277,888	12.63	NA	NA	203,917	NA
Cambridge	Disease 38	2,110,646	289	0.01	NA	NA	289	NA
Cambridge	Disease 39	1,016,694	24,639	2.42	NA	NA	21,817	NA
Cambridge	Disease 40	738,246	5,724	0.78	NA	NA	5,433	NA
Cambridge	Disease 41	2,096,784	171,075	8.16	NA	NA	146,212	NA
Cambridge	Disease 42	3,247,232	2,295,486	70.69	NA	NA	813,457	NA
Cambridge	Disease 43	554,096	3,318	0.60	NA	NA	3,291	NA
Cambridge	Disease 44	1,544,178	903,730	58.52	NA	NA	452,230	NA
Cambridge	Disease 45	2,648,994	132,091	4.99	NA	NA	115,464	NA
Cambridge	Disease 46	2,091,574	112,650	5.39	NA	NA	73,381	NA
Cambridge	Disease 47	1,559,028	1,492	0.10	NA	NA	1,489	NA
Cambridge	Disease 48	2,670,292	26,551	0.99	NA	NA	25,670	NA
Cambridge	Disease 49	2,056,264	5,923	0.29	NA	NA	5,352	NA
Cambridge	Disease 50	3,119,110	81,631	2.62	NA	NA	63,264	NA
Cambridge	Disease 51	945,096	188,307	19.92	NA	NA	63,761	NA
Cambridge	Disease 52	1,047,788	21,311	2.03	NA	NA	20,805	NA
Cambridge	Disease 53	947,354	78,380	8.27	NA	NA	35,183	NA
Cambridge	Disease 54	1,729,208	736,357	42.58	NA	NA	163,249	NA
Cambridge	Disease 55	2,589,948	3,042	0.12	NA	NA	3,019	NA
Cambridge	Disease 56	631,500	19,934	3.16	NA	NA	19,197	NA
Cambridge	Disease 57	2,524,278	4,144	0.16	NA	NA	4,121	NA
Cambridge	Disease 58	1,362,442	245,797	18.04	NA	NA	87,344	NA
Cambridge	Disease 59	2,806,760	1,194,774	42.57	NA	NA	250,341	NA
Cambridge	Disease 60	2,213,584	544,664	24.61	NA	NA	170,054	NA
Cambridge	Disease 61	2,412,222	356	0.01	NA	NA	353	NA
Cambridge	Disease 62	2,639,446	41,137	1.56	NA	NA	38,139	NA
Cambridge	Disease 63	3,800,714	66,612	1.75	NA	NA	57,373	NA
Cambridge	Disease 64	4,179,442	32,927	0.79	NA	NA	30,340	NA
Cambridge	Disease 65	2,927,868	418	0.01	NA	NA	415	NA
Cambridge	Disease 66	4,635,050	592,274	12.78	NA	NA	366,585	NA
Cambridge	Disease 67	4,126,486	140,083	3.39	NA	NA	117,236	NA
Cambridge	Disease 68	2,277,294	2,755	0.12	NA	NA	2,697	NA
Cambridge	Disease 69	2,963,800	1,116,595	37.67	NA	NA	346,965	NA
Cambridge	Disease 70	1,859,700	1,163	0.06	NA	NA	1,158	NA
Cambridge	Disease 71	2,307,386	10,457	0.45	NA	NA	10,051	NA
Cambridge	Disease 72	2,041,424	8,242	0.40	NA	NA	7,926	NA
Cambridge	Disease 73	2,621,932	227,027	8.66	NA	NA	161,484	NA
Cambridge	Disease 74	2,007,874	219	0.01	NA	NA	219	NA
Cambridge	Disease 75	3,611,160	831,708	23.03	NA	NA	366,327	NA
Cambridge	Disease 76	1,596,724	1,087	0.07	NA	NA	1,079	NA
Cambridge	Disease 77	2,670,084	1,511	0.06	NA	NA	1,469	NA
Chicago	Disease 78	2,299,262	1,010	0.04	NA	NA	1,005	NA
Chicago	Disease 79	2,565,346	139	0.01	NA	NA	138	NA
Chicago	Disease 80	2,342,156	68,152	2.91	NA	NA	49,162	NA
Chicago	Disease 81	3,298,802	1,464,153	44.38	NA	NA	626,228	NA
Chicago	Disease 82	2,539,358	1,901	0.07	NA	NA	1,855	NA
Chicago	Disease 83	2,981,682	281,269	9.43	NA	NA	201,063	NA
Chicago	Disease 84	2,394,688	91,325	3.81	NA	NA	77,979	NA
Chicago	Disease 85	3,830,506	1,266,358	33.06	NA	NA	715,601	NA
Chicago	Disease 86	3,502,954	16,400	0.47	NA	NA	15,722	NA
Chicago	Disease 87	3,352,598	149,741	4.47	NA	NA	73,399	NA
Chicago	Disease 88	2,929,918	331,619	11.32	NA	NA	252,323	NA
Chicago	Disease 89	3,473,508	1,074,750	30.94	NA	NA	509,394	NA
Chicago	Disease 90	1,214,702	3,317	0.27	NA	NA	3,220	NA
Chicago	Disease 91	3,825,838	21,981	0.57	NA	NA	20,243	NA
Chicago	Disease 92	2,431,028	388,718	15.99	NA	NA	110,424	NA
Chicago	Disease 93	2,125,146	248,822	11.71	NA	NA	131,995	NA
Chicago	Disease 94	2,879,428	469,166	16.29	NA	NA	115,545	NA
Chicago	Disease 95	2,267,290	270,505	11.93	NA	NA	63,267	NA
Chicago	Disease 96	2,804,470	579,565	20.67	NA	NA	187,402	NA
Chicago	Disease 97	3,202,992	184,491	5.76	NA	NA	64,034	NA
Chicago	Disease 98	2,759,690	24,435	0.89	NA	NA	23,751	NA
Chicago	Disease 99	1,092,508	6,756	0.62	NA	NA	6,499	NA
Chicago	Disease 100	3,637,688	328,733	9.04	NA	NA	141,947	NA
Chicago	Disease 101	3,119,986	396,859	12.72	NA	NA	183,655	NA
Chicago	Disease 102	2,732,006	1,316,100	48.17	NA	NA	522,713	NA

^*Initially, total reads (unduplicated) were mapped to ALL contigs. In healthy individuals, all reads were deduplicated and then mapped to our contigs. Because deduplication did not modify our results, in disease individuals only used reads (reads that mapped to our contigs initially) were further trimmed, deduplicated, and remapped.

^†Individuals from this dataset were not analyzed separately.

Fig. 1.

Heat map indicating the presence of the 72 complete bacteriophage genomes in the globally distributed 62 healthy individuals represented in the Norman dataset (7). The percentage of healthy individuals from three global locations (Boston, Cambridge, Chicago) that harbor each of the complete bacteriophage genomes is indicated. Nine bacteriophages, including crAssphage (bacteriophage 31), were present in >50% of the individuals (core), and 46% were found in >20% of the individuals (core and common). Each line is a bacteriophage (phage).

Fig. S3.

Detection of HGP bacteriophage using increasing detection threshold. Core healthy gut phageome (HGP) bacteriophages were identified in individuals by read recruitment of high-quality deduplicated reads to bacteriophage contigs. If there was one read recruited, the virus was considered present in an individual. The percentage of individuals in which we could detect these bacteriophages using increasing detection thresholds (1 read versus 2, 5, 10, 50, or 100) is represented.

To add confidence to these results, we reversed the process by first independently assembling the bacteriophage community from healthy individuals within the Norman dataset, followed by read recruitment from all of the individuals. We observed the same general findings. Of the 13,707 bacteriophage contigs from the 62 Norman healthy individuals, 1,408 (9%) were found in more than 20% of the individuals and 115 (0.83%) were found in more than 50% of the individuals. The average size of these contigs was 7 kb, compared with 32 kb of the HGP bacteriophages from our ALL dataset. This suggested that these contigs could be partial fragments from the core and common bacteriophages identified in the two test individuals. When we compared the HGPs from the two test individuals with the Norman HGP, we identified all of the HGP bacteriophages initially identified in our ALL set. Of the 1,408 shared bacteriophages in the Norman dataset, 1,013 (71%) were highly related to the HGP. We also identified additional potential core bacteriophages (n = 10) and common bacteriophages (n = 86) (Table S4). These results show that our approach of using only two test individuals but ultradeep sequencing their bacteriophage community was sufficient to detect a HGP. We speculate that a much larger HGP will be identified by analyzing additional healthy individuals.

Table S4.

HGP contigs identified in Norman individuals not previously identified in study individuals

Phage contig	Network group*	Length, bp	Cohort	Type of phage	% Individuals with contig
Cross_Contig_26	1	37,893	Cross	Common	23
Cross_Contig_43	1	21,615	Cross	Common	38
Cross_Contig_241	1	12,042	Cross	Common	34
Boston_contig_469	1	6,505	Boston	Common	27
Boston_contig_638	1	5,530	Boston	Common	31
Boston_contig_682	1	5,315	Boston	Common	28
Chicago_contig_492	1	5,284	Chicago	Common	20
Cambridge_contig_529	1	5,267	Cambridge	Common	23
Boston_contig_353	2	8,332	Boston	Common	20
Boston_contig_194	3	13,150	Boston	Common	28
Cambridge_contig_359	3	6,330	Cambridge	Core	58
Cambridge_contig_378	3	6,169	Cambridge	Core	56
Boston_contig_524	3	6,163	Boston	Common	38
Chicago_contig_388	3	6,120	Chicago	Core	55
Cambridge_contig_392	3	6,079	Cambridge	Core	53
Boston_contig_540	3	6,068	Boston	Common	30
Boston_contig_1250	3	5,858	Boston	Common	41
Chicago_contig_459	3	5,565	Chicago	Common	38
Chicago_contig_515	3	5,176	Chicago	Common	25
Cross_Contig_124	11	10,476	Cross	Common	34
Cambridge_contig_553	11	5,122	Cambridge	Common	20
Cross_Contig_11	20	94,996	Cross	Common	23
Cambridge_contig_58	20	20,080	Cambridge	Common	20
Cross_Contig_309	22	5,673	Cross	Common	34
Boston_contig_86	23	22,335	Boston	Common	22
Cross_Contig_21	29	14,389	Cross	Common	20
Cross_Contig_67	29	13,787	Cross	Common	22
Boston_contig_579	29	5,891	Boston	Common	22
Boston_contig_11	41	51,927	Boston	Common	30
Cross_Contig_198	41	30,179	Cross	Common	30
Cross_Contig_240	41	12,059	Cross	Common	22
Cross_Contig_16	42	42,054	Cross	Common	31
Cambridge_contig_463	42	7,029	Cambridge	Common	22
Cross_Contig_288	42	6,952	Cross	Common	20
Cross_Contig_327	42	5,106	Cross	Common	25
Cambridge_contig_313	49	6,699	Cambridge	Common	39
Chicago_contig_530	49	5,040	Chicago	Common	34
Chicago_contig_310	59	6,806	Chicago	Common	38
Boston_contig_480	59	6,446	Boston	Common	22
Cambridge_contig_722	59	6,289	Cambridge	Common	33
Chicago_contig_670	59	5,589	Chicago	Common	27
Cambridge_contig_544	59	5,176	Cambridge	Common	23
Cross_Contig_125	68	10,471	Cross	Common	48
Cambridge_contig_289	68	7,087	Cambridge	Common	44
Chicago_contig_309	69	6,807	Chicago	Core	52
Boston_contig_455	69	6,683	Boston	Core	53
Boston_contig_472	69	6,487	Boston	Common	31
Cambridge_contig_380	69	6,161	Cambridge	Core	56
Cross_Contig_265	75	8,774	Cross	Common	23
Chicago_contig_362	76	6,258	Chicago	Common	28
Boston_contig_62	77	26,488	Boston	Common	20
Boston_contig_246	77	10,888	Boston	Common	20
Cambridge_contig_480	80	5,603	Cambridge	Core	50
Boston_contig_76	82	23,429	Boston	Common	42
Boston_contig_446	83	6,793	Boston	Common	20
Cambridge_contig_326	83	6,605	Cambridge	Common	20
Chicago_contig_389	83	6,115	Chicago	Common	25
Cross_Contig_107	86	30,526	Cross	Common	28
Cross_Contig_10	86	29,265	Cross	Common	33
Cross_Contig_203	86	27,447	Cross	Common	39
Cross_Contig_232	86	14,021	Cross	Common	27
Chicago_contig_156	86	10,806	Chicago	Common	30
Cross_Contig_136	86	6,963	Cross	Common	31
Cambridge_contig_413	86	5,961	Cambridge	Common	27
Cross_Contig_32	90	15,929	Cross	Common	28
Cambridge_contig_372	100	6,226	Cambridge	Common	25
Cambridge_contig_381	100	6,158	Cambridge	Common	20
Chicago_contig_391	100	6,097	Chicago	Common	27
Boston_contig_624	100	5,627	Boston	Common	27
Boston_contig_607	103	8,783	Boston	Common	28
Cambridge_contig_1460	103	6,335	Cambridge	Common	30
Boston_contig_15	112	46,981	Boston	Common	20
Boston_contig_197	116	13,014	Boston	Common	34
Cross_Contig_194	128	54,756	Cross	Common	28
Chicago_contig_34	128	27,151	Chicago	Common	23
Cross_Contig_237	128	12,633	Cross	Common	20
Boston_contig_389	128	7,556	Boston	Common	22
Cambridge_contig_482	128	5,598	Cambridge	Common	20
Boston_contig_238	154	11,135	Boston	Common	20
Cross_Contig_112	171	20,983	Cross	Common	20
Cambridge_contig_99	259	14,728	Cambridge	Common	30
Cambridge_contig_349	355	6,389	Cambridge	Common	28
Boston_contig_498	355	6,330	Boston	Common	25
Cambridge_contig_360	355	6,328	Cambridge	Common	31
Boston_contig_499	355	6,325	Boston	Common	28
Boston_contig_534	355	6,090	Boston	Common	34
Boston_contig_805	355	6,073	Boston	Core	50
Chicago_contig_471	355	5,479	Chicago	Common	41
Chicago_contig_517	355	5,166	Chicago	Common	34
Boston_contig_712	374	5,185	Boston	Common	23
Cambridge_contig_347	387	6,407	Cambridge	Common	23
Boston_contig_512	387	6,218	Boston	Common	34
Cambridge_contig_448	387	5,779	Cambridge	Common	20
Cambridge_contig_318	478	6,674	Cambridge	Common	25
Cross_Contig_82	678	7,268	Cross	Common	23
Cross_Contig_65	686	15,244	Cross	Core	53

“Cross” indicates reference sequence from contigs from different cohorts assembled together through Geneious.

^*Network groups from this analysis do not correspond to the groups from the ALL dataset viral network reported in the main text.

Erroneous mapping of sequencing reads was tested by mapping over 38 million viral metagenomic reads from other environments (marine and hot spring environments) and a synthetic dataset to ALL contigs. No reads were recruited. We used the synthetic dataset to estimate the percentage of true positives by mapping the synthetic reads back to a collection of 1,197 Caudovirales (including the 50 genomes from which the synthetic reads were derived). The error rate used to create this dataset was set at 1.5% (a value set by the program) compared with the estimated Illumina error rate of 0.0046% (14). We found that 93% of the contigs called were true positives (Table S5) and that increasing the number of reads required to call a contig did not significantly change this percentage (Table S5). Given the high error rate in the synthetic dataset, the frequency of miscalled contigs in our experimental datasets is likely much lower (approximately one order of magnitude), providing confidence that using even a single read to identify shared bacteriophages is a valid approach.

Table S5.

Mapping error

Number of reads used to deem a contig present	1 read	5 reads	10 reads	50 reads	100 read
Number of miss-called contigs	88	75	68	65	47
Percentage of miss-called contigs	7.35	6.27	5.68	5.43	3.93
Percentage of reads assigned correctly	92.9	93.6	94.2	95.4	95.7

To investigate whether HGP bacteriophages play a role in maintaining human health, we compared the prevalence of core bacteriophages in healthy individuals to those with irritable bowel disease (IBD) [n = 66 with ulcerative colitis (UC); n = 36 with Crohn’s disease (CD)] (7) (Fig. 2). Although core bacteriophages are found on average in 62% of healthy individuals, their prevalence was reduced by 42% and 54% on average in UC and CD patients, respectively (value of P < 0.0001) (Fig. 2A). Moreover, healthy subjects harbored on average 62% of the 23 core bacteriophages; however, in UC and CD patients, the average is significantly reduced to 37% and 30%, respectively (value of P < 0.0001) (Fig. 2B). These results suggest that the HGP is significantly perturbed in diseased patients.

Fig. 2.

Reduction of the HGP present in IBD patients. (A) Comparison of percentage of healthy individuals (H) (n = 62) versus IBD-diseased individuals (CD and UC) (n = 102) that harbor each of the 23 core HGPs identified in this study. (B) Percentage of core, common, and unique bacteriophages (unique and low-overlap bacteriophages were pooled) present in all healthy and IBD individuals. Median is shown with a horizontal black bar. The number of core and common bacteriophages that IBD individuals harbor is significantly lower compared with healthy individuals. Significance was calculated using a Sidak’s multiple-comparison test with a 95% confidence interval.

To analyze the HGP structure, we conducted a network analysis using ALL contigs. The network groups viruses based on sequence homology to identify related viruses and to organize short contigs from partial genomes that dominate metagenomic datasets (9). A total of 240 statistically supported groups was identified (clustering coefficient value of 0.667 and modularity value of 0.675), with 44 containing five or more contigs (Fig. 3A and Table S6). Groups with less than five contigs represented only 11% of the total reads and were excluded to focus on the largest groups of related sequences. Of the 72 complete genomes, 45 (63%) were present and belonged to 17 different groups. One of these groups contained crAssphage and three additional complete crAssphage-related genomes. Previously classified phage genomes were mapped onto the network to test the accuracy of the grouping, and in nearly all cases (25 of 28), phages belonging to the same family grouped together. We also evaluated 172 artificially fragmented Caudovirales genomes. The fragmented genomes clustered as expected, with 20 of 26 being family-specific groups. Only six groups had fragments from more than one family, which is consistent with the high level of horizontal gene transfer shown between bacteriophages (15). These results suggest that the 44 network-defined groups can be considered roughly as family-level groups. Only 14 of the 44 network-defined groups contained classified bacteriophages, supporting that novel bacteriophages are abundant in the human gut. The network analysis proved useful in reducing the complexity of metagenomic datasets by organizing contigs into biologically relevant groups. Whether the members of these groups carry out similar functions in the gut is a subject for future study.

Fig. 3.

Network-based analysis of the bacteriophage community in the human gut from two healthy individuals. (A) Graphical representation of the 44 network groups (colored by group) with greater than five contig membership. Each dot represents a contig. (B) Core groups (present in more than one-half of the individuals) are highlighted and labeled in red. (C) Heat map representing the percentage of individuals that have at least one member of each phage group. (D) Schematic organization of the bacteriophage community associated with the healthy gut microbiome.

Table S6.

Network phage groups membership

Phage group	Partial or complete phage in group	Complete phage in group	Type of group
0	8	0	Common
1	10	0	Common
2	15	0	Common
3	12	0	Unique
4	401	1	Low overlap
5	13	0	Low overlap
6	87	6	Core
7	55	6	Core
8	9	1	Common
9	7	0	Low overlap
10	31	1	Unique
11	5	2	Common
12	21	2	Core
13	38	4	Core
14	8	0	Common
15	34	3	Common
16	8	2	Common
17	6	0	Unique
18	14	0	Common
19	23	0	Unique
20	11	1	Unique
21	10	0	Low overlap
22	6	0	Unique
23	23	2	Core
24	17	2	Low overlap
25	6	0	Unique
26	86	0	Low overlap
27	13	0	Unique
28	5	0	Common
29	70	4	Core
30	21	0	Common
31	39	0	Common
32	18	0	Unique
33	9	0	Unique
34	27	0	Common
35	6	1	Low overlap
36	6	0	Low overlap
37	76	4	Core
38	24	0	Core
39	10	0	Low overlap
40	11	0	Low overlap
41	65	3	Core
42	5	0	Low overlap
43	5	0	Low overlap
Not in network	2,927	27	NA

The network analysis supports the global distribution of the HGP. The majority (41 of 44) of network-defined groups were present in at least one of our test subjects (Table S7); in the Norman study (Fig. 3C), 12 groups were present in 2–20% of individuals (low overlap group); 13 were present in 20–50% (common groups); and 9, including the crAssphage and 2 other groups with complete genome representatives, were present in >50% of healthy individuals (core groups, Fig. 3B). Overall, the viral network represents a more encompassing view of the viral community than can be seen by only looking at the limited number of specific taxa classified by fragmented sequence information and supports the existence of a definable HGP.

Table S7.

Percentage abundance of groups identified by network analysis in study individuals

Light gray indicates group with more than 0.001% of abundance. Dark gray indicates groups that have no members in the specific individual and time point.

*Groups with more than 1% of abundance in at least one individual at one time point.

Discussion

The overall objective of this study was to investigate the presence of a HGP. We propose that the phageome of healthy people can be divided into three classes (Fig. 3D): (i) the core, found in more than one-half of all individuals; (ii) the common, shared among many individuals; and (iii) the low overlap/unique, found in a limited number of individuals. The core is composed of at least 23 bacteriophages in this study and is significantly reduced in UC and CD patients, suggesting that these bacteriophages play a role in maintaining human health. Our results do not account for temporal dynamics in either the GM or phageome. Therefore, our estimates of sharing (>50%, 20–50%, 2–20%) for these bacteriophage categories are conservative and may need to be refined with temporal data from more subjects.

The role of bacteriophages in the human GM is poorly understood. In most bacteria-dominated ecosystems, bacteriophages play a key role in shaping community structure and function (16, 17). Thus, it is expected that the human gut ecosystem is strongly affected by the activity of bacteriophages. Microbe–virus interactions are often dominated by lytic interactions in a prey–predator dynamic (18). However, community sequence analysis of the human gut indicates that lysogenic bacteriophages (prophages) dominate over lytic bacteriophages (6, 8, 19). It is thought that the majority of bacterial cells contain at least one prophage, and that selective activation of a subset of these prophages is what can be detected as the active phageome in stool samples. The dynamics that arise from a lytic–lysogenic balance in the gut are still unknown. However, the active gut bacteriophages have been shown to influence bacterial population dynamics in a murine gut microbial ecosystem model (20), and to confer advantages to certain microbiome members during stress, such as antibiotic exposure (21). We presume that the active phageome of healthy people plays a critical role in maintaining the structure and function of a healthy gut ecosystem. Although the HGP represents a small proportion of the overall potential gut bacteriophage community, elucidating the role of the active phageome in a healthy gut will be essential to fully understanding its potential roles in disease or rehabilitation following episodes of microbiome imbalance.

Identification of a human HGP was somewhat surprising. Currently, two somewhat contrasting views of human gut phageome exist. On one hand, there is evidence of more phageome similarity among relatives and household members compared with unrelated individuals (7, 8). On the other hand, there is evidence that some phages are common to unrelated people (13, 19, 20, 22). Stern et al. (19) identified 991 bacteriophage sequences from cellular metagenomes that were shared across 124 individuals. The majority of these bacteriophages appeared to be integrated prophages, and only a small fraction was present as VLPs. We found an expanded set of shared phages as well as phages that were unique to individuals. It is likely that the core HGP is hosted by core gut bacteria detected by others (3, 12). Our work provides a more comprehensive view and accurate estimate of phageome overlap. We presume even more overlap will be discovered with greater sequencing depth, the application of similar network analytical approaches, and consideration of more individuals.

In humans, viruses are often viewed as deleterious disease-causing agents, but the existence of an HGP suggests that the net influence of bacteriophages in the human gut is not deleterious, but rather beneficial. Despite large differences in the species composition of gut bacteria between people, there is a reproducible, higher-level taxonomic signature (4). This higher-level structure is thought to be a marker of a healthy gut ecosystem. We posit that bacteriophages play an essential role in this ecosystem and that the human HGP, arising from the large pool of cellular prophages, likely contributes to the proper function of the healthy GM by influencing the stability and maintenance of a healthy ecosystem through active lysis and predator–prey dynamics.

Our results increase the known repertoire of shared bacteriophages in healthy humans and expand the number of complete core HGP genomes from one (crAssphage) to nine representatives. Network analysis identified groups of related viruses, and provided a powerful tool to organize and reduce the complexity of fragmented metagenomic data. We suspect that our efforts (sequencing and computation) underestimated the full extent of the human HGP, and that larger studies will generate even more accurate estimates of HGP prevalence, which will allow for a better understanding of the cosmopolitan human phageome. Future studies are needed to address the role of the HGP in health and disease, and to assess the potential use of core bacteriophages for clinical therapeutics and controlled manipulation of the human gut ecosystem.

Materials and Methods

Sample Collection.

Samples were collected from two individuals of 26 and 55 y at two time points over a period from time 0–3 or 15 mo, respectively, under a protocol approved by the Internal Review Board of Montana State University. Both individuals were healthy, with normal bowel movement (at least once every 2 d and no more than three times a day), and had not received antibiotics for >2 mo before the collection of the samples. Collections were carried out with subject’s informed consent under an approved institutional review board protocol.

VLP Isolation, DNA Extraction, and Sequencing.

VLPs were purified based on Minot et al. (6) (modification: 0.45-μm filtration). The 1.5 g/cm³ fractions were pooled, dialyzed, and DNase treated, and DNA was extracted using PureLink Viral DNA/RNA extraction kit (Invitrogen) and sequenced using MiSeq Paired End Illumina Technology.

Bacteriophage Sequences Assemblies.

Bacteriophage sequences from two test individuals were initially assembled using MIRA 4.0 (23) and VICUNA (24) [quality trimmed reads using Trimmomatic v0.32 (25), 25-bp overlap at a minimum of 90% identity, and a minimum contig size of 1,000 bp]. Contigs generated by MIRA and VICUNA were further assembled to each other using Geneious 9.0.2 (contigs had to overlap across 250 bp at 98% identity to be assembled) (26). An ALL dataset was generated by assembling contigs from all time points and individuals together, at 500-bp overlap at a minimum of 98% identity, using Geneious. Norman dataset reads were quality trimmed using Trimmomatic. Individual reads were grouped by cohort, and pooled reads were cross-assembled with IDBA-UD (27) using minimum and maximum kmer lengths of 20 and 120, respectively (7). Contigs longer than 1,000 bp obtained were pooled together and assembled using Geneious and manually curated using the same parameters as in our experimental approach.

Bacteriophage Community Analysis.

Reads were recruited to contigs using Geneious with high-stringency parameters [a read needed to be >98% identical across its entire length (300 bp) to be recruited]. The number of reads per contig was normalized to the length of the contig (number of total base pairs aligned to contig/base pair of contig). The percentage of total normalized reads to a contig was considered the percentage of the bacteriophage community accounted for by a particular contig. BC distances and between communities and Shannon index were calculated using vegan package in R.

Taxonomic Classification and Detection of Bacteriophage Genes.

ORFs were predicted using the WebMGA server. Proteins longer than 100 aa with less than 1% of the length of ambiguities were queried against ACLAME database (e value < 1e-10) and contigs were classified based on Waller (28).

Detection of Contigs in Publicly Available Metagenomes.

Reads from the Norman dataset (7) were quality trimmed using Trimmomatic v0.32 (25) and deduplicated using cd-hit-dup (29). Trimmed unique reads were mapped to ALL contigs using high-stringency parameters [90% identity across the full length of the maximum read length (250 bp) using Geneious]. Reads shorter than this length will not be aligned; thus, these parameters only allow for high-quality deduplicated reads to be recruited to contigs.

Read Recruitment Controls.

Because read length from the hot springs datasets (9) was very heterogeneous (340–855 bp), reads were aligned at 90% over the minimum length of these datasets. Reads from marine metagenomes [deep ocean hydrothermal vent (SRR3133481)] were mapped with the same aligning parameters as the reads from the Norman dataset, adjusting the read length to the size of the reads from this dataset (202 bp). A synthetic dataset was generated from randomly selected 50 genomes from a list of 1,197 complete Caudovirales genomes downloaded from National Center for Biotechnology Information with Grinder (30). The generated 2,000,000 Illumina reads (250 bp) were mapped to ALL contigs and to the set of 1,197 genomes using the same parameters as for Norman reads (see above).

Network Analysis.

The network analysis was carried out based on Bolduc et al. (9). Briefly, network analysis is based on an all-versus-all BLASTn/leuvin search that incorporates both the number and strength of matches between contigs to group-related sequences into discrete groups. BLASTN analysis was performed with default parameters, except e-value cutoff was set to 1E-10 and max target sequences to 10,000. Matches were filtered based on the parameters of high-scoring segment pair (HSP) length > 200 (minimum alignment length) and 75% nucleotide identity. The network algorithm measures the number and weight (HSP and e value) of these connections and clusters sequences into groups of highly related sequences. The network was manually curated and visualized using Gephi with ARF view. A network control was performed on 172 complete Caudovirales genomes associated with bacterial host from the human gut (31). Genomes were fragmented into 9,000 with 50-bp overlap using an in-house python script (9), and network analysis was performed using the same parameters used in the ALL dataset network.

Identification of HGP Bacteriophages in Norman Assemblies.

HGP bacteriophages were identified via network analysis. Contigs from our ALL and Norman dataset were pooled together. An all-versus-all BLASTN analysis was performed and network analysis was applied (see above). Contigs that were in the same group as HGP bacteriophages identified in our analysis were considered closely related HGP bacteriophages.