홀로비온트(holobiont) 인간, 동물, 식물+미생물은 하나의 생물학적, 생태학적 단위

[문형철]

홀로비온트(holobiont) 

호스트(숙주: 동물, 식물, 산호 등 다세포 생물)와

그와 공생하는 미생물(박테리아, 고세균, 바이러스, 진균 등) 전체를

하나의 생물학적·생태학적 단위로 보는 개념.

호스트와 미생물은

서로 긴밀하게 상호작용하며,

함께 진화·적응·기능을 수행하는 ‘통합된 개체’.

이 개념은

홀로게놈(hologenome)(호스트 게놈 + 미생물 게놈 전체)과 함께

진화의 단위로 확대 해석되기도 합니다.

초기에는 공생(symbiosis) 연구에서 출발했으나,

미생물 군집(metagenomics) 기술 발달로

2000년대 후반부터 폭발적으로 연구되었습니다.

사용자님이 요청하신 대로,

개념을 처음 주창한 논문부터 차례로 10개 주요 논문을

연대순으로 정리했습니다.

초기 논문은

Nature/Science/Lancet 수준의 초고영향력 저널이 아니었지만(개념 자체가 이론 생물학·공생학에서 시작),

이후 미생물학·진화생물학 분야에서

고영향력 저널

(Nature Reviews Microbiology, PNAS, PLOS Biology, Cell Metabolism, Science, Nature Communications 등)로

확산되었습니다.

각 논문의 핵심 기여와 개념 발전 과정을

간단히 설명하겠습니다.

1. Margulis, L. (1990). Words as battle cries: Symbiogenesis and the new field of endocytobiology. BioScience, 40(9), 673–677.

최초로 “holobiont” 용어를 사용한 논문. Lynn Margulis(엔도공생설의 주창자)는 호스트와 공생자(bionts)의 결합체를 “holobiont”(전체 생물체)로 명명하며, 진화에서 공생이 핵심임을 강조했습니다. 초기에는 주로 단일 공생체(예: 조류+균류 리켄)를 의미했으나, 현대 미생물 군집 연구의 기초가 되었습니다.

논문은
홀로비온트(holobiont)라는 용어를
생물학 문헌에 최초로 사용한 역사적인 짧은 에세이입니다.

Lynn Margulis(엔도공생설의 주창자)는
기존의 다윈주의 중심 진화론에 도전하며,
공생생성(symbiogenesis)을 진화의 주요 메커니즘으로 강조합니다.

주요 내용 요약

• 전투의 함성( Words as Battle Cries ) 비유: 과학 용어는 단순한 단어가 아니라 “전투의 함성”처럼 새로운 사고방식을 만들어내는 강력한 도구라고 주장합니다. 기존 생물학이 “개별 개체(individual)” 중심이었다면, 공생을 통해 새로운 관점이 필요하다는 메시지입니다.
• 엔도사이토바이올로지(Endocytobiology)의 제안: 세포 내부 공간(intracellular space)을 “oligogenetic ecosystem”(소수의 유전체가 함께 사는 생태계)으로 재정의합니다. 즉, 진핵세포(eukaryotic cell) 내부의 미토콘드리아, 엽록체 등은 원래 별개의 원핵생물이었으며, 공생을 통해 통합된 결과물이라는 엔도공생설(serial endosymbiosis theory)을 바탕으로 새로운 학문 분야를 제안합니다.
• 홀로비온트(holobiont) 용어의 최초 등장: 서로 다른 종(species)의 파트너들이 물리적으로 접촉하며 공생하는 복합체를 holobiont라고 명명합니다.
  • 두 개의 서로 다른 개체(예: 정자와 난자 → zygote)가 합쳐져 새로운 개체가 되는 것처럼,
  • 두 개의 공생자(symbionts)가 합쳐져 holobiont라는 새로운 생물학적 개체(individual)가 형성된다고 설명합니다. 이 과정에는 통합(integration)과 해리(dissociation) 메커니즘이 존재하며, 매 세대마다 반복된다고 봅니다.
• 주요 주장:
  • 공생은 일시적 현상이 아니라, 진화에서 지속적이고 창조적인 힘입니다.
  • 생물학적 개체성(biological individuality)의 경계는 고정된 것이 아니라, 공생 파트너들과 함께 유동적입니다.
  • “우리는 결코 단일 개체가 아니었다(We have never been individuals)”라는 후속 생각의 씨앗이 여기서 시작됩니다.

의의

이 논문 자체는
이론적·철학적 에세이 성격이 강하며,
실험 데이터보다는 개념적 제안과 역사적·철학적 논의가 중심입니다.

Margulis는
이 글에서 holobiont라는 단어를 처음 썼지만,
이후 1991년 책 챕터 등에서
“symbiotic complex”나 “product of association between bionts”로 더 자주 설명했습니다.

현대 홀로비온트 연구자들은
이 1990년 논문을 용어의 기원으로 가장 많이 인용

2. Rosenberg, E. et al. (2007). The role of microorganisms in coral health, disease and evolution. Nature Reviews Microbiology, 5, 355–362.

고영향력 리뷰(Nature Reviews)로, 산호 홀로비온트(corals + Symbiodinium + bacteria)를 모델로 미생물이 호스트 건강·질병·진화에 미치는 역할을 체계화. hologenome 개념의 초기 틀을 제시하며, 미생물이 빠른 적응(rapid evolution)을 가능케 한다고 강조. 고영향력 저널에서 개념을 본격적으로 확산시킨 논문입니다.

이 논문은
산호 건강, 질병, 진화에서
미생물이 하는 핵심 역할을 종합적으로 리뷰한 글입니다.

전 세계적으로 급속히 파괴되고 있는
산호초 문제를 이해하고 보전하기 위해,
산호와 그 공생 미생물의 상호작용을 중점적으로 다룹니다.

핵심 개념

• 홀로바이언트(Holobiont): 산호 숙주 + 공생 미생물(박테리아, 고세균, 바이러스 등)의 전체 공동체를 의미합니다.
• 산호 프로바이오틱 가설(Coral Probiotic Hypothesis): 산호와 공생 미생물 사이에 동적인 관계가 존재하며, 환경 조건(특히 온도 변화)에 따라 가장 적합한 미생물 군집을 선택한다는 가설입니다. 이를 통해 산호가 병원체에 대한 저항성과 환경 적응력을 얻는다고 설명합니다.
• 홀로게놈 이론(Hologenome Theory of Evolution): 위 가설을 일반화한 것으로, 미생물이 동물과 식물의 진화 과정에서 매우 중요한 역할을 한다는 진화 이론입니다.

산호와 미생물의 관계 산호의 점액층(mucus), 골격, 조직에는 매우 다양한 미생물이 서식하며 다음과 같은 역할을 합니다:

• 광합성 지원 (zooxanthellae와 함께)
• 질소 고정
• 영양소 공급 및 소화
• 병원체 침입 방어 (항균 물질 생산 등)

질병과 백화 현상

• 산호 백화(Coral Bleaching): 고온 등 스트레스로 공생 조산자(zooxanthellae)가 빠져나가 산호가 하얗게 되는 현상.
• 대표 병원체: Vibrio shiloi (Oculina patagonica 산호 백화의 원인균). 고온에서 병독성이 강해지며, 초과산화물 디스뮤타아제(superoxide dismutase) 같은 인자가 관여합니다.
• 블랙 밴드 질병(Black band disease) 등 다른 질병에서도 특정 박테리아 군집이 중요한 역할을 합니다.

중요한 발견 2000년대 초 연구에서, 지중해 산호 Oculina patagonica가 처음에는 Vibrio shiloi에 감염되어 백화되었으나, 몇 년 후 자연적으로 저항성을 획득한 사례가 관찰되었습니다. 이는 미생물 군집의 변화로 설명되며, 산호 프로바이오틱 가설의 핵심 증거가 되었습니다.
결론과 함의 기후 변화, 수질 오염, 과도한 어업 등으로 산호 질병이 급증하고 있습니다. 저자들은 산호 미생물학 연구가 산호초 보존에 필수적이라고 강조하며, 다음과 같은 점을 제안합니다:

• 동적인 공생 관계를 활용한 프로바이오틱 접근 (유익한 미생물 도입)
• 미생물이 진화의 주요 동력이라는 새로운 관점 (홀로게놈 이론)

이 논문은 산호를 단순한 동물이 아니라 미생물과 함께 살아가는 복합 생물(holobiont)로 바라보는 관점을 제시한 중요한 리뷰로, 이후 산호 미생물학 연구의 기초가 되었습니다.

3. Zilber-Rosenberg, I. & Rosenberg, E. (2008). Role of microorganisms in the evolution of animals and plants: the hologenome theory of evolution. FEMS Microbiology Reviews, 32(5), 723–735.

홀로게놈 이론(hologenome theory of evolution)을 정식 제안.

홀로비온트를

선택의 단위(level of selection)로 보고,

미생물이 호스트 진화에 기여하는 메커니즘(수직 전파, 미생물 다양성 변화 등)을 설명.

개념을 진화생물학의 새로운 패러다임으로 끌어올린 핵심 논문입니다.

이 논문은
홀로게놈 이론(Hologenome Theory of Evolution)을
본격적으로 제시한 대표 리뷰입니다.

이전
2007년 Nature Reviews Microbiology 논문(산호 중심)에서 제안된 아이디어를
동물과 식물 전반으로 확장한 버전입니다.

홀로게놈 이론의 기본 개념

• 홀로바이언트(Holobiont): 숙주(동물 또는 식물) + 그와 공생하는 모든 미생물(박테리아, 고세균, 바이러스, 진균 등)의 총체.
• 홀로게놈(Hologenome): 숙주의 게놈 + 미생물 군집의 전체 유전 정보의 합.

이 이론은 진화의 단위를 개별 생물이 아닌 홀로바이언트 전체로 본다는 점이 핵심입니다. 미생물이 숙주의 적응·발달·진화에 매우 중요한 역할을 한다는 관점입니다.

이론의 4가지 일반화 (Four Generalizations)

1. 모든 동물과 식물은 미생물과 공생 관계를 맺는다.
2. 공생 미생물은 숙주에게 유익한(또는 때로는 해로운) 영향을 미치며, 숙주의 적합도(fitness)에 기여한다.
3. 미생물 군집은 환경 변화에 따라 빠르게 변할 수 있다 (미생물의 빠른 증식과 수평 유전자 이동 덕분).
4. 미생물의 획득·전달 방식(수직 전달, 수평 전달 등)이 다양하며, 이는 홀로바이언트의 진화에 유연성을 제공한다.

주요 주장

• 미생물은 단순한 ‘공생자’가 아니라, 숙주의 진화 과정의 주요 동력이다.
• 미생물 군집의 변화(획득, 소실, 구성 변화)를 통해 홀로바이언트가 환경에 더 빠르게 적응할 수 있다. 이는 다윈주의적 진화(돌연변이 + 자연선택)에 라마르크적 요소(획득 형질의 유전)를 일부 보완하는 형태로 볼 수 있다.
• 산호 백화 연구에서 출발했으나, 곤충, 척추동물, 식물 등 다양한 생물군으로 확대 적용

4. Gilbert, S. F. et al. (2012). A Symbiotic View of Life: We Have Never Been Individuals. The Quarterly Review of Biology, 87(4), 325–341. (인용 1,600회 이상)

“우리는 결코 개별 개체가 아니었다”는 철학적·생물학적 재해석.

Margulis의 공생 관점을

현대 미생물학으로 확장해,

모든 다세포 생물이 미생물과 함께 ‘공생적 개체’임을 주장.

개념의 철학적·교육적 영향력이 매우 컸습니다.

“우리는 결코 개별(individual) 존재가 아니었다.”

동물(그리고 식물)이라는 존재는
오랜 기간 미생물과 함께 살아온 복합체(composite)이며,
전통적인 ‘생물학적 개체(individual)’ 개념이 더 이상 유효하지 않다는 강력한 주장을 펼친 리뷰 논문입니다.

주요 주장

최신 유전체 분석(고속 시퀀싱, RNA 기술 등)을 바탕으로
생물학의 여러 분야(해부학, 생리학, 발생학, 면역학, 유전학, 진화학)에서
‘개체’ 개념이 어떻게 도전받고 있는지를 종합적으로 검토합니다.

1. 해부학·생리학적 관점 동물은 미생물 공생체 없이 정상적인 대사 경로를 완성할 수 없다. 다양한 공생 미생물이 영양 공급, 소화, 대사 등 핵심 생리 기능을 담당한다.
2. 발생학(Development) 관점 동물의 발생(발달 과정)은 공생 미생물(symbionts)이 없으면 불완전하다. 미생물이 발생 신호, 장기 형성 등에 필수적으로 관여한다.
3. 유전학적 관점 미생물은 두 번째 유전적 유전 방식(second mode of genetic inheritance)을 제공한다. 이는 숙주 게놈과 별도로, 자연선택의 대상이 되는 유전적 변이를 공급한다.
4. 면역학 관점 면역 체계는 미생물과 ‘대화’하며 발달한다. 미생물을 단순히 ‘적’이 아닌, 몸의 일부로 통합하는 역할을 한다.

핵심 개념: 홀로바이언트(Holobiont)

• 다세포 진핵생물(숙주) + 그 안에 지속적으로 존재하는 공생 미생물 군집을 하나의 단위로 본다.
• 이 홀로바이언트가 해부학, 발생, 생리, 면역, 진화의 진정한 단위가 되어야 한다는 주장.

5. McFall-Ngai, M. et al. (2013). Animals in a bacterial world, a new imperative for the life sciences. Proceedings of the National Academy of Sciences (PNAS), 110(9), 3229–3236.

PNAS 고영향력 리뷰.

동물이 “박테리아 세계”에 살고 있으며,

미생물이 동물 발달·면역·진화에 필수적임을 종합.

“새로운 생명과학의 명령”으로서

홀로비온트 관점을 주류 생물학으로 끌어올린 획기적 논문입니다.

“동물은 세균(박테리아)의 세계 속에서 진화하고 살아간다.
생명과학 전반에 새로운 패러다임이 필요하다.”

이 논문은
미생물(특히 박테리아)과의 상호작용을 고려하지 않은
기존 동물 생물학이 더 이상 충분하지 않다는 강력한 선언입니다.

유전체·메타게놈 기술의 발달로 밝혀진
방대한 동물-박테리아 상호작용을 바탕으로,
생명과학의 기본 전제를 재고해야 한다고 주장합니다.

주요 내용 (5가지 핵심 질문 중심)

논문은 다음과 같은 5가지 근본 질문에 대해 최근 연구 성과를 종합적으로 검토합니다:

1. 박테리아가 동물의 기원과 진화에 어떻게 기여했는가? 동물의 진화 과정에서 박테리아가 핵심적인 역할을 했으며, 홀로바이언트 관점이 진화 이해에 필수적이다.
2. 동물과 박테리아가 서로의 게놈에 어떤 영향을 미치는가? 공생을 통해 유전자 교환, 게놈 안정성, 적응력이 영향을 받는다.
3. 정상적인 동물 발생(발달)은 박테리아 파트너 없이 가능한가? 많은 동물에서 발생 과정(장기 형성, 면역 발달 등)이 미생물 신호에 의존한다. 미생물이 없으면 불완전하거나 비정상적이다.
4. 동물과 공생 미생물 사이의 항상성(균형)은 어떻게 유지되는가? 면역 체계, 대사 조절 등을 통해 숙주-미생물 균형이 유지되며, 이 과정이 동물 생리의 핵심이다.
5. 생태학적 접근이 동물-박테리아 상호작용 이해에 어떻게 도움이 되는가? 환경, 군집 수준에서의 다층적 상호작용을 고려해야 한다.

이전 논문과의 연계 (시리즈 흐름)

• 2007 Rosenberg et al. (산호 중심)
• 2008 Zilber-Rosenberg & Rosenberg (홀로게놈 이론)
• 2012 Gilbert et al. (“우리는 결코 개별 존재가 아니었다”)
• 2013 McFall-Ngai et al. → 위 논문들을 바탕으로 더 광범위하고 선언적인 종합 리뷰로 마무리. 생명과학 전체에 “새로운 명령(new imperative)”을 제시.

의의 이 논문은 홀로바이언트/마이크로바이옴 패러다임을 생명과학의 주류로 끌어올린 대표적 작품입니다. “동물 생물학 = 미생물 없이 설명 불가능”이라는 관점을 명확히 하여, 이후 수많은 연구와 교육 방향에 큰 영향을 미쳤습니다.

6. Bordenstein, S. R. & Theis, K. R. (2015). Host Biology in Light of the Microbiome: Ten Principles of Holobionts and Hologenomes. PLOS Biology, 13(8), e1002226.

홀로비온트·홀로게놈의 10가지 원칙을

체계적으로 제시한 가장 영향력 있는 논문 중 하나(인용 1,500회 이상).

PLOS Biology(고영향력)에서 개념을 명확히 정의하고,

오해를 바로잡아 현대 연구의 기초를 마련했습니다.

 이 논문은
홀로바이언트(Holobiont)와 홀로게놈(Hologenome) 개념을
더 명확하고 체계적으로 정리한 중요한 리뷰/에세이입니다.

이전 논문들
(2007 Rosenberg 산호, 2008 Zilber-Rosenberg 홀로게놈 이론,
2012 Gilbert “우리는 결코 개별이 아니었다”,
2013 McFall-Ngai “동물은 박테리아의 세계 속에 있다”)에서 제기된 아이디어를 종합·정제하면서,
생명과학에서 이 개념들이 무엇이고 무엇이 아닌지를 10가지 원칙으로 명확히 제시합니다.

목적은
홀로바이언트/홀로게놈 이론이 기존 진화·생태 이론을 지지하면서도 확장한다는 점을 보여주고,
향후 연구의 로드맵을 제공하는 것입니다.

주요 내용 논문은
10가지 원칙(Ten Principles)을 중심으로 구성되어 있습니다.

이 원칙들은
홀로바이언트와 홀로게놈의 정의, 범위, 한계, 그리고 생물학적 함의를 체계적으로 설명합니다.

대표적인 원칙의 핵심 아이디어 (간략히):

• 홀로바이언트는 숙주 + 모든 공생 미생물의 생태적 단위이며, 진화의 중요한 단위가 될 수 있다.
• 홀로게놈은 숙주 게놈 + 미생물 게놈의 총합으로, 미생물의 빠른 진화와 수평 유전자 이동이 숙주의 적응력을 높여준다.
• 미생물 군집은 환경 변화에 빠르게 반응하며, 숙주의 발생·생리·면역·행동에 필수적이다.
• 이 개념은 다윈의 자연선택 이론을 보완(라마르크적 요소 포함)하지만, 기존 이론과 충돌하지 않는다.
• 개념의 오용(예: 모든 미생물이 항상 유익하다고 보는 과도한 일반화)을 경계한다.

저자들은 이 10가지 원칙이 통합 이론(unified theory)으로 기능할 수 있도록, 개념의 명확한 경계를 제시하고 실험적 검증의 방향을 제안합니다.

7. Postler, T. S. & Ghosh, S. (2017). Understanding the Holobiont: How Microbial Metabolites Affect Human Health and Disease. Cell Metabolism, 25(2), 234–246.

Cell Metabolism(고영향력)에서 인간 홀로비온트에 초점. 미생물 대사산물이 숙주 대사·면역에 미치는 구체적 메커니즘을 설명하며, 의학적 적용(비만, 염증 등)을 강조. 개념을 임상·의학 분야로 확장한 논문입니다.

https://pmc.ncbi.nlm.nih.gov/articles/PMC5535818/

Understanding the Holobiont: How microbial metabolites affect human health and shape the immune system - PMC

 인간과 장내 미생물(microbiome)은
단순한 공생 관계가 아니라 하나의 복합 생물(holobiont)을 이룬다.

미생물이 생산하는 대사산물(metabolites)이
숙주(인간)의 대사, 면역 체계, 그리고 전반적인 건강에 결정적인 영향을 미친다는 점을
중점적으로 다룬 리뷰 논문입니다.

미생물 군집의 교란(dysbiosis)이
비만, 대사 질환, 염증성 장질환, 암 등 다양한 질병과 연관된다는 최근 연구 성과를 종합합니다.

주요 내용

1. 홀로바이언트로서의 인간
   • 평균 성인 남성 몸에는 약 3.8 × 10¹³개의 미생물이 살고 있으며, 이는 인간 세포 수(약 3.0~3.7 × 10¹³)와 비슷하거나 많다.
   • 미생물 유전자 수는 인간 단백질 코딩 유전자보다 약 150배 많다.
   • 장내 미생물은 주로 Firmicutes와 Bacteroidetes 두 문(phylum)이 지배하며, 숙주에게 비타민 B·K 합성, 복잡한 탄수화물 분해 등 필수 대사 기능을 제공한다.
   • 혈액 내 대사산물의 약 10%가 장내 미생물에서 유래한다.
2. 미생물 대사산물의 역할
   • 단쇄지방산(SCFAs): 아세테이트, 프로피오네이트, 부티레이트 등. 에너지 공급, 장벽 강화, 면역 조절(예: Treg 세포 유도, HDAC 억제), 항염증 효과.
   • 담즙산(bile acids), 트립토판 대사물, 폴리아민 등 다양한 대사산물이 호스트의 대사·면역 신호 전달에 관여.
   • 이러한 대사산물은 GPCR 수용체(GPR41, GPR43 등)나 히스톤 탈아세틸화효소(HDAC) 억제를 통해 숙주 세포의 유전자 발현, 염증 반응, 대사 항상성을 조절한다.
     
     
3. 미생물 교란(dysbiosis)과 질병
   • 건강한 미생물 군집에서는 상호이익(mutualistic) 관계이지만, dysbiosis 시에는 비만, 동맥경화, 대사증후군, 염증성 장질환(IBD), 암 등에서 해로운 방향으로 작용할 수 있다.
   • 반대로 특정 “무해한” 미생물의 부재가 면역 세포 발달을 저해하기도 한다.
   • 미생물 대사산물이 대사 질환과 면역 질환의 중요한 매개체임을 강조.

 
이전 논문 시리즈와의 연계

• 2007~2015년 홀로바이언트/홀로게놈 시리즈(Rosenberg, Gilbert, McFall-Ngai, Bordenstein 등)가 개념적·이론적 기반을 마련했다면,
• 2017 Postler & Ghosh 논문은 이를 인간 건강과 질병, 특히 대사 및 면역 분야로 구체화·실증한 실용적 리뷰입니다.
• 미생물이 단순히 “공생자”가 아니라, 대사산물을 통해 숙주의 생리를 직접 조절하는 핵심 파트너라는 점을 명확히 보여줍니다.

이 논문은
미생물 대사산물이 호스트 생리학의 핵심 조절자라는
현대 마이크로바이옴 연구의 중요한 전환점을 제시했습니다.

이후
프로바이오틱스,
프리바이오틱스,
대사산물 기반 치료 전략(예: 부티레이트 보충 등)의 과학적 근거를 강화하는 데 크게 기여

8. Morris, J. J. (2019). Do hosts and their microbes evolve as a unit? PNAS, 116(29), 14485–14487. (또는 관련 PNAS perspective)

PNAS에서 홀로비온트가 진화 단위인지 논쟁을 정리. 미생물-호스트 공동 진화(co-evolution)와 선택 압력을 분석하며, 개념의 철학적·이론적 논의를 심화했습니다.

이 글은
과학 뉴스 피처(News Feature)로,
홀로바이언트(hologenome) 이론을 둘러싼 진화생물학자들 사이의 격렬한 논쟁을 균형 있게 소개합니다.

“숙주(호스트)와 그 미생물이 하나의 단위(unit)로 진화하는가?”라는 질문을 중심으로,
지지파와 반대파의 주요 주장과 증거를 정리한 리뷰 형식의 기사입니다.

주요 내용

• 홀로게놈 이론 지지 측 (예: Eugene Rosenberg 등)
  • 숙주 + 미생물 군집 = 홀로바이언트(holobiont)
  • 이들의 전체 유전 정보 = 홀로게놈(hologenome)
  • 미생물 군집은 숙주 게놈보다 훨씬 빠르게 변화할 수 있어 (빠른 증식, 수평 유전자 이동, 미생물 획득/소실), 환경 변화에 홀로바이언트가 신속하게 적응할 수 있다고 주장.
  • 산호, 곤충, 일부 척추동물 등에서 미생물이 숙주의 적응과 진화에 기여한 사례를 증거로 제시.
  • “숙주 게놈은 천천히 진화하지만, 미생물은 빠르게 변한다”는 점이 홀로바이언트의 적응 우위를 만든다고 강조.
• 반대 측 (비판자들)
  • 미생물 군집(특히 장내 미생물처럼 환경에 크게 의존하는 경우)은 세대 간 안정적으로 전달되지 않는다.
  • 따라서 숙주와 미생물이 진화의 단일 단위(unit of selection)로 작용하기 어렵다.
  • 미생물은 숙주의 “환경”이나 “확장된 표현형”으로 보는 것이 더 적절하며, 기존 다윈 진화 이론으로 충분히 설명 가능하다고 주장.
  • 일부는 홀로게놈 이론이 과도한 일반화이며, 실험적 증거가 부족하다고 지적.
• 현재 상황: 논쟁은 여전히 뜨겁게 진행 중이며, 양측 모두 일부 사례에서는 홀로바이언트 관점이 유용하지만, 모든 생물에 보편적으로 적용하기는 어렵다는 데 어느 정도 공감대가 형성되고 있음.

이전 논문 시리즈와의 연계

• 2007 Rosenberg (산호) → 2008 Zilber-Rosenberg (홀로게놈 이론) → 2012 Gilbert → 2013 McFall-Ngai → 2015 Bordenstein (10가지 원칙) → 2017 Postler & Ghosh (인간 건강 적용)까지 이어진 홀로바이언트/홀로게놈 시리즈의 개념을 2019년 시점에서 객관적으로 조명한 메타적 기사입니다.
• 이론이 제안된 지 10여 년이 지나면서 생겨난 과학계 내 논쟁의 현황을 잘 보여주는 자료

9. Bordenstein, S. R. et al. (2024). The disciplinary matrix of holobiont biology. Science.

Science(최고 영향력) perspective. 미생물이 보이는 생명(호스트)과 보이지 않는 생명(미생물)을 통합한 ‘홀로비온트 생물학’의 학문적 틀을 제시. 농업·보존·의학 등 실용적 응용까지 연결하며, 개념을 21세기 생물학의 핵심 프레임워크로 승격시켰습니다.

미생물(microbiome) 연구의 급속한 발전으로
미시생물학과 거시생물학(micro- and macrobiology)이 수렴하고 있다.

저자는
홀로바이언트 생물학(holobiont biology)을
하나의 다학제적 프레임워크(multidisciplinary framework)로 제시하며,
이를 “disciplinary matrix” (학문적 매트릭스)로 개념화합니다.

이 매트릭스는
보이는 세계(숙주: 동물, 식물 등 거시생물)와
보이지 않는 세계(미생물: 바이러스, 박테리아, 고세균, 진균 등)를 통합하여
생물학 연구의 개념적 진보를 이끈다고 강조합니다.

주요 개념

• 홀로바이언트(Holobiont): 숙주(host)와 그에 공생하는 모든 미생물 세포들의 총체.
• 홀로게놈(Hologenome): 홀로바이언트 내 모든 다종(genomic material)의 유전 정보 총합.
• 이 개념들은 미생물 공생(symbiosis)을 거시생물의 구조(structure), 기능(function), 진화(evolution)에 본질적으로 통합합니다.
• 숙주는 더 이상 “단일 개체”가 아니라, 미생물을 포함한 동적 집합체(dynamic assemblage)로 재정의됩니다.

Disciplinary Matrix의 의미

• Seen realm (보이는 세계): 전통적인 동물학, 식물학, 해부학, 생리학 등 거시생물 중심 연구.
• Unseen realm (보이지 않는 세계): 미생물학, 메타게놈학, 바이러스학 등 미시생물 중심 연구.
• 이 두 영역을 하나의 매트릭스로 연결함으로써, 농업 생산성 향상, 생물 다양성 보전, 인간 질병 치료(마이크로바이옴 기반 솔루션) 등 광범위한 응용 분야에 기여할 수 있습니다.

이 매트릭스는
생물학의 기존 패러다임을 확장하며,
미생물이 생물권(biosphere)에서 중심적인 역할을 한다는 점을 강조합니다.

Darwin이 말한
“endless forms most beautiful”에
미생물 공생이라는 새로운 차원을 추가하는 개념적 진보로 평가됩니다.

이전 시리즈와의 연계 Seth R. Bordenstein은
2015년 “Host Biology in Light of the Microbiome: Ten Principles of Holobionts and Hologenomes”
논문의 저자이기도 합니다.

이번 2024년 Science Perspective는
2007년 Rosenberg의 산호 연구부터 시작된
홀로바이언트/홀로게놈 시리즈의 최신 종합·성숙 단계로 볼 수 있습니다.

이전 논쟁(2019 PNAS News Feature에서 제기된
“숙주와 미생물이 하나의 단위로 진화하는가?”)을 넘어,
홀로바이언트 생물학을 통합적 학문 매트릭스로 정립하려는 시도

10. Wang, W. et al. (2024). Decoupling of strain- and intrastrain-level interactions of microbiomes in a sponge holobiont. Nature Communications, 15, 8205. (사용자님이 제공한 논문)

Nature Communications(고영향력) 최근 실증 연구. 스폰지(해면동물) 홀로비온트에서 미생물 strain 수준 상호작용이 호스트 유전적 다양화에 미치는 역할을 밝힘. 초기 이론을 실험·유전학적으로 검증하며, 개념의 현대적 적용(미생물 다양성·진화 메커니즘)을 보여주는 대표적 사례입니다.

논문 상세 설명 (2024년 Nature Communications)

제목: Decoupling of strain- and intrastrain-level interactions of microbiomes in a sponge holobiont 저자: Wenxiu Wang (You Wang), Bu Xu, Lu Fan 등 (주저자: Lu Fan 연구팀, 호주·중국 공동 연구) 저널: Nature Communications (Volume 15, Article number: 8205) 발행일: 2024년 9월 18일 (Open Access, 무료 전체 공개) DOI: 10.1038/s41467-024-52464-6 링크: https://www.nature.com/articles/s41467-024-52464-6

1. 연구 배경과 핵심 질문

이 논문은

홀로바이언트(holobiont)의 유전적 다양성이 어떻게 형성되는지를

균주(strain) 수준과 균주 내(intrastrain) 수준으로 나누어 분석합니다.

이전 홀로게놈 이론(2008 Zilber-Rosenberg 등)에서는

미생물이 숙주의 적응과 진화에 중요하다고 했지만,

개별 홀로바이언트 사이의 유전적 유사성/차이를 결정하는 정확한 메커니즘은 아직 불분명했습니다.

연구팀은 모델 해绵(스폰지) Cymbastela concentrica (C. concentrica)을 이용해:

• 숙주(스폰지) 유전자가 미생물 군집에 미치는 영향
• 바이러스(주로 박테리아·고세균을 감염시키는 phage)와의 상호작용
• strain-level(균주 수준) vs. intrastrain-level(균주 내 수준) 상호작용의 decoupling(분리/해리) 현상

을 밝혀냈습니다.

결론적으로,

“strain-level과 intrastrain-level 상호작용의 decoupling이

홀로바이언트의 유전적 다양화(genetic diversification)의 핵심 요인”이라고 주장합니다.

2. 연구 방법 (간략)

• 샘플: 호주 시드니 Bare Island 인근 작은 지역(~2500 m², 약 200개체)에서 7~31개체의 건강한 스폰지 채취.
• 시퀀싱: 단편·장편 DNA/RNA 시퀀싱 + Hi-C (DNA proximity linkage) 기술로 숙주-미생물-바이러스 연결 고해상도 분석.
• 분석: MAGs (Metagenome-Assembled Genomes) 37개 (박테리아·고세균), vOTUs (viral operational taxonomic units) 15,400개. nucleotide diversity(π), popANI, CRISPR, provirus, defense systems(PADLOC) 등 다양한 bioinformatic pipeline 사용.

3. 주요 발견 (핵심 결과)

1. Strain-level (균주 수준)
   • 스폰지 숙주 유전자가 미생물 군집과 바이러스 군집의 조성을 강하게 결정 (phylosymbiosis-like 패턴 확인).
   • 바이러스는 고도의 strain-level 특이성 (94.9%가 하나의 균주만 감염).
   • 외부 바이러스 침입과 종간 재조합(interspecific recombination)이 강력히 차단됨.
2. Intrastrain-level (균주 내 수준)
   • 개체 간 nucleotide diversity가 매우 높은 중복(overlap) → 지역 개체군 내에서 미생물 세포와 바이러스가 빈번하게 교환·전달됨.
   • Immune-evasive arms race (면역 회피 무기 경쟁): 바이러스와 숙주 미생물 사이에 지속적인 돌연변이·재조합이 일어나 홀로바이언트 특이적 변이를 생성.
3. Decoupling (분리 현상)
   • Strain-level: 숙주가 강하게 통제 (안정적, 특이적).
   • Intrastrain-level: 숙주 영향이 약하고, 지역 개체군 내 공유·교환이 활발 (유연하고 동적).
   • 이 decoupling이 홀로바이언트의 빠른 유전적 다양화와 적응력을 제공하는 핵심 메커니즘.

4. 그림 설명 ( 논문 Figure 6)

이 그림은 논문의 가장 중요한 개념적 모델(conceptual model)로, 위 모든 발견을 한 장에 시각화한 것입니다.

• 상단 두 개의 원형 다이어그램 (서로 다른 두 미생물 군집)
  • 왼쪽/오른쪽 모두 “strong immune protection” (강력한 면역 보호) 표시.
  • Environmental viral attack (환경 바이러스 공격) → 빨간 X로 차단.
  • Interspecific recombination (종간 재조합) → 빨간 X로 차단. → 스폰지의 면역 체계가 외부 침입과 다른 종 간 유전자 교환을 막는다는 의미.
• 오른쪽 상단 파란 상자 (Virus-host interactions)
  • high strain-level specificity (고도의 균주 수준 특이성)
  • immune & evasive balance (면역과 회피의 균형)
  • high intracellular copies (세포 내 높은 복제 수)
  • intrastrain-level co-evolution (arms race & recombination) (균주 내 수준 공동진화: 무기 경쟁 + 재조합)
• 하단 두 개의 스폰지 모양
  • 스폰지 유전자가 strain-level microbiome을 형성 (“sponge genetics shapes strain-level microbiome”).
  • Intrastrain-level microbial transmission (균주 내 수준 미생물 전달): 양방향 화살표로 개체군 내 빈번한 교환 표시.
• 범례
  • 스폰지 C. concentrica
  • bacterial/archaeal symbionts (보라·주황 막대)
  • viruses of bacterial/archaea (파랑·빨강 phage)

한 줄 요약:

스폰지는 외부는 철저히 차단하지만,

내부(균주 내)에서는 미생물·바이러스 간 활발한 교환과 공동진화가 일어나

홀로바이언트의 다양성을 유지·증가시킨다.

5. 결론과 의의

• 홀로바이언트의 유전적 다양성은 단순한 숙주-미생물 공생이 아니라, strain vs. intrastrain decoupling이라는 이중 구조로 설명된다.
• 이는 이전 홀로게놈 이론(2008~2015 시리즈)을 한 단계 더 구체화·실증한 결과로, 스폰지 같은 고대 공생계가 어떻게 장기적으로 안정성과 유연성을 동시에 유지하는지를 보여줍니다.
• 함의: 농업·의학·생태 보전에서 마이크로바이옴 조작 시 균주 수준과 균주 내 수준을 구분해서 접근해야 한다는 실마리 제공.

이 논문은

2007년 Rosenberg 산호 논문부터 시작된

홀로바이언트 시리즈의 최신·가장 구체적인 실증 연구 중 하나

이 10개 논문은

홀로비온트 개념이

이론적 제안(1990) → 산호·미생물 적용(2000s) → 진화 이론화(2008) → 주류 생물학 통합(2010s) → 현대 실증·응용(2020s)으로

발전한 흐름을 잘 보여줍니다.

초기 논문은 BioScience·FEMS 등이었으나,

2007년 이후 Nature Reviews, PNAS, Science, Nature Communications 등 top-tier 저널에서

지속적으로 다뤄지고 있습니다.

1. nature  
2. nature communications  
3. articles  
4. article

Decoupling of strain- and intrastrain-level interactions of microbiomes in a sponge holobiont

Download PDF

• Article
• Open access
• Published: 18 September 2024

Decoupling of strain- and intrastrain-level interactions of microbiomes in a sponge holobiont

• Wenxiu Wang, 
• Weizhi Song, 
• Marwan E. Majzoub, 
• Xiaoyuan Feng, 
• Bu Xu, 
• Jianchang Tao, 
• Yuanqing Zhu, 
• Zhiyong Li, 
• Pei-Yuan Qian, 
• Nicole S. Webster, 
• Torsten Thomas & 
• Lu Fan 

Nature Communications volume 15, Article number: 8205 (2024) Cite this article

•  
•  
•  
•  

Abstract

Holobionts are highly organized assemblages of eukaryotic hosts, cellular microbial symbionts, and viruses, whose interactions and evolution involve complex biological processes. It is largely unknown which specific determinants drive similarity or individuality in genetic diversity between holobionts. Here, we combine short- and long-read sequencing and DNA-proximity-linkage technologies to investigate intraspecific diversity of the microbiomes, including host-resolved viruses, in individuals of a model marine sponge. We find strong impacts of the sponge host and the cellular hosts of viruses on strain-level organization of the holobiont, whereas substantial overlap in nucleotide diversity between holobionts suggests frequent exchanges of microbial cells and viruses at intrastrain level in the local sponge population. Immune-evasive arms races likely restricted virus-host co-evolution at the intrastrain level, generated holobiont-specific genome variations, and linked virus-host genetics through recombination. Our work shows that a decoupling of strain- and intrastrain-level interactions is a key factor in the genetic diversification of holobionts.

Similar content being viewed by others

Disentangling direct vs indirect effects of microbiome manipulations in a habitat-forming marine holobiont

Article Open access29 March 2024

The coral microbiome in sickness, in health and in a changing world

Article 04 March 2024

Persistence of functional microbiota composition across generations

Article Open access24 September 2021

Introduction

Holobionts are ecological and evolutionary entities consisting of host animals or plants and their associated communities of symbiotic microbial eukaryotes, bacteria, archaea, and viruses (collectively the microbiome)1,2. Information on the biological processes that structure microbiomes and generate genetic variability within and between holobiont individuals is essential to understanding holobiont health and evolution. In addition to processes such as ecological drift and intra-host diversification, animal hosts can structure their microbiome composition by selective processes (e.g., immune interactions, cell-cell recognition, and metabolic mutualism) or by dispersal restriction (e.g., strict vertical transmission of microbial symbionts and geographic isolation)3,4. Eval‎uating the relative contribution of these host-driven processes, along with the combination of non-host processes, such as horizontal symbiont transmission and genome recombination, requires careful methodological and experimental design5.

Sponges (phylum Porifera) are early-diverging sessile animals6 that can reproduce both sexually and asexually and play increasingly important roles in reef ecosystems, where corals are declining due to climate change or anthropogenic impacts7. Studies have revealed that complex communities of sponge microbiomes are often highly specific to a particular sponge species8,9,10,11, thereby implying a degree of co-evolution9,12. the existence and extent of horizontal symbiont transmission in sponge holobionts are still being actively debated. For example, previous studies reported that strict vertical transmission of microbial symbionts between sponge generations was unlikely to exist13 and that sponges shared a large component of their microbial symbionts with sediments and other reef invertebrates by horizontal transmission14. In contrast, another study identified cophylogeny between sponge hosts and microbial symbionts over their long evolutionary history, suggesting vertical transmission15. All these studies were based on sequencing of the partial 16S rRNA genes, resolving at best species-level taxonomy1. As microbial cells of the same species can vary substantially in metabolism, virulence, cell surface characteristics, and consequently their interactions with holobiont hosts16, sponge individuals may selectively host distinct strains of the same microbial species. Therefore, to resolve the ongoing debates in microbiome transmission modes in sponge holobionts, it will be critical to analyze sponge-associated microorganisms with intraspecific taxonomic resolution, such as at strain or intrastrain levels.

Like cellular microbiomes, the taxonomic composition of viromes is also generally specific to a sponge species and distinct from the ambient free-living viral communities11,17,18. The hosts of viruses (i.e., the animal/plant host plus the cellular microbial symbionts) should be a dominant factor in influencing the virome structure of a holobiont. However, due to the technical difficulties in linking viruses to their cellular hosts at the community level19, the mechanisms of virome assembly in holobionts are still largely unknown. Moreover, technological limitations in taxonomic resolution have obscured an understanding of the mode of viral transmission. For example, the extent of maternal transmission in human infant viromes was reported at various levels in studies based on viral assemblies of short sequencing reads, which provides only consensus sequences of viral populations20,21. Similarly to cellular microbiome studies, the availability of analytical tools with nucleotide resolution to study at strain or intrastrain levels may provide more robust approaches to understanding viral transmission between holobiont individuals22.

The collection of genomes from all holobiont members provides a rich pool of genetic diversity, which underpins the potential of a holobiont to adapt to environmental changes23,24. Although many mechanisms are involved, virus-host interactions are expected to be an important factor in driving intraspecific diversity of both the viruses and their cellular hosts within and between holobiont individuals24. In principle, viral activities may promote intraspecific genetic variability by selectively killing microbial strains, facilitating mutations in the virus-host arms race, and transferring, disrupting, or rearranging genes in and between microbial hosts25. Pioneering studies have demonstrated that bacteria26 and viruses26,27 in human gut microbiomes are in rapid evolution. However, it is still uncertain how much of their evolutions, especially in the generation of intraspecific diversity, are contributed by virus-host interactions.

In this study, we apply Oxford nanopore sequencing, high-throughput chromosome conformation capture (HiC) metagenomics (a DNA proximity linkage technology), and single-nucleotide variation analysis to investigate the strain- and intrastrain-level organization, interactions, and co-evolution of cellular microbial symbionts and their viruses in a subset of a local population of Cymbastela concentrica, a widely distributed and ecologically important temperate sponge in the Indo-Pacific ocean with a relatively stable microbial community28,29,30. These technologies have recently been successfully applied to recover high-quality viral genomes31, to identify virus-host interactions previously missed by clustered regularly interspaced short palindromic repeats (CRISPR)-based methods32, and to reveal bacterial and viral transmission between holobiont individuals22, respectively. We elucidate the intraspecific genetic diversity of holobiont members in individuals of C. concentrica and then test whether the genetics of the sponge host impact the microbiome structure at the strain and intrastrain levels. Host specificity and intracellular activity of sponge-associated viruses are then eval‎uated by revealing their infectious relationships with their cellular hosts. Finally, we investigate the level of microbial transmission and the mechanisms of virus-host interactions in generating intrastrain genetic variations within and between sponge individuals.

Results

Rich intraspecific diversity in the C. concentrica holobiont

The definitions of taxonomic ranks below species of bacteria, archaea, and viruses are still under ongoing discussion and are largely study-dependent16,22,33. For a practical purpose, we here define “strain-level diversity” as the genetic difference between bacterial, archaeal, or viral genomes sharing average nucleotide identity (ANI) higher than 95% but lower than 99%. The lower limit of 95% is taken as commonly suggested by the scientific community, including in a recently published review16, whereas the upper limit of 99% is generally recommended for distinguishing metagenome-assembled genomes (MAGs) when using Illumina short reads34. Accordingly, we define “intrastrain-level diversity” as single nucleotide polymorphism between cells or viral variants within the above defined “strain”.

From a subset of a local population of C. concentrica at Bare Island, Sydney, Australia, DNA samples of C. concentrica (CC) individuals CC1-CC7 were sequenced on the Illumina, Roche 454, and Oxford nanopore platforms, and RNA samples of individuals CC8-CC31 were sequenced on the Illumina platforms in this and our previous studies35,36,37 (Supplementary Data 1). The sequencing reads were used to reconstruct the genomes of the sponge host, the cellular microbial symbionts, and the viruses (Supplementary Fig. 1). The analysis of strain-level community composition and intrastrain-level nucleotide diversity was performed with Illumina sequencing reads of CC1-CC4 and CC8-CC28 (Supplementary Data 1).

In total, 37 bacterial and two archaeal non-redundant (pairwise ANI < 99%) MAGs (Fig. 1a and Supplementary Data 2) were generated through a customized pipeline combining short- and long-read sequences (Supplementary Fig. 1a). The average completeness and contamination of these MAGs was 80.70% ( ± 12.56%) and 0.94% ( ± 0.84%), respectively, as eval‎uated by CheckM38, and 80.63% ( ± 15.09%) and 2.31% ( ± 1.77%), respectively, as eval‎uated by CheckM239. We note that the completeness of some MAGs could have been underestimated. For example, one complete circular MAG (M411805 Nitrosopumilus) was estimated to be of only 85.04% and 65.18% completeness by CheckM and CheckM2, respectively (Supplementary Data 2).

Fig. 1: Strain-level diversity and intrastrain-level genetic variations in individuals of the sponge C. concentrica.

The alternative text for this image may have been generated using AI.

Full size image

a Phylogeny and abundance of archaeal and bacterial MAGs. Gray boxes show zero values. The maximum likelihood tree was constructed based on 40 single-copy marker proteins. The archaeal tree is rooted when the bacteria are treated as the outgroup, and the bacterial tree is rooted when the archaea are treated as the outgroup. The tree scale is shown. Branch supporting values ≥ 95% are shown as dots based on 1000 iterations of ultrafast bootstrapping of IQ-TREE. b Nucleotide diversity of the archaeal, bacterial, diatom, and sponge host MAGs. c Abundance of viral families. Gray boxes show zero values. d Nucleotide diversity of the viral families. For boxplots, data points of independent samples and the sample sizes are shown. The center lines represent median values. Box limits represent upper and lower quartiles. Whiskers represent 1.5 × interquartile range. Outliers are detected based on chi-squared scores using the R package outliers and are shown as dark red points. Source data are provided as a Source Data file.

The combination of long- and short-read sequencing enabled us to obtain a partial genome of the sponge host and a nearly complete genome of a symbiotic diatom Bacillariophyta (Fig. 1a). The other MAGs were classified into one archaeal and eight bacterial phyla. Read recovery analysis shows these MAGs were present in almost all sponge individuals but varied in their relative abundance.

Genetic analysis at a single-nucleotide resolution of holobiont members identified substantial intrastrain genetic variations in sponge individuals. In individuals, CC1-CC4, the nucleotide diversity (π) of cellular microbial symbionts ranged from 0 to 0.0065 (mean = 0.0028) (Fig. 1b and Supplementary Data 2). The highest values were observed in the bacteria M10 Rhizobiaceae (mean = 0.0052 ± 0.0016), M38 Pseudohongiellaceae (mean = 0.0037 ± 0.0013), and M11238 Alphaproteobacteria (mean = 0.0036 ± 0.0012), whereas the lowest in M45 Bdellovibrionales (mean = 0.0017 ± 0.0010), M6398 Opitutales (mean = 0.0019 ± 0.0014), and M9 Verrucomicrobiales (mean = 0.0020 ± 0.0015). Moreover, intrastrain-level nucleotide diversities of the sponge hosts (mean = 0.0024 ± 0.0009) were close to the median of the microbial symbionts (0.0029), suggesting each sponge individual contains a population of genetically diverged sponge cells (Fig. 1b and Supplementary Data 2). A similar trend was observed in the metatranscriptomes of individuals CC8-CC28, where the π of the sponge ranked 23 in all 41 MAGs (Fig. 1b).

Sequences of 15,400 viral operational taxonomic units (vOTUs) clustered based on an ANI of 99% and 66 plasmids were obtained from metagenome and metatranscriptome assemblies of sponge tissue, “microbial cell fraction” (particle size > 0.22 and < 3 µm) and “virion fraction” (particle size < 0.22 µm) samples of C. concentrica individuals (Supplementary Data 1, Supplementary Data 3, and Supplementary Fig. 1b). Completeness assessment using CheckV40 found that 0.3% of vOTUs were complete, 1.3% were of high completeness, 0.9% were of medium completeness, and 10.3% were of low completeness. However, the completeness for the majority of vOTUs (87.1%) failed to be determined by CheckV. This proportion is much higher than those of seawater (3.3%)41 but close to those of deep-sea sediments (73.2%)42. Thus, like the deep-sea sediments, the C. concentrica holobiont may possess a substantial diversity of newly discovered viruses, whose completeness was difficult to eval‎uate with the current CheckV database. Supporting this hypothesis, only 933 (6.1%) vOTUs could be annotated to viral families. Double-strand DNA (dsDNA) viruses were the most abundant viral type, followed by single-strand DNA (ssDNA) and RNA viruses (Fig. 1c and Supplementary Data 3). The abundance of plasmids was generally low, ranging from 0 to 6.55 reads per kilobase per million mapped reads (RPKM) (Supplementary Data 3), consistent with previous reports on sponge holobionts43. The most abundant vOTUs were present in all sponge individuals with different relative abundances (Supplementary Fig. 2).

The intrastrain nucleotide diversities of vOTUs ranged from 0 to 0.1 (mean = 0.0032) in the metagenome of the microbial cell fractions of CC1-CC4, 0 to 0.16 (mean = 0.0019) in the metagenome of the virion fractions of CC1-CC3, and 0 to 0.16 (mean = 0.0009) in the metatranscriptome of the sponge tissue of CC8-CC28 (Supplementary Data 3). Such a level of intrastrain nucleotide diversity is even higher than the previously reported intraspecific nucleotide diversity of seawater viruses (approximately mean = 0.0004)44, suggesting faster evolutionary rates in sponge-associated viruses. The highest π values were observed in the viral families Podoviridae-like (single value = 0.0054), Pithoviridae (mean = 0.0048 ± 0.0008), and Pithoviridae (mean = 0.0039 ± 0.0064), whereas the lowest values were observed in the families Lipothrixviridae-like (mean = 0.0015 ± 0.0010), Caliciviridae-like (mean = 0.0002 ± 0.0002) and Siphoviridae-like (mean = 0.0002 ± 0.0006) in the metagenomes of the microbial cell fractions of CC1-CC4, the metagenomes of the virion fractions of CC1-CC3, and the metatranscriptome of the sponge tissue of CC8-CC28, respectively (Fig. 1d).

High strain-level host-specificity and restricted interspecific gene transfer in viruses associated with C. concentrica

Metagenomic HiC is an approach to retrieve physical proximity between DNA fragments and has the potential of linking extrachromosomal DNA like viruses and plasmids to their cellular hosts45. We identified 186 potential virus-host interactions based on their DNA proximity (140, through HiC-read pairing of individuals CC1-CC3), immune memory (33, through CRISPR spacer-protospacer matching), and/or provirus insertion (15, through virus-provirus matching) (Fig. 2 and Supplementary Data 4). No plasmid could be linked to any MAG, possibly due to the low abundance of plasmids found in the data (Supplementary Data 3). Substantial variation in the numbers and types of viral interactions was found among the MAGs — some MAGs and viruses were mainly linked through spacer-protospacer matching (e.g., M411805 Nitrosopumilus), whereas others by virus-provirus matching (e.g., M6398 Opitutales), or solely by HiC read pairing (e.g., M18 Mariniblastus). With two exceptions, all virus-host interactions were exclusively detected by only one method, highlighting the capability of the HiC approach in capturing virus-host interactions that are not immunized by CRISPR-Cas systems or in the form of proviruses. In addition, HiC was effective at discovering viruses of eukaryotic hosts, including the sponge host and the diatom, which are difficult to predict from metagenome sequences using only bioinformatic tools46.

Fig. 2: Virus-host interactive relationships and genome recombination events in the C. concentrica holobiont.

The alternative text for this image may have been generated using AI.

Full size image

Virus-host links based on CRISPR, provirus, and HiC evidence are shown as lines in different colors connecting MAGs and vOTUs. HiC-links are retrieved from the metagenomes of the tissue of individual CC1 and the microbial cell fractions of individuals CC1-CC3. The thickness of a HiC link reflects the number of HiC read pairs supporting the link. Genome recombination events between MAGs and vOTUs are shown as gray lines. Two MAGs with ANI > 95% and < 99% are connected by spiral lines.

Whereas the sponge host and many cellular microbial symbionts were infected by multiple viruses of different species (ANI < 95%), 94.9% (167 out of 176) of viruses each only infected one host strain (ANI < 99%) and 98.3% (173 out of 176) each only infected one host species (ANI < 95%), indicating a high degree of host-specificity for viruses in the C. concentrica holobiont (Fig. 2). Exceptions included a virus V957 linked to M10 and M11238, which both belonged to Alphaproteobacteria but shared a low genome identity (average amino acid identity, AAI = 58.5%), and V1226 and V6338, which were linked to two and three Verrucomicrobiota MAGs, respectively (AAI < 71.0%). Another six vOTUs were linked to two closely related Acidimicrobiales MAGs (AAI = 97.5%). Consistent with the paucity of broad-host-range viruses, we found very few (43) possible horizontal gene transfer (HGT) events between bacterial symbionts in the C. concentrica holobiont (Supplementary Fig. 3 and Supplementary Data 5). This HGT frequency was much less than what was recently reported in the seaweed surface biofilm using the same approach (i.e., 1.1 versus 2.1 transfers per sponge microbial symbiont versus seaweed microbial symbiont)47.

High intracellular viral activity and potential chronic infection

Many host-linked viruses of the C. concentrica holobionts maintained high intracellular abundance and activity showing potential chronic infection features. We used the read coverage of a virus in the tissue of CC1 relative to its bacterial or archaeal host in the tissue of CC1 (VT/HT ratio) to identify active viruses. We also used the read coverage of a virus in the microbial cell fractions of individuals CC1-CC4 relative to its bacterial or archaeal host in the microbial cell fractions of individuals CC1-CC4 (VC/HC ratio) to specify intracellularly active viruses. In total, 62.0% (67 out of 108) of viruses linked to bacteria or archaea were found either to be “active” (VT/HT > 2) or “intracellularly active” (VC/HC > 2) (Supplementary Fig. 4 and Supplementary Data 4). These proportions were approximately seven times higher than that reported in gut microbiomes45,48, suggesting intensive and ongoing microbe-virus interactions in the sponge holobiont. Of these, 28.7% (31 out of 108) were determined to be intracellularly active in at least one of the individuals CC1-CC4. The intracellular activity of viruses was generally supported by their transcriptional level in individuals CC8-CC28 (Supplementary Fig. 4 and Supplementary Data 4). It is likely that these viruses were maintaining a high number of genome copies in host cells without packing and releasing virions, indicating chronic infection. 58.8% (20 out of 34) links between intracellularly active viruses and their hosts were based on CRISPR-matching, whereas only 15.1% (13 out of 86) links between non-intracellularly active viruses and their hosts were based on this method. This result implies that viruses with high intracellular genome copies are more likely to be captured by CRISPR-Cas adaptative immune systems than those maintaining low intracellular copies.

Host impacts on the strain-level compositions of the C. concentrica microbiomes

The individuals of C. concentrica investigated here contained structured genetic backgrounds. A pattern like inter-specific phylosymbiosis (i.e., that the “microbial community relationships recapitulate the phylogeny of their host”49) between the sponge genetics and the strain-level compositions of sponge host viromes and cellular microbiomes was observed, respectively (Fig. 3a). Specifically, Illumina metagenomic sequences of individuals CC1-CC4 were used to extract the gene sequences of ribosomal protein S6e of the sponge host (Supplementary Fig. 5). Phylogenetic analysis of this protein shows that CC1 and CC3 were genetically closer to each other in comparison to CC2 and CC4, whereas CC4 was genetically distant from the other three individuals (Fig. 3a). The community composition (beta diversity) of viruses infecting the sponge host largely mirrored the genetic heterogeneity of the latter (Robinson-Foulds test, dist = 0, P = 0.042) indicating a strong host influence (Fig. 3a). Similarly, the clustering of CC1-CC4 based on dissimilarity of their cellular microbiome composition was consistent with the genetic relationship of their respective sponge hosts (Robinson-Foulds test, dist = 0, P = 0.042).

Fig. 3: Phylosymbiosis-like pattern and nucleotide diversity overlap of holobiont members between individuals of C. concentrica.

The alternative text for this image may have been generated using AI.

Full size image

Illumina metagenomes of microbial cell fractions of individuals CC1-CC4 are used for analysis. a The analysis for phylosymbiosis-like patterns between sponge host genetics and community composition of the sponge host viromes (HiC-linked), the cellular microbiomes, and the viromes of cellular microbial symbionts (HiC-linked). The maximum likelihood tree of the sponge host is inferred from the consensus sequences of sponge ribosomal protein S6e in sponge individuals. The tree scale is shown. The tree is rooted to CC4 as its sequence shows the largest deviation from the others. Branch support value is based on 1000 iterations of ultrafast bootstrapping in IQ-TREE. The community compositions are clustered based on Bray-Curtis dissimilarity using the average method. The information on virus-host links is based on the results of Fig. 2. RF, Robinson-Foulds test (one-sided test). b The analysis for phylosymbiosis-like patterns between the popANI relationships of bacterial symbionts and the community compositions of their viromes. The virome compositions are clustered based on Bray-Curtis dissimilarity using the average method. RF, Robinson-Foulds test (one-sided test). c and d popANIs (c) and SGCs (d) of the sponge host viromes, cellular microbial symbionts, and their viromes between sponge individuals in comparison with the phylogenetic distance of the sponge host. For boxplots, data points of independent samples and the sample sizes are shown. The center lines represent median values. Box limits represent upper and lower quartiles. Whiskers represent 1.5 × interquartile range. Outliers are detected based on chi-squared scores using the R package outliers and are shown as dark red points. e and f Consensus trees of sample clusters based on popANIs (e) and SGCs (f) of cellular microbial symbionts (n = 33). Source data are provided as a Source Data file.

In comparison, such a consistency in sponge individual relationship was not found in the viromes of cellular microbial symbionts — individuals CC1 and CC4 were clustered together, whereas CC2 and CC3 were found similar (Robinson-Foulds test, dist = 1, P = 1, Fig. 3a). To further test the potential impact of cellular microbial symbionts on their virome compositions, we clustered sponge individuals based on the population ANIs (popANIs) or shared genome coverages (SGCs)22 of their microbial symbionts. These two values were used as a proxy for the genetic relationships of these microbial symbionts between sponge individuals. The higher the values are, the more similar the genetic background of the microbial cell populations are shared by two sponge individuals. In two Acidimicrobiales of the four studied bacterial symbionts, each with over eight HiC-linked vOTUs, their strain-level virome composition was shaped by their genetic patterns as indicated by popANI-based clustering (Robinson-Foulds test, dist = 0, P = 0.042), whereas in the other two, this influence is weaker (Fig. 3b). No significant host impact was found by SGC-based clustering (Supplementary Fig. 6).

Intrastrain-level genetic variations between sponge holobionts

We further investigated the intrastrain-level genetic diversity of the cellular microbial symbionts and viruses between sponge individuals CC1-CC4 by using the Illumina reads of their microbial cellular fractions (Supplementary Data 1). For each MAG or vOTU, which represents strain-level classification, its nucleotide variation (i.e., intrastrain-level heterogeneity) between sponge individuals was compared by calculating the popANI or SGC values.

Although it shows clear heterogeneity in phylogenetic distance between the four individuals of sponge hosts, the intrastrain nucleotide diversities of their associated microorganisms were generally highly similar, as shown by median popANI value of 1 for sponge host viruses, values between 0.99979 and 0.99992 for cellular microbial symbionts, and value of 1 for viruses of cellular microbial symbionts (Fig. 3c). The correlations between sponge host phylogenetic distance and average popANI values were not significant for sponge viruses, cellular microbial symbionts or viruses of cellular microbial symbionts (Supplementary Fig. 7a). The inconsistency between the phylogenetic distances of sponge hosts and nucleotide diversities of their microbial cells and viruses is further supported by SGC-based analysis (Fig. 3d and Supplementary Fig. 7b). These results suggest the host selection and/or isolation signals in the intrastrain-level nucleotide variation between sponge individuals were weak.

A more specific analysis identified a few cases of host impact on intrastrain-level nucleotide variation of cellular microbial symbionts or viruses. For example, in only one of the 33 analyzed cellular microbial symbionts, including the diatom, symbiont popANI reflected sponge host phylogenetics (Robinson-Foulds test, dist = 0, P = 0.042), whereas in others, such coherence was not supported (Supplementary Data 6). In SGC patterns, this ratio was five in 33. The topologies of sponge individual clusters based on the popANIs or SGCs of the 33 cellular microbial symbionts variated greatly. Thus, a consensus tree of these clusters with strong branch supports (> 80%) was not obtained (Fig. 3e, f). Furthermore, at single nucleotide resolution, the popANI patterns of nine in 35 bacteriophages reflected the popANI structures of their specific bacterial hosts (Robinson-Foulds test, dist = 0, P = 0.042) (Supplementary Data 6). In SGC patterns, this ratio is six in 35. Common topology in popANI- or SGC-based sponge individual clustering was also rare between viruses of the same bacterial host. Strong branch supports (> 80%) were only found in parts of the consensus trees for the viruses of M10 Rhizobiaceae (popANI, Supplementary Fig. 8a), M5 Acidimicrobiales and M61 Acidimicrobiales (SGC, Supplementary Fig. 8b).

The intensive immune-related arms race between holobiont members

Our previous studies found enriched defense systems against foreign DNA in sponge microbiomes, such as CRISPR-Cas and restriction-modification (R-M) systems, in comparison to planktonic microbiomes30,50. This finding was confirmed and expanded here by analyzing more recently identified defense systems, such as Wadjet, Zorya, and the cyclic-oligonucleotide-based anti-phage signaling system51,52 (Fig. 4a). The most preval‎ent systems were R-M, followed by CRISPR-Cas and Abi, and they were actively transcribed in sponge samples (Supplementary Fig. 9a). In total, 56 CRISPR spacer arrays were identified in bacterial and archaeal MAGs, of which eight contained 50 to 100 spacers and three contained over 100 spacers (Supplementary Data 7), indicating intensive ongoing battles between the adaptive CRISPR-Cas systems of the microbial symbionts and viruses in the C. concentrica holobiont. For example, M411805 Nitrosopumilus encoded a type I CRISPR-Cas system containing 158 spacers, to which ten vOTUs found from the C. concentrica holobiont could be matched, showing successful immunity by this CRISPR-Cas system (Fig. 2 and Supplementary Data 7). Supporting this finding, only one virus with ongoing infection was found for this archaeon by HiC-links. In comparison, M6398 Opitutales might benefit from homoimmunity by harboring proviruses that inhibit secondary infections by closely related viruses25.

Fig. 4: Immune-related proteins encoded in C. concentrica-associated bacterial and archaeal symbionts and their viruses.

The alternative text for this image may have been generated using AI.

Full size image

a Abundance of defense systems against mobile genetic elements in archaeal and bacterial symbionts. Gray boxes show zero values. Cbass, cyclic-oligonucleotide-based anti-phage signaling system. The phylogenetic trees are the same as the ones in Fig. 1a. b Presence of proteins encoding immune- and transposase-related functions in host-linked bacterial and archaeal viruses. Detailed information is in Supplementary Data 8.

In comparison, many bacterial and archaeal viruses also encode a variety of immune system-related proteins, which might contribute to viral immune evasion or to the host’s immune ability against other viruses53,54. These genes were also actively transcribed, as shown by the metatranscriptome data (Supplementary Fig. 9b). R-M-related DNA methyltransferases or methylases were frequently encoded, especially in viruses of the abundant bacterial group Acidimicrobiales (Fig. 4b and Supplementary Data 8). These proteins might help viruses counteract the R-M systems of their bacterial hosts by protecting their DNA from restriction digestion54. In addition, viruses can encode CRISPR-Cas systems to counteract the defense island of the host55. One CRISPR-associated Cas1 and two Cas2 proteins were encoded by the virus V1128 infecting M1 Verrucomicrobiales. This virus was matched by 72 spacers of the CRISPR-Cas system of M1 and was active in individual CC1, suggesting a balanced coexisting and an ongoing arms race between immune protection by the CRISPR-Cas system of Verrucomicrobiales and immune evasion by the virus (Supplementary Fig. 4).

Holobiont-specific intrastrain genetic diversity driven by the virus-host arms race

Arms races between viruses and the immune systems of microorganisms are a primary driver of genetic variation in both viral and cellular host populations. Based on the nucleotide diversity data of individuals CC1-CC4, we found that 31 and 28 genes encoding proteins potentially involved in virus-host interactions, such as CRISPR-associated proteins, integrases, methylases, and methyltransferases were in negative frequency-dependent selection56 (maximum pN/pS ratio > 1) in prokaryotic symbionts and viruses, respectively (Supplementary Data 9). Further analyses revealed heterogeneity of nucleotide mutation and genome structure variation in prokaryotic symbionts and their viruses between sponge individuals directly caused by immune interactions. For example, site mutation is a mechanism for viruses to evade the restriction digestion by microbial hosts and consequently cause genome variations in viral populations54. We found frequent mutations in recognition sites of viral DNA targeted by host R-M systems (Fig. 5a). Of the recognition sites in host-linked viruses, 86.2% (25 out of 29) were found mutated in at least one sponge individual of CC1-CC4. These individuals also varied in the proportion of mutated recognition site loci and the degree of base mutation (Fig. 5a).

Fig. 5: Intrastrain genetic variations driven by immune interactions and genome recombination between cellular microbial symbionts and their viruses.

The alternative text for this image may have been generated using AI.

Full size image

a Mutation proportion and degree at restriction recognition sites in genomes of prokaryotic host-linked viruses in the metagenomes of microbial cell fractions of individuals CC1-CC4. Mutated proportion is calculated by dividing the number of mutated loci by the number of all recognition loci found in the viruses. The mutation degree is calculated by summing up the proportion of mutated bases in all recognition sites found in viruses. Gray boxes show viruses with recognition sites having average mapped bases of less than 100, which are excluded from the analysis. b Read coverage of bacterial contigs containing CRISPR arrays in the metagenomes of microbial cell fractions of individuals CC1-CC4. The array regions are shown with a shaded background. Sequencing depth is the number of bases mapped to each site. c The Spearman correlation (two-sided test) between intrastrain nucleotide diversity of viruses and their bacterial/archaeal hosts in their entire genomes in the metagenomes of microbial cell fractions of individuals CC1-CC4. d The Spearman correlation (two-sided test) between nucleotide diversity of viruses and their bacterial/archaeal hosts in their recombination regions in the metagenomes of microbial cell fractions of individuals CC1-CC4. A sequence identity threshold of 95% is used in the mapping step of inStrain to ensure read coverage. e Genomic map of the transposable virus V957 infecting the bacterium M10 Rhizobiaceae and M11238 Alphaproteobacteria. f Genome recombination regions (shaded sections) between V63657 and M6 Arenicellales, V36967 and M6398 Opitutales, and V4473 and M8024 Alphaproteobacteria. Source data are provided as a Source Data file.

Differences were apparent in the length and content of CRISPR arrays in populations of bacterial symbionts from sponge individuals CC1-CC4, as indicated by analyzing the distribution of sequencing depth over contigs containing CRISPR arrays calculated from mapped reads. For example, the CRISPR arrays in contigs M1_10 matching virus V1128 and M1_96 matching viruses V1128 and V6338 of M1 Verrucomicrobiales were mostly recovered by reads of CC1, whereas the array in contig M1_20 was mostly recovered by reads of CC2 and CC3 (Fig. 5b). Moreover, half of the CRISPR array in contig M6_3 matching virus V3791 was mapped exclusively by reads of M6 Arenicellales in CC2 and CC3, but the other half was also found in CC1 (Fig. 5b). This variation in CRISPR array configuration between sponge individuals reflects the holobiont-specific co-evolutionary history of microbial symbionts and viruses, and points to a mechanism for generation of genetic variants across sponge individuals.

Genome recombination and coupled nucleotide diversity between viruses and their hosts likely promoted by virus-encoded transposases

In virus-host pairs, nucleotide diversity of the paired genomes was positively correlated, suggesting coupled generation of genetic diversity in their long-term co-evolution in the C. concentrica holobiont (r = 0.58, P = 9.66e-33) (Fig. 5c). Although immune interactions, as indicated above, may explain this observation, exchange of genome fragments between viruses and their cellular hosts could have also directly generated linked genomic nucleotide diversities. Indeed, 61.9% (109 out of 176) vOTUs showed sequence similarity to their hosts (alignment length ≥ 300 bp, identity ≥ 90%) including 14 matched to proviruses (Fig. 2). The virus-host coupling in nucleotide diversity were stronger in these recombination regions than in the entire genomes (r = 0.65, P = 3.17e-16) (Fig. 5d). These preval‎ent remains of DNA recombination events in the genomes of viruses and their specific hosts likely recorded their long history of DNA exchange. 90.1% of the recombination activities were detected within each strain-level virus-host pair, suggesting such genome communication was restricted at intrastrain levels (Fig. 2) when interspecific HGT was largely blocked (Supplementary Fig. 3). Moreover, virus-host recombination was only found in HiC- or provirus- but not CRISPR-inferred virus-host pairs implying recombination activity was restricted between viruses and immunized hosts (Fig. 2).

Abundant transposases detected in the genomes of sponge-associated symbionts and viruses here and in previous studies50,57 may explain frequent recombination events (Fig. 4b and Supplementary Data 8). Viruses carrying transposons may jump into the host genomes upon infection58 and inadvertently bring adjacent viral genomic sequences into the host genomes. Similarly, host genomic sequences may be carried to viral genomes through transposon activity. Multiple transposases, such as DDE and Mu-like prophage FluMu transposases, were found in the viruses interacting with symbionts of the taxa Verrucomicrobiales, Acidimicrobiales, and Alphaproteobacteria (Fig. 4b). Specifically, the virus V957 infecting the abundant bacterium M10 Rhizobiaceae and M11238 Alphaproteobacteria was identified as a transposable phage (Fig. 5e). In addition, REP-associated tyrosine transposase was found at the recombination regions of M6398 Opitutales and its virus V36967 (Fig. 5f and Supplementary Data 10). Insertion sequences belonging to the IS66 and IS1380 families were identified at the recombination regions of M8024 Alphaproteobacteria and its virus V4473, and M6 Arenicellales and its virus V63657, respectively.

Discussion

The organization of holobiont microbiomes is a complex process, with many factors potentially driving community assembly and composition3,4. Among these factors, host impacts are often dominant, reflected by congruence between the host’s genetic heterogeneity and microbiome composition dissimilarity, namely phylosymbiosis49. However, the host’s genetic signal can often be confounded by factors such as diet and habitat preference, especially when conspecific host populations with small-scale genetic differences are studied59,60. In this study, individuals of the sponge C. concentrica from the same location were investigated to control for variations in host metabolism and geography. Sponge individuals with different levels of genetic variation were sampled to assess the strength of host impact on microbial and viral community compositions. In line with previous studies based on 16S rRNA genes15,59,60, we detected a strong and significant phylosymbiosis-like pattern between host phylogeny and the strain-level composition of cellular microbiomes (Figs. 3a, 6). Moreover, few studies have reported phylosymbiosis between a host and its virome17,61. Here we used HiC metagenomics to identify virus-host links and detected various levels of impact from the cellular hosts to the strain-level compositions of viruses that infected them. Like the cellular microbiomes, the congruence between sponge host phylogeny and the composition of viruses infecting the sponge host was strong in individuals CC1-CC4 (Fig. 3a). In contrast, congruence between the genetics of prokaryotic symbionts and their viromes were found to be inconsistent, suggesting that the virome composition of C. concentrica holobiont might be influenced by both their cellular hosts and other unknown factors (Fig. 3b and Supplementary Fig. 6). However, we acknowledge that this study would benefit from a larger sample size by recruiting more individuals of the local sponge population. We consider this an important objective for future research.

Fig. 6: A model describing the microbial ecology and evolution at strain and intrastrain levels in individuals of the sponge C. concentrica holobiont.

The alternative text for this image may have been generated using AI.

Full size image

At the strain level (ANI < 99% MAGs and vOTUs as defined in this study), the community composition of the microbial symbionts and sponge viruses is shaped by the genetic heterogeneity of the sponge hosts. At the intrastrain level (ANI > 99% MAGs and vOTUs down to single cells or viruses), this host impact generally disappears, and microbial cells and viruses are frequently exchanged between sponge hosts. The specificity between viruses and their cellular microbial hosts in sponge holobionts is preserved at the strain to species levels, and interspecific HGTs are generally blocked because of strong immunity. At the intrastrain level, interactions and coevolution between virus-host pairs are frequent resulting in holobiont-specific genetic diversity.

The extent of vertical transmission of microbial symbionts is an essential parameter when testing the mode of selection in holobiont evolution1,62,63. However, the methodology is known to affect predictions of vertical transmission13,14,15. We investigated potential patterns of transmission for both cellular microbiomes and viromes between C. concentrica individuals at single nucleotide resolution. Most cellular microbial symbionts and viruses shared a high overlap in nucleotide diversity between sponge individuals with various genetic distances, suggesting an overall genetic decoupling between the sponge host and other holobiont members at the intrastrain level (Fig. 3c, d). This could be best explained by a scenario in which symbiotic microbial cells and viral particles are frequently exchanged between sponge individuals of the local population (Fig. 6). Detailed analysis of each MAG supports this notion that only a small portion of microbial symbionts showed a signal of vertical transmission (Supplementary Data 6), supporting the long-standing speculation that sponge holobionts acquire their symbionts through a mix of vertical and horizontal transmission10,64. Furthermore, it is rational to expect that viruses are transmitted along with their specific bacterial or archaeal hosts between holobionts. However, whereas strain-level virus-host specificity was consistent between sponge holobionts in our study, nucleotide diversity analysis of each MAG and their linked viruses showed there were only a small number of co-transmission cases (Supplementary Data 6). This suggests a shared pool of genetic variations for most microbial symbionts and viruses in the local sponge population (Fig. 6). While their infectious specificity was maintained, microbial cells and viruses were thus likely assembled among sponge individuals following independent rules. Many bacterial vOTUs had very high levels of genetic overlap (up to 100% popANI and 99.5% SGC) between sponge individuals (Fig. 3c, d). These potentially shared viral variants thus showed epidemic-like characteristics in the local C. concentrica population.

At the strain level, 94.9% of viruses specifically infected a single host, suggesting strict host specificity (Fig. 2). In contrast, recent HiC metagenome studies in other environments generally found lower host specificity. For example, host specificity was found in 70% of viruses in human guts at the species level and in 68% of viruses in wastewater at the genus level45,65, whereas 7.1% and 13.9% of viral species in hydrothermal mats were found to infect only one host based on CRISPR-matching and HiC-ligation, respectively66. In addition, HGT frequency between archaeal and bacterial species in C. concentrica was very low, again reflecting the lack of multi-host viruses in the sponge holobiont (Supplementary Fig. 3). Limited HGT by the prokaryotic microbial symbionts of C. concentrica is also supported by no plasmid being linked to any MAGs. In comparison, much higher gene exchange activity occurs in animal guts67, which are considered melting pots of HGT68. Intensive immune interactions during the evolution of viruses and cellular hosts may have contributed to their high host specificity (Fig. 6). As sponges continuously pump large amounts of seawater containing diverse viruses through their bodies10,64, microbial symbionts living in highly dense communities need to combat high level of viral attack and thus have developed powerful anti-viral systems under this long-term selective force (Fig. 4a). As a side effect, it is likely that these defense systems limit HGT across species of cellular microbial symbionts by suppressing the existence of plasmids and broad-spectrum viruses. In contrast, sponge-associated viruses have developed mechanisms to evade host immune effect, for example, by encoding methyltransferases (Fig. 4b). The balance between defense and anti-defense systems of microbial symbionts and their viruses resulting from their long-term interactions and coevolution may have led to viruses maintaining a chronic infection lifestyle with high intracellular viral copies69, as exemplified by M9 Verrucomicrobiales and its viruses V6338, V1226 and V4379 (Supplementary Fig. 4). Long-term exposure between the viruses and the host immune systems iteratively evolves new resistances and infection targets70 and may also drive the high host specificity of viruses associated with C. concentrica.

While C. concentrica-associated viruses showed strict host specificity at the strain level, their transmission and evolution at the intrastrain level were substantial in local sponge populations forming shared (as discussed above) and holobiont-specific genetic patterns, respectively (Fig. 6). Specifically, we identified the involvement of at least three forms of mutation generation by virus-host interaction. First, randomly generated point mutations during replication in the restriction recognition sites of viral genomes were selected to evade digestion by host restriction endonucleases. Heterogenous mutation patterns in these sites were identified in the viromes between sponge individuals (Fig. 5a). Second and on the host side, microorganisms encoding CRISPR-Cas systems acquired viral DNA fragments and developed specific spacer arrays reflecting holobiont-specific history of viral attack. This process generated holobiont-specific genome structural variations in microbial symbionts between sponge individuals (Fig. 5b). Third, frequent exchange of genome fragments by viruses with their highly specific cellular hosts by recombination in their prolonged intrastrain-level coevolutionary history (Fig. 2). This process might have linked the genetic diversification of the viruses and the cellular hosts (Fig. 5d). Supporting this speculation, genome recombination was exclusively found in HiC- and provirus-linked virus-host pairs but not in CRISPR-linked ones (Fig. 2). We propose that the intracellular recombination activities of viruses targeted by host CRISPR-Cas systems might be suppressed and consequently these viruses had little chance for genetic exchange with their immunized host during their coevolutionary history. Moreover, we posit that the activity of the highly abundant transposases may have driven these recombination events, including the identification of virus-encoded transposases (Fig. 5e, f), which would have promoted genome reorganization and increased genetic diversity within the symbiont population71.

In summary, using the marine sponge C. concentrica as a model, we revealed the intraspecific organization and evolution of microbiomes in a subset of its local population. We found evidence for the sponge host’s impact on the strain-level composition of both the cellular microbiomes and the sponge host viromes and observed that intrastrain-level transmission of microbial cells and viruses was likely frequent between sponge individuals. Strict specificity between viruses and their cellular hosts was identified at strain level and virus-host coevolution was restricted at intrastrain level within these virus-host pairs. Immune and evasive interactions were likely responsible for establishing this specificity and simultaneously drove holobiont-specific genetic variances both in the cellular hosts and the viruses. These results provide a structured map for understanding the ecological and evolutionary processes occurring within marine holobionts.

Methods

Overview of experiment design and sample processing

Sponge samples were collected under permit FP23/6, and the guidelines within issued by the Department of Primary Industries of the New South Wales State Government. No ethical approval is required by the University of New South Wales’s institutional policies when sampling invertebrates. The geographic range of the sampling site was in the order of 2500 m2, and there were around 200 individuals of sponge C. concentrica in a patchy distribution. In total, data of 31 sponge individuals as a subset of the local population were analyzed (Supplementary Data 1). Individuals CC1-CC3 and CC7 were collected and sequenced in this study and sequencing reads of all other individuals were generated in our previous studies35,36,37. Oxford nanopore sequences of sponge tissue DNA of CC7, Roche 454 sequences of microbial cellular DNA of CC4-CC6, paired-end (CC1-CC3) and mate-pair (CC4) Illumina sequences of microbial cellular DNA of CC1-CC4 were used to generate MAGs. All but the Roche 454 DNA data plus the Illumina sequences of sponge tissue total RNA of CC8-CC31 were used for genera viral contigs. All Illumina data except those of CC29-31 were used for MAG and vOTU abundance analyses. All Illumina data except those of CC29-31 and sponge tissue of CC1 were used for nucleotide diversity analyses. HiC sequencing information of CC1-CC3 were used in reconstructing the virus-host network.

Sponge sampling and processing

The samples of the sponge C. concentrica were collected at the Bare Island, Sydney, Australia at a depth of 8 m in July 2020. Phenotypically healthy sponge individuals were randomly sampled within a radius of 20 m. Parts of the sponge tissue were directly snap frozen in liquid nitrogen on-site, while the remaining samples were transported in ambient seawater to the laboratory (about 20 min drive), and then directly processed to obtain the microbial cell fractions and virion fractions. Specifically, 50 g of sponge tissue was washed three times in sterile calcium-and magnesium-free seawater on a platform rocker for 3 mins, homogenized in 50 ml of sterile calcium-and magnesium-free seawater and subsequently filtered through a 125-µm pore-sized sieve. The filtrate was centrifuged twice at 300 × g and 4 °C for 15 mins. The supernatant was collected and filtered twice using 11-µm and then 3-µm pore-sized filters. The final filtrate was centrifuged at 15,000 × g and 4 °C for 15 mins to pellet bacterial and archaeal cells. Ice-cold polyethylene glycol 8000 (Sigma-Aldrich) was added at final concentration of 6% v/v to the remaining supernatant, which was then incubated overnight at 4 °C. The mixture was then centrifuged at 76,000 × g and 4 °C for 30 mins to collect the virion pellets, which were immediately snap-frozen in liquid nitrogen for further processes.

DNA sequencing of the sponge tissue and the microbial cell fraction

For nanopore sequencing, total DNA was extracted from the tissue of individual CC7 (randomly selected) using the Blood and Cell Culture DNA mini kit, genomic-tip 20/G (Qiagen). A nanopore library was constructed using the SQK-LSK110 kit (Oxford Nanopore Technologies, City, UK). Metagenomic sequencing and base calling were performed on the GridION platform, and low-quality reads (scores < 7) were removed. For HiC metagenome sequencing, snap-frozen tissue of individual CC1 and enriched microbial cell fractions from individuals CC1, CC2, and CC3 were used to generate the HiC DNA libraries following the standard ProxiMeta protocol developed by Phase Genomics (Seattle, WA, USA). Metagenomic sequencing was conducted on the Novaseq 6000 platform (Illumina) at Shanghai Biozeron Biotechnology Co., Ltd (Shanghai, China).

DNA Sequencing of the virion fraction

The virion fractions from CC1, CC2, and CC3 were processed for DNA extraction following a previously described SDS method72. Extracted viral DNA was then amplified with the REPLI-g Cell WGA & WTA Kits (150054, Qiagen, Hilden, Germany). Illumina sequencing libraries were generated for the amplified viral DNA using the NEB Next UltraTM DNA Library Prep Kit for Illumina (New England Biolabs, City, MA, USA). Sequencing was performed on the Novaseq 6000 platform (Illumina) at Magigene (Guangzhou, China). For the same amplified viral DNA, nanopore PromethION libraries were constructed using the SQK-LSK109 kit (Oxford Nanopore Technologies, City, UK). Sequencing was performed on the PromethION platform (Oxford Nanopore) at Magigene (Guangzhou, China), and low-quality reads of scores < 7 were removed.

Read processing

End N/Xs of raw reads were trimmed by using ngsqctoolkit (v.2.3.3, -t5 -t3 -n 75)73. Illumina adapters and low-quality bases were trimmed, and reads with an average Q score below 25 or shorter than 75 nt were excluded using Flexbar (v.3.5.0)74 and Skewer (v.0.2.2)75. For Illumina HiC paired-end reads, the 3’-end of the ligation site in each read was removed.

Metagenome assembly, MAG generation and characterization

The assembly and binning processes are summarized as a flowchart in Supplementary Fig. 1a. Specifically, Illumina mate-paired reads were assembled using metaSPAdes (v.3.15.3)76 with default setting. Illumina HiC paired-end reads were assembled using metaSPAdes with default setting and MEGAHIT (v.1.1.3, --presets meta-sensitive)77. Illumina contigs longer than 1 K nt were processed along with nanopore reads for binning using BASALT (v.1.0.0, --autopara more-sensitive --refinepara deep)78. In parallel, Illumina contigs were pooled with the contigs of 454 sequencing that we generated previously50 and processed through CDHIT (v.4.8.1, -c 1)79 to remove redundant sequences. The non-redundant contigs were further assembled with nanopore reads using Opera-MS (v.0.9.0)80 with the default setting. Nanopore reads were also separately assembled using Flye (v.2.9, -meta)81. These contig sets were then pooled and processed through CDHIT (-c 1), and polished by using Illumina reads with Nextpolish (v.1.4.0) with default setting82. Polished contigs were processed for binning using VAMB (v.3.0.7)83 with the default setting. Bins generated in the above two independent processes were pooled with polished nanopore contigs longer than 500 K nt and analyzed using both CheckM (v.1.1.3, lineage_wf) and CheckM2 (v.1.0.1) with the default setting. The contamination values from CheckM were adjusted by excluding duplicated markers with AAI no less than 95%, which were expected to come from the same species84 (adjusted contamination = contamination × (100 - heterogeneity) / 100). Bins with completeness over 50% and contamination lower than 5% eval‎uated by both CheckM and CheckM2 were pooled with our previously obtained MAGs28,37, with an exception of the dominant MAG M8024, which has a contamination value of 0.87 and 7.88 as eval‎uated by CheckM and CheckM2, respectively (Supplementary Data 2). dRep (v.2.6.2, -c 0.99)34 was used to obtain the final prokaryotic MAGs. Bins with a length over 2.5 Mb and having more than 90% eukaryotic content as estimated with Tiara85 with the default setting, were identified as eukaryotic MAGs. Prokaryotic and eukaryotic MAGs were further decontaminated using ACDC86 with default setting and their qualities were finally assessed using CheckM, CheckM2, and BUSCO (v.5.3.2)87 with default setting.

Taxonomic classifications of prokaryotic and eukaryotic MAGs are conducted using GTDB-Tk (v.1.7.0)88 based the GTDB taxonomy r202 database and CAT89 based on the NCBI NR database with default settings, respectively. A phylogenetic tree of prokaryotic MAGs was inferred based on 40 single-copy marker proteins90 using IQ-TREE (v.2.0, -m MFP -B 1000 -bnni)91. Pairwise ANI and AAI comparisons of MAGs were calculated using FastANI (v.1.33)92 and CompareM (https://github.com/dparks1134/CompareM) with default settings, respectively.

Viral genome assembly and classification

The bioinformatic workflow for vOTU generation and plasmid identification is shown in Supplementary Fig. 1b. Specifically, Illumina reads of metagenomes were assembled using metaSPAdes in the virus and plasmid assembly mode. Non-redundant contigs were further assembled with nanopore reads using Opera-MS with the default setting. Viral DNA nanopore reads were also individually assembled using Flye (-meta). The two contig sets were pooled, processed through CDHIT (-c 1), and polished by using Nextpolish with default setting. Illumina reads of metatranscriptomes were assembled using both Trinity (v.2.13.2)93 and metaSPAdes with the transcriptome mode. Contigs retrieved by the two methods were then pooled and processed through CDHIT (-c 0.99).

Putative viral contigs were identified from the above pooled metagenome and metatranscriptome contigs with lengths greater than 1.5 K nt (for circular) or 5 K nt (for linear) using VirSorter94, VirSorter295, VirFinder96, Deepvirfinder97, and CAT similarly to the method previously described44. Briefly, contigs ≥ 5 K nt and sorted as VirSorter categories 1 and 2, and/or VirSorter2 score > 0.9 and/or VirFinder score > 0.9 and P < 0.05 and/or Deepvirfinder score > 0.9 and P < 0.05, and contigs ≥ 1.5 K nt sorted as “circular” by VirSorter were defined as viral genome fragments. Contigs that had over two records in the following four conditions were also classified as viral genome fragments: VirSorter categories 3, VirSorter2 score 0.7–0.9, VirFinder score 0.7-0.9 and P < 0.05, Deepvirfinder score 0.7–0.9 and P < 0.05. Contigs that had only one record were further run through CAT and only those having over 40% genes classified as viral genome fragments were added to the viral contig set. The identified viral contigs were then grouped into vOTUs by using anicluster40 based on 99% pairwise ANI and 85% sequence coverage (Supplementary Data 3, vOTUs with prefix V were from metagenomes, and with Vr were from metatranscriptomes). Plasmids were identified using Mash distance search against PLSDB98 with a max distance = 0.1 (Supplementary Data 3, prefix P).

The completeness of the identified vOTUs was assessed using CheckV with default setting. Taxonomic classification of vOTUs with length ≥ 10 K nt was conducted using PhaGCN2 based on the ICTV virus taxonomy profile (2021)99. Taxonomy of the remaining vOTUs were defined at the family level if > 50% of the genome’s proteins were assigned to the same viral family using BlastP (v.2.8.1, identity > 30%, bit-score > 50, and e-value < 0.001)100 against the Viral RefSeq (release 201).

Abundance, activity and nucleotide diversity analysis of MAGs and vOTUs

rRNA and tRNA genes of MAGs and repeat regions of MAGs and viral contigs were masked using dustmasker (https://www.ncbi.nlm.nih.gov/IEB/ToolBox/CPP_DOC/lxr/source/src/app/dustmasker/) with default setting. Illumina reads of the metagenomes and metatranscriptomes were mapped to these masked genomes using BWA-MEM (v.0.7.17, −5SP)101. Low-quality mappings were filtered using Samtools (v.1.9, -q 20, -F 2314)102 and Bamm (v.1.7.3, --percentage_id 0.99 --percentage_aln 0.9) (https://github.com/Ecogenomics/BamM). Mapped reads were counted using BBMap (v.38.73)103 with default setting and normalized to each genome size by calculating RPKM values. Gene expressive levels in metatranscriptomes were also calculated as RPKM values. The community-wide differences of the microbiomes or viromes in sponge individuals were eval‎uated as Bray-Curtis dissimilarity and visualized as a cluster using the average distance method in the R package vegan104. Robinson-Foulds test was used to compare the congruency of dendrogram topologies of sponge host genetics and microbial community composition using the R package phangorn (https://cran.r-project.org/web/packages/phangorn/index.html). Robinson-Foulds P-values were determined by the probability of 100,000 randomized bifurcating dendrogram topologies yielding equivalent or more congruent phylosymbiotic patterns than the tested dendrogram as specified previously105.

Reads were further filtered to keep those with Q score ≥ 30 to conduct nucleotide diversity analyses. Intrastrain-level nucleotide diversity (π) and gene pN/pS ratio of sponge host, microbial symbionts and viruses were analyzed using inStrain (v.1.6.4) by setting mapping identity threshold to 99%. Nucleotide diversity was calculated based on the formula: 1-[(frequency of A)2 + (frequency of C)2 + (frequency of G)2 + (frequency of T)2]. The genetic overlap ratios between sponge individuals were calculated using inStrain with the compare mode. Nucleotide diversity overlaps of MAGs and vOTUs between sponge individuals were assessed using the popANI and SGC values generated with inStrain and relevant clustering of sponge individuals were obtained. The consensus trees of the clusters were generated using IQ-TREE (-con).

Putative coding genes of prokaryotic MAGs were predicted using Prodigal (v.2.6.3, -meta)106. Protein sequences of the genes were searched against the REBASE database (http://rebase.neb.com/rebase) using BlastP (e-value < 1e-5, identity ≥ 50%) to classify restriction enzymes. Their potential recognition sites were informed from the database. Metagenomes of microbial cell fractions of individuals CC1-CC4 were selected for site mutation calculation as viruses in these fractions had sufficient sequencing depth for nucleotide diversity analysis. The recognition sites were then searched in host-linked viruses and the base mutation probabilities were calculated using reads with Q score ≥ 30. Only the recognition sites of a virus having averagely > 100 mapped bases per site per sample were included for analysis.

Virus-host pairing

Links between viruses and the cellular hosts they infected were generated based on three methods. (1) Provirus-based links. Putative proviruses in MAGs were identified using Phaster (http://phaster.ca) with scores > 70. The identified provirus regions were searched in vOTUs by using Minimap2107 with default setting; (2) CRISPR-based links. CRISPR spacers in prokaryotic MAGs were identified using CRISPRDetect3108. Spacer sequences were then matched to vOTUs and plasmids using SpacePHARER (v.5.c2e680a)109 with default setting; and (3) HiC-based links. HiC Illumina reads were mapped to sequences of MAGs, vOTUs and plasmids using BWA-MEM (− 5SP). To keep only reliable HiC-links, we filtered out the aligned read pairs that (a) had MAPQ score < 20, nucleotide identity < 99%, or minimum aligned coverage < 90% (to remove unspecific alignments); (b) mapped to multi-sites with flags XA:Z: and SA:Z: (to exclude potential intragenomic repeats or potential virus-provirus links); or (c) were likely duplicates of another read pair identified using CDHIT (-c 0.99) (to remove remaining amplification artifacts generated during Illumina library preparation). The contig links were extracted from the filtered BAM files. Links between a MAG and a vOTU, that is, in a read pair when one read mapped to a contig of a MAG and the other read mapped to a vOTU contig, supported by two or more HiC read pairs were considered as potential virus-host infectious links. Background noise in HiC data likely resulted from links between extracellular free DNA were further removed by calculating the ratio of HiC read pairs with two read ends mapped to a vOTU and a MAG, respectively, versus those with both read ends mapped to vOTUs. The top 140 vOTU-MAG links with the highest ratio values were kept.

The recombination regions between vOTUs and MAGs were identified using BlastN (v.2.8.1, e-value < 1e-5, alignment length ≥ 300 bp, identity ≥ 90%)100 on repeat region-masked genome sequences. The combined host-virus network was visualized in Cytoscape (v.3.8.2)110 with colored lines representing three prediction methods of host-virus linkages, and genome recombination events. HGT events between prokaryotic MAGs and vOTUs were predicted using MetaCHIP111 with default setting and visualized in Cytoscape.

Viral activity classification

Community-wide virus/host abundance ratios greater than 1 or 2 have previously been considered to define active viruses45,48. Here, we used the virus/host ratio in the CC1 tissue sample (containing intracellular viral DNAs and extracellular virions) to identify active viruses and used the virus/host ratio in the microbial cell fraction samples (containing mostly intracellular viral DNAs) to identify intracellularly active viruses. Specifically, we defined viruses linked to prokaryotic MAGs as generally active (VT/HT ratio > 2), intracellularly active (VC/HC ratio > 2), and dormant (VT/HT ratio < 2 or VC/HC ratio < 2).

Functional annotation of MAGs and viral contigs

Annotation of prokaryotic putative coding sequence was performed by searching them against PFAM112 and KOfam113 using HMMER (v.3.3.2) (http://hmmer.org) with an e-value threshold of 1e-15 and coverage of 35%. Antiviral defense system in prokaryotic MAGs was identified and classified using PADLOC (v.1.1.0)114 with default setting.

Putative coding sequences of vOTUs were predicted using PHANOTATE (v.1.5.0)115 with default setting. Functional annotation of proteins was performed based on PFAM, KOfam, and VOG databases using HMMER with an e-value threshold of 1e-15 and coverage of 35%. Transposable phages were identified based on the co-occurrence of conserved maker proteins using BlastP (e-value < 1e-5) following a previously described instruction116. Insertion sequence elements in identified similar DNA fragments were detected using ISEScan (v.1.7.2.3)117 with default setting.

Sponge phylogeny and potential genetic relationship

During the fractionation process, some sponge cells could have been lysed. The released DNA could have passed through the filter pores and contaminated the microbial cell pellets. Therefore, sponge sequences were present in the sequencing data of the microbial fraction, and they were used to study the sponge’s genetics. The ribosomal protein S6e of the sponge host was selected as a marker gene to infer the phylogeny of sponge individuals CC1-CC4 because this gene contains no repeat regions and had sufficient read coverage in all four individuals. Metagenome Illumina reads of individuals CC1-CC4 were mapped to this gene in the sponge host MAG and consensus sequences of each sponge individual were then aligned using BWA-MEM (− 5SP). The multiple sequence alignment was manually checked and columns containing missing data in any of the four individuals were removed (Supplementary Fig. 5). Phylogeny was processed using IQ-TREE (-m MFP -B 1000 -bnni). The phylogenetic distance between branches was calculated using MEGA-11118 with the p-distance model.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All raw sequencing data generated in this study have been deposited both in the NCBI database under project PRJNA944799 and the National Omics Data Encyclopedia under the project OEP004196. Accession numbers of samples are listed in Supplementary Data 1. The prokaryotic genome sequences are available in LMSG under accession numbers LMSG_G000026884.1-LMSG_G000026920.1. The specific links for genomes are provided in Supplementary Data 2. The sequences of vOTUs and eukaryotic genomes are available in figshare (https://doi.org/10.6084/m9.figshare.24540976.v4). Source data are provided in the Source Data file. Source data are provided with this paper.

Code availability

The code used in this study can be found at Zenodo (https://doi.org/10.5281/zenodo.13293443)119.

References

1.  
2.  
3.  
4.  
5.