Tsimane和Hadza群体古肠道微生物与人类迁徙的协同演化证据
Prehistoric global migration of vanishing gut microbes with humans
研究对Bolivia Tsimane(133个粪便样本)和Tanzania Hadza(已有137人数据)进行宏基因组测序,发现87.4%(1231种)的Tsimane肠道菌种与Hadza共享。
文章展示了两个遥远古人群共享肠道微生物组深层测序证据,可为微生物演化研究提供方法范例。
译文尚不完整,完整内容请切换到原文。
主
共生肠道菌群在人类生物学中扮演着复杂角色1,但这些微生物关系的进化历史仍知之甚少9。识别人类与微生物之间的长期共生关系对于理解它们的协同进化7,10,11至关重要,这或许能揭示对人类健康和生理机能至关重要的微生物物种与功能。
通过对古粪便进行测序,已直接观察到历史上的微生物关系12。该研究提示,古代肠道微生物组比工业化人群的更相似于当代非工业化人群(此处“工业化”既指后工业化社会,也指处于高级或进行中工业化阶段的社会)。然而,现有古粪便的年代相对较近(约1,000–2,000年),加之降解古DNA分析的困难,使我们难以确定这些微生物谱系是否在漫长的进化时间尺度上(约10,000–100,000年)被忠实地传递,还是在较近的时间才被获取或传播。
比较基因组学方法提供了一种替代途径,可从当代微生物基因组的遗传多样性模式中推断进化历史8,13,14,15,16,17,18,19。在非人类灵长类动物的先前工作基础上2,8,19,20,21,菌株分辨宏基因组学的进展使得可以将其应用于来自全球人类群体的肠道微生物物种。既往工作已发现肠道微生物基因组与其来源人类群体之间的系统发育相关性7,可能与长期共生关系相一致18,22,23。然而,解读这些相关性仍然困难,因为需要在亚种或菌株水平进行比较。常规系统发育方法24在此类情境下常常失效,因为菌株间的种内基因转移与同源重组产生镶嵌状祖先模式,无法用单一系统发育树表示25,26,27,28。这些高重组率使得难以将古代共迁移与其他群体结构来源(如地理或宿主生活方式差异)区分开来22,23。此外,现有的菌株水平分析偏向工业化人群的肠道微生物物种29。这些采样偏差阻碍了我们推断最可能是祖先共迁移候选物的微生物类群的进化历史。
为了应对这些挑战,我们对玻利维亚的Tsimane农耕者粪便样本进行了深入宏基因组测序(该群体此前已通过16S核糖体RNA测序进行表征6),并将其微生物组与坦桑尼亚Hadza狩猎采集者的微生物组进行比较3。这两个队列代表独特的且在地理上相距遥远的现代群体,其祖先在数千年前就已分离,这为考察广泛候选祖先微生物类群中人类与微生物共同迁徙模式提供了难得的机会。我们在两个群体之间识别出超过1,200种共享微生物物种,其中大多数(60%)在工业化群体中罕见或缺失。通过明确考虑微生物重组的群体遗传学分析,我们提供了菌株水平生殖隔离比在可比工业化群体微生物中观察到的更强的证据。这些发现表明,肠道微生物群落通过代际传递在古人类迁徙过程中得以维持,为理解水平基因转移和宿主-微生物共同进化在人类微生物组长期构建中的作用提供了见解。
测序和基因组恢复
Tsimane人是生活在玻利维亚亚马逊地区的原住民狩猎-农耕者。该队列中的个体居住在Maniqui河沿岸的六个村庄以及附近森林区域的三个村庄6。我们对2009年及2012–2013年从85名Tsimane成年人处收集的133份粪便样本进行了深入宏基因组测序(中位深度 = 31.9吉碱基;扩展数据图1a)。我们恢复了代表1,408种细菌和古生菌物种的12,746个宏基因组组装基因组(MAG)(图1a、方法和补充表1)。然后,我们将这些基因组与先前从坦桑尼亚137名Hadza狩猎采集者的深度测序微生物组样本中获得的类似收集的32,034个MAG进行比较3(扩展数据图1b),以量化过去100,000年来这些群体肠道微生物物种的分化程度。
a,从Tsimane和Hadza微生物组中恢复的微生物物种分类学多样性,按细菌门着色。 b,群体之间共享的微生物物种。 c,来自相同群体与不同群体的同种基因组对的平均核苷酸identity(ANI),汇总自具有至少四个MAG的636种共享物种(n = 1,012,124种内物种间两两基因组ANI比较)。插图,ANI值大于99.75%的相应分布(n = 10,927种内物种间两两基因组比较)。H–H,Hadza–Hadza;T–H,Tsimane–Hadza;T–T,Tsimane–Tsimane。中心线代表中位数;箱体边界代表第25和第75百分位数(四分位距,IQR);须延伸至1.5 × IQR范围内的最极端值;点为单独绘制的离群值。每种物种的结果在扩展数据图6中显示。
Tsimane人和Hadza人共享许多物种
尽管两个群体之间存在巨大的地理隔离且历史上没有直接联系,Tsimane队列中87.4%(1,231个)的物种也在Hadza个体中被发现(图1b)。其中一些共享物种(例如Ruminococcus bromii)在全球范围内也很常见,但大多数(60.2%;848个物种)在工业化人群的粪便微生物组中极为罕见或缺失(扩展数据图2和3及补充表2)。在对不同群体间的测序深度异常值进行控制(扩展数据图2d)或使用基于比对的分析以排除共同的组装偏差(扩展数据图3)时,获得了类似的结果。然而,队列间菌株水平的异质性差异和/或测序深度分布差异可能共同导致某些稀有MAG的回收率存在差异。远处群体间的物种共享可能源于古老的共同迁移,或近期从环境或人类接触中获得。我们试图通过检测这些共享物种中的遗传变异,并将这些模式与历史人类迁移的预期进行比较,来区分这些情景。
作为第一步,我们计算了Tsimane和Hadza队列中每个拥有至少四个MAG的636个共享物种内所有MAG对之间的平均核苷酸同一性(ANI)(扩展数据图4)。这些ANI值提供了每个MAG对最近共同祖先时间上限的估计。肠道细菌的分子钟估计表明,突变至少以每个位点每年10−7的速率积累(扩展数据图5和补充方法1)。Tsimane和Hadza MAG之间的典型ANI(约98%;图1c)因此表明它们在过去10万至20万年间拥有一个共同的遗传祖先(扩展数据图5)。大多数存在于Tsimane样本中的物种在此时间框架内与Hadza中的物种拥有共同祖先这一事实(扩展数据图6),对其之前的进化历史施加了严格的约束,并表明Tsimane的祖先在迁移到美洲期间或之后并未获得一套全新的肠道微生物。
通过比较Tsimane–Hadza基因组对及其对应的群体内对照(Tsimane–Tsimane、Hadza–Hadza)的ANI值,可以获得进一步的见解。与预期一致,这些数据表明,来自同一人类群体的同种基因组对平均比来自不同群体的基因组更加相似(P < 0.001,方差分析,图1c)。然而,效应量小于每个群体内部的变异(平均固定指数(FST)= 0.10;扩展数据图6)。这些分布之间的广泛重叠表明,Tsimane微生物群体中大部分菌株水平的多样性早于地理隔离的开始。
近期菌株共享减少
尽管平均ANI值在地理上表现出较小的差异,但我们观察到在分布的高ANI尾部(>99.75%)存在强得多的趋势,表明种群内部的成对比较相比种群之间存在显著的富集(P < 2.2 × 10−16,Wilcoxon秩和检验;图. 1c,插图)。这种种群内部的富集表明,在近期的进化历史中,这些种群之间的菌株传播受到限制,这与由历史分离驱动的分化而非持续交流是一致的。然而,估算这种分离的时间可能具有挑战性,因为这些高ANI值通常受到与其他菌株重组的强烈影响25,28,30。先前的工作表明,这些高ANI对代表几乎克隆的菌株,其基因组尚未被重组覆盖25,28,31(扩展数据图. 5 和 补充方法 3)。它们的克隆遗传谱系可以从大段几乎相同的DNA序列中推断出来,这些序列之间夹杂着来自其他菌株的重组片段。我们使用具有相同DNA序列的核心基因的比例,估算了每对MAG的克隆遗传谱系总量(图. 2a)。通过用垂直遗传基因组区域中突变积累速率来校准这些估算,我们确定,具有超过10%相同基因的MAG对对应于在大约过去5000年内拥有共同祖先的克隆菌株(图. 2a,b,扩展数据图. 7,补充表 4 和 补充方法 3)。使用这一阈值,我们识别了图. 1c 中636个共享物种中所有这样的克隆菌株对,以研究Tsimane和Hadza之间近期菌株传播的景观。
a,b,Bulleidia sp905194705(a)以及 Tsimane 和 Hadza 之间共享的 636 个 abundant 细菌物种(b)中 MAG 对之间 ANI 与 identical gene sharing 的关系。灰色虚线代表用于界定近期分离谱系的 10% identical gene 阈值;该阈值因物种而异,从 Bulleidia sp905194705 的约 4,000 年,到 b 中物种组的中位数约 2,000 年(95% 百分位区间:500–5,000 年)(Extended Data Fig. 8)。展示了位于灰色虚线上方的 MAG 对比例。星号表示 Bulleidia sp905194705(校正后 P < 0.001,单侧 permutation test,Benjamini–Hochberg 校正)和所有物种(P < 10−4,单侧 permutation test)中共享超过 10% identical genes 的 MAG 比例显著高于随机预期。 c,用于全球谱系共享分析的五个队列的位置。来自亚洲32、欧洲33、北美洲34 和 Hadza 人群3 的数据来自先前发表的研究。灰色圆点和灰色虚线分别代表人群内部的近期谱系共享率(H–H 和 T–T)以及人群之间的近期谱系共享率(T–H)。 d,T–H strain-sharing rates 为 0%、0–5% 和 5–100% 的物种数量(总 n = 636)。T–H strain-sharing rates 为 0% 的物种数量显著高于随机预期(P < 10−4,单侧 permutation test)。插图展示了 5–100% 谱系近期共享的物种中所有人群间比较的 strain-sharing rates。数据为物种间的均值 ± s.e.m.。 e,具有不同行为特征的八个物种的人群内和人群间共享率。上部,四个 T–H strain sharing rates 为 0% 的物种。下部,四个 T–H strain sharing rates > 5% 的物种。点大小表示每个队列中的 MAG 数量,点的位置如 c 所示(无点表示无 MAG);线宽表示近期谱系共享率。
对所有 636 个物种的汇总比较显示,按照此标准,共有 78,746 个 MAG 对(7.8%)是克隆相关的,意味着它们共享的最近期共同祖先最多在 5,000 年前(5 ka;Fig. 2b,c)。其中大多数克隆相关对(93.8%)出现在人群内比较中,表明大多数近期关系存在于同一人类群体中(P < 10−4,permutation test)。此外,Tsimane 和 Hadza 之间共享的少数克隆株对在各物种中分布不均,636 个物种中有 545 个(85.6%)显示两个群体之间没有近期的 strain sharing(Fig. 2d;P < 10−4,permutation test)。这些结果表明,在过去大约 5,000 年中,Tsimane 和 Hadza 之间在大多数共享物种中发生了非常有限的 strain sharing。
然而,这种有限的群体间菌株共享趋势中存在一些显著的例外。某些物种的Tsimane–Hadza菌株共享水平与其群体内基线相当,这与近期传播的特征相符。这些物种包括Akkermansia muciniphila和Bacteroides ovatus(图2e,底部),它们在工业化人群中往往更为富集。为了进一步探索这一模式,我们将分析范围扩展,纳入了来自欧洲、亚洲和北美的额外8,855个MAG32,33,34。结果显示,这些近期共享的Tsimane–Hadza物种在全球范围内也分布有亲缘关系密切的MAG(图2c–e和补充数据图8)。这表明,具有较高Tsimane–Hadza菌株共享率的物种是通过与其他人类群体更近期的互动被引入Tsimane和Hadza群体的。
值得注意的是,与工业化MAG的比较还揭示了另一类物种:这类物种在不同生活方式中普遍存在,但Tsimane和Hadza之间近期几乎未发生菌株传播(图2e和补充数据图8)。这些类群包括诸如Agathobacter rectalis和Ruminococcus bromii等常见物种,这表明尽管这类物种在工业化人群中有广泛的全球分布和近期的共享,Tsimane和Hadza的菌株仍可保持隔离状态。
水平基因转移导致的基因流减少
菌株共享指标的一个局限性在于,克隆对的总数受多种因素影响,例如宿主接触网络35,36和全基因组选择性清除37,这些因素在群体之间难以进行比较。为了进一步佐证我们关于Tsimane和Hadza之间微生物隔离的观察结果(图2),我们采用了一种正交方法,其中包含了剩余MAG对(即那些相同基因占比<10%的对)的信息。尽管其原始克隆祖先已被与其他菌株的重组所覆盖,但这些远缘菌株祖先之间的近期重组事件会产生较短的、几乎完全相同的DNA片段,而背景全基因组ANI较低(95–99%)(图3a)。随后的突变和重组事件会随时间逐渐降解这些片段,从而在相同DNA片段长度与转移事件年龄之间建立起联系。这些事件的时间可用于衡量人类群体内部及群体之间历史微生物基因流的速率(图3a)——这是一种传统的遗传标志,表明它们曾属于同一微生物群体38。
a,重组在不同菌株之间产生较长的相同基因组片段,而随后的突变会缩短这些片段。这导致了片段长度与转移年代之间具有特征性的关系(补充方法 4)。生殖隔离导致不同群体之间的长片段源自隔离前的较古老转移,使得与群体内比较相比,不同群体间比较的聚合片段长度更短。这种差异可用于推断遗传隔离的时间。Syn.,同义;SNV,单核苷酸变异。b,两个示例物种的片段长度经验累积分布函数图,比较Tsimane和Hadza群体之间以及群体内部的情况。bp,碱基对。虚线表示L99长度。校正 P 值(Padj)对应单侧置换检验结果(经Benjamini–Hochberg校正,补充方法 4)。c,多个群体对之间L99指标的比较(Tsimane–Hadza、亚洲–北美以及北美–欧洲)。P 值表示比较每个群体对之间和群体内部L99值结果的Wilcoxon秩和检验结果。d,每个物种估计的群体分裂时间。每个物种的95%置信区间(补充方法 4)列于补充表 3。***P < 10−4 适用于Tsimane–Hadza分裂时间估计值与亚洲–北美以及欧洲–北美估计值的比较(双侧Mann–Whitney U检验)。
我们通过识别636个共有物种(显示代表性物种于图 3b)中每个MAG对之间最长的相同同义位点片段来量化此信号。我们排除了克隆对(>10%相同基因),以关注基因流而非通过共同祖先获得的DNA(如分析于图 2中)。群体内与群体间比较的片段长度差异在较长片段中最为显著,这与近期的重组事件相一致(图 3b 和补充方法 4)。因此,为了系统地量化这一趋势,我们计算了636个物种中片段长度的第99个百分位数(L99长度)。
与我们其他结果一致,大多数物种在Tsimane和Hadza人群之间的比较中,其L99长度短于同一人群内的比较,这表明近亲关系在近期受到遗传隔离的影响(图3c;P < 10−4,Wilcoxon秩和检验)。然而,一些例外——包括大肠杆菌——在不同人群中具有相似的L99长度,反映了它们近期较高的菌株共享水平(图2)。我们对来自亚洲、欧洲和北美的MAG重复了此分析(图3c)。对于这些工业化人群,同一人群内与不同人群之间的L99长度差异小于Tsimane和Hadza人群(Wilcoxon秩和检验;亚洲–北美,P = 0.01;欧洲–北美,P = 0.41)。这表明Tsimane和Hadza人群之间的基因流障碍强于工业化人群。在跨越不同生活方式的常见物种中也观察到了同样的趋势(扩展数据图9,P = 0.018,Wilcoxon秩和检验),这表明它不是由被比较的底层微生物群系中的系统性差异驱动的。
分裂时间与人类迁移相吻合
接下来,我们使用观察到的L99长度(图3c)来估计Tsimane和Hadza微生物组之间的遗传隔离时间。对于每个物种,我们在同一人群内的MAG对中添加了模拟突变,直到它们的L99长度达到观察到的不同人群间的水平。通过将这些计算机模拟的人口分裂与肠道共生菌的估计突变率进行校准28,我们可以估计特定两个人群的肠道菌株最后一次交换DNA以来的年数(图3d,补充表5和补充方法 4)。这些估计显示,工业化人群(欧洲–北美:中位数 = 1,109 年,s.e.m. = 377 年;亚洲–北美:中位数 = 4,637 年,s.e.m. = 1,267 年)之间的基因交换比非工业化人群(Tsimane–Hadza:中位数 = 17,090 年,s.e.m. = 863 年,P < 10−4,双侧Mann–Whitney U-检验)要近得多。值得注意的是,Tsimane和Hadza微生物组之间遗传隔离开始的时间大致与人类最初离开非洲(约60–70 ka)39,40和定居美洲(约16–23 ka)41,42之间的时间框架相吻合,考虑到我们的分子钟估计中的不确定性(补充方法 1)。
为了获得关于微生物分支时间的额外独立估计,我们使用了一种现有的群体推断方法43。该方法不使用相同DNA片段,而是采用完全正交的方法,分析单核苷酸变异(SNV)频率在群体中的分布(图 4a,b,补充方法 5)。尽管将该模型应用于微生物存在挑战,如样本量小和复杂的群体结构(例如,存在支系或亚种),但我们确定了15个具有足够数据和适当群体结构以估计隔离时间的物种(补充表 6;详细选择标准见补充方法 5)。对于每个物种,推断的模型与基于L99长度的模型大致一致(图 3d 和扩展数据图 10),并明确支持微生物种群历史中存在古老的群体分支,而非两个无隔离的种群(图 4b,c 和补充方法 5)。综合来看,这些结果表明,许多肠道微生物物种与人类共存的时长至少与人类在全球范围内的迁徙时长相当。
a,种群分支示意图;每个有色片段代表一个SNV的等位基因频率轨迹。较新的突变在遗传隔离的种群间传播的可能性较小。b,一个示例物种(Vescimonas sp900551995)的种群动态情景比较。该图展示了双种群SNV频率分布的一部分(补充方法 5),说明了SNV群体频率与仅存在于一个种群的SNV比例之间的关系。来自Tsimane和Hadza队列的观察数据(黄色圆点)与两个情景的预测值进行比较:使用完整SNV频率数据推导的推断种群动态模型(黄色曲线;95%置信区间代表Tsplit的不确定性)以及假设自由迁徙的无分支情景(黑色曲线;补充方法 5)。c,右图,Tsimane和Hadza之间共享的15个肠道细菌物种的推断种群分支时间。左图,为供对比,展示了两个人类种群分支的推断时间——走出非洲迁徙和亚洲–美洲分支。这些分支(以±2 s.e.显示)基于先前发表的一项研究48,该研究使用了相同的种群动态建模方法49。白令海峡淹没的日期基于先前发表的一项研究50。每个点代表一个物种(中心值)的最大似然分支时间估计;误差棒显示±2 s.e.每个物种的标准差(s.e.)是根据该物种的位点频率谱,从矩型复合似然拟合的Fisher信息矩阵获得的渐近标准差(方法);每个物种的生物独立样本数量和同义SNV的确切数量列于补充表 6。
讨论
使用几种互补的分析方法,我们证明数百种共生物种在人肠道中栖息了数万年的历史。这些共生体在非洲通过代际传播,并随着人类的全球迁徙而传播。这些发现揭示了塑造当今肠道微生物组物种地理分布的一些历史动态。许多这些物种此前已被鉴定为 VANISH 类群(基于它们在 Hadza 狩猎采集者中的富集,表现为易变和/或与人类工业化社会呈负相关)3,44,45。我们的遗传分析表明,许多 VANISH 物种也存在于 Tsimane 和 Hadza 的常见祖先中,这表明普遍存在的物种丢失可能解释了这些物种在工业化人群中缺失的原因。我们还观察到 BloSSUM 物种(在城市化或现代化社会中激增或被选择)4的例子,它们在工业化和非工业化人群之间广泛共享菌株;这可能是由于物种从工业化人群向 Tsimane 和 Hadza 传播所致。最后,我们确定了第三类物种(稳定且独立于生活方式 [Stable Independent of Lifestyle, StIL]),该类物种在各种生活方式中普遍存在,但具有与史前共同迁徙一致的 Tsimane–Hadza 隔离特征。值得注意的是,在这个 StIL 物种子集中,Tsimane–Hadza 隔离的程度通常比在类似隔离的工业化人群之间观察到的隔离程度更强(图 2e 和扩展数据图 8)。这些观察结果表明,工业化人群中持续的交流可能削弱了共同迁徙的信号;这可能在一定程度上解释了为何之前对这些信号的观察是有限的。
我们的研究结果强调,通过聚焦于接触工业化程度有限的当代人群的微生物组,我们可以更好地理解延伸至人类早期历史的长期人–微生物组关联。尽管这些是当代人群,但它们的肠道细菌之间的遗传关系可以揭示其过去的进化历史,这与在人类基因组研究中包含多样化人群所带来的好处相似46。随着我们扩展对农村和原住民人群微生物组的了解,坚持伦理研究实践并与当地社区建立合作至关重要,在整个研究过程中尊重他们的自主权,确保互惠互利,促进开放对话并建立信任。值得注意的是,随着工业化 VANISH 类群的丢失,这与其增加的几种慢性代谢和炎症性疾病风险相关,这提出了重新引入这些物种可能有助于恢复与人类生物学共同进化的微生物功能并改善人们健康的可能性47。这一努力需要研究人员、伦理学家和原住民社区之间深思熟虑和包容性的对话。
方法
伦理与包容性声明
知情同意在三个层级从 Tsimane 人群中获取:Gran Consejo Tsimane(Tsimane 治理机构)、社区领导者和研究参与者。所有研究方案均经加州大学圣塔芭芭拉分校(3-21-0652)和新墨西哥大学(07-157, 15-133, 17-230)的机构审查委员会批准。在玻利维亚,研究方案经玻利维亚科恰班巴马约尔·德·圣西蒙大学(Universidad Mayor de San Simón)和 Gran Consejo Tsimane 批准。关于样本出口,获得了 Gran Consejo Tsimane、玻利维亚科恰班巴马约尔·德·圣西蒙大学医学院以及玻利维亚圣克鲁斯热带疾病国家中心(CENETROP)的批准。由玻利维亚和美国组成的跨国 Tsimane 健康与生命史项目(THLHP)团队在玻利维亚工作已超过二十年,并培训了数十名由该团队雇用的 Tsimane 研究助理。我们的所有研究均与当地 Tsimane 社区及 Gran Consejo Tsimane 合作开展。收集粪便样本的初衷是分析其微生物组成。然而,作为对参与者的益处,样本还通过野外显微镜检查分析了可用抗蠕虫或抗原虫药物治疗的适用寄生虫种类。研究目的向参与者进行了口头说明。THLHP 还通过教育进行能力建设(与美国非政府组织 One Pencil Project 合作,该组织为 Tsimane 学生就读大学或医学院的公立学校提供教育用品和奖学金)。THLHP 投资了市政医院的卫生基础设施,多年来雇用了十几名玻利维亚医生,在超过 65,000 次患者就诊中提供直接医疗服务,培训了当地卫生官员,建设了正式的医疗和牙科诊所,并为两家当地医院提供了专业知识、人员和设备。
我们还分析了先前发表的关于 2013–2014 年从 Hadza 狩猎采集者收集的粪便样本的宏基因组数据3。在获得 Hadza 参与者的知情同意后(向参与者提供了研究项目的经翻译的口头描述),与坦桑尼亚野外向导协商后合法且合乎伦理地收集了样本。与坦桑尼亚国家医学研究所(National Institute for Medical Research in Tanzania)签署的物资转让协议规定,收集的粪便样本仅用于学术目的。研究许可来自国家医学研究所和坦桑尼亚科学与技术委员会(MR/53i 100/83 和 NIMR/HQ/R.8a/Vol.IX/1542)。近期一项研究的结果3被转化为使用当地语言的信息图,并由一名人类学家和一名坦桑尼亚野外向导传达给 Hadza 人。
样本收集
Tsimane 粪便样品是先前发表的一项研究6的一部分,并作为当前工作的一部分接受了额外的测序。简而言之,两组样品(2009 年和 2012–2013 年)是在 THLHP51 的主持下从九个村落的参与者中收集的,该计划收集 Tsimane 参与者的健康和人口统计数据,并为他们的社区提供初级医疗保健。本研究中使用的粪便样品组来自年龄至少 3 岁的 Tsimane 个体(平均年龄 = 28.4 岁,标准差 = 13.2 岁)。野外样品采集方案是根据后勤挑战制定的,并且在 2009 年和 2012–2013 年队列之间保持一致。在数据收集期间,参与村庄中没有家庭拥有管道,使用旱厕或尿布,或能够使用冷藏设备6。粪便样品被收集在给参与者的前一天发放的无菌尿液标本杯中。研究人员在上午 7 点至 9 点之间访问参与者家中以收集样品。样品在收集后 1–2 小时内被运往野外实验室,使用装有可重复使用冰包的冷却器运输。样品在收集杯中均质化,然后使用非无菌木制压舌板分装到 2 毫升无菌冻存管中。冻存管立即储存在液氮中,然后转移到 −20 °C 冰箱,在运往美国之前被放置在干冰上。样品在美国的 −80 °C 实验室冰箱中储存直至分析。Hadza 队列的样品收集信息在先前研究中提供45。简而言之,在采集样品时,在可用的情况下,我们记录了年龄(由个人提供或根据回忆估算)、性别、体重及相关信息。在从参与者处取回样品后,样品在野外立即放入液氮容器中,以冷冻方式运输至 Arusha,然后以干冰方式空运至美国,在美国维持 −80 °C 直至进一步处理。
文库制备和测序
DNA 使用 MoBio PowerSoil 试剂盒(Qiagen)按照制造商的说明从粪便样品中提取。文库使用 Nextera Flex 试剂盒制备,目标为 10 ng DNA,使用 12 bp 独特双索引条形码,进行 12 个循环以最小化扩增偏差。双端测序(2 × 140 bp)在 UCSD 微生物组与代谢组学核心设施上使用 NovaSeq 6000 的 S4 流动细胞进行。样品在运行中随机化,并反复测序直至达到目标深度。每个样品的最低目标深度为 1 亿对双端 reads(约 28 千兆碱基对)。共生成了 5.25 太碱基对的宏基因组数据。
宏基因组质量控制和组装
原始测序读段使用自定义流程进行解复用和数据处理(https://github.com/MrOlm/nf-genomeresolvedmetagenomics)。简要而言,使用FastP对适配器进行裁剪52。利用Bowtie2去除与人类和PhiX基因组比对上的读段(参考文献53)。使用FastQC确保读段质量54。使用SPAdes组装宏基因组55,并使用Bowtie2将读段映射到组装结果上。使用CoverM测定 contig 覆盖率56。使用QUAST评估组装大小和 contig 指标57,并将组装结果筛选为长度≥1,500 bp 的 contig,用于所有下游分析。
宏基因组组装基因组的回收
Genome binning was performed using METABAT258, and genome bin quality was assessed using CheckM59. Genomes with ≥50% completeness and <10% contamination according to CheckM were retained, in accordance with MIMAG standards60. Genomes with ≥50% completeness and <10% contamination were classified as medium-quality genome bins, and those with ≥90% completeness and <5% contamination as high-quality genome bins. We next used GTDB-Tk61 (r220) to determine the taxonomic classification of each medium- and high-quality genome bin. Genome bins derived from publicly available studies32,33,34 were also re-run using GTDB-Tk r220 to ensure up-to-date taxonomic classification.
Mapping-based prevalence calculations
To determine species presence across samples, quality-filtered metagenomic reads were mapped against the species-level representative genome database using Bowtie2. Alignment files were processed using inStrain quick_profile (v.1.2.14) and CoverM (v.0.4.0) to calculate per-genome coverage statistics. A species was considered present in a given sample if it achieved a breadth of coverage of ≥0.5 (at least 50% of the representative genome was covered by mapped reads). Prevalence for each species was then calculated as the percentage of metagenomes in which it was detected.
Sequence divergence and clonal strain-sharing analyses
To determine the all-versus-all ANI in each GTDB-assigned species, we used dRep62,63 (v.3.5.1) with the following parameters: (dRep compare --S_algorithm goANI --SkipMash). This command ensures that all genomes in a GTDB-assigned species are compared using the goANI algorithm64,65. In brief, goANI identifies open reading frames using Prodigal (v.2.6.3)66, aligns their nucleotide sequences with NSimScan67 and calculates ANI as the average sequence identity of all aligned genes. We used these ANI values to estimate an upper bound on the corresponding divergence times using molecular clock analysis described in Supplementary Methods 1.
To evaluate the statistical significance of the ANI differences within versus between populations, we performed a one-way analysis of variance for each species comparing the ANI distributions of intra- and inter-population strain pairs. We then applied Benjamini–Hochberg corrections to the P values for each species to obtain a false discovery rate-corrected q value for each species. We next used Fisher’s sumlog method to combine these q values across species into a single final P value. We also calculated the fixation index FST (a standard measure of population differentiation) using the procedure described in Supplementary Methods 2.
To perform the strain-sharing analysis (Fig. 2), we first calculated the number of identical genes shared by each genome pair. The set of genes aligned between each genome pair was filtered to include only those with at least 500 bp aligned. The percentage of identical genes was then calculated as the number of gene pairs with 100% nucleotide identity divided by the total number of aligned gene pairs with at least 500 bp aligned.
We defined strains as clonally related when they shared more than 10% identical genes. To estimate the divergence time between such clonal pairs, we used a hidden Markov model-based method28 to identify recent recombination events and obtain recombination-corrected divergence time estimates. Further details on this analysis and the corresponding inference procedure are provided in Supplementary Methods 3.
We used a permutation test to determine whether within-population genome pairs are more clonally related than are between-population genome pairs. For each species, we computed the test statistic (pHH + pTT)/2 − pHT, where pXX denotes the proportion of genome pairs with >10% identical genes in each population combination. We generated a null distribution by permuting population labels across genomes within each species (n = 10,000 permutations) and calculated empirical P values as (n* + 1)/(n + 1), where n* is the number of permutations with a test statistic of at least the observed value. For species-specific permutation tests, P values were corrected for multiple comparisons using the Benjamini–Hochberg procedure. For the global test, we permuted labels within species and pooled pairs across all species to calculate a single test statistic and P value.
SNV catalogues and gene flow analysis
For each species sufficiently represented in the Hadza and Tsimane, we generated a catalogue of SNVs present in the core genomes of that species (core genomes are defined as the core genes present in >90% of MAGs in a species). To create an SNV catalogue for a given species, we first chose one genome in the species as the representative genome based on the highest score determined using the following formula: Completeness − 5 × Contamination + 3 × log10(N50) + Length/106 (N50 is the contig length at which 50% of the total assembled sequence length is contained in contigs of that length or longer). We then used a previously published pipeline to annotate the sites in the representative genome that had SNVs compared with all of the other genomes in the species68 (https://github.com/zjshi/snv_analysis_almeida2019). We next filtered this whole-genome SNV catalogue to include only sites located in core genes. To identify core genes in the pangenome of each species, we called open reading frames for all genomes in the species using Prodigal (v.2.6.3)66, clustered all open reading frames in the pangenome at 90% identity using MMseqs269 and determined which gene clusters were present in at least 90% of the genomes. On the basis of these core genes, we created an additional mask that selected fourfold degenerate sites, thus focusing on sites where all mutations would be synonymous. The resulting SNVs were used to perform the gene flow analysis shown in Fig. 3. Further details of these analyses, including the underlying assumptions and motivations, are provided in Supplementary Methods 4.
Demographic inference using Moments
We used the SNV catalogues to perform the demographic inference shown in Fig. 4. For each species, we computed the frequency of each SNV in the Tsimane and Hadza populations to obtain a joint frequency spectrum across all fourfold degenerate sites. We then used the Moments software package43 to fit these data to a demographic model of a historical population split. Further details of these analyses, including the quality-control criteria used to select the species in Fig. 4, are provided in Supplementary Methods 5.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
Individual-level data are stored in the European Genome-Phenome Archive (EGA), governed by the Tsimane Health and Life History Project (THLHP), and available through restricted access for ethical reasons. The highest priority of the THLHP is to protect the study participants and minimize risk. The THLHP adheres to the CARE Principles for Indigenous Data Governance (collective benefit, authority to control, responsibility and ethics), which ensure that the Tsimane people (1) have sovereignty over how the data are shared; (2) are the primary gatekeepers determining ethical use; (3) are actively included in data collection; and (4) benefit from the data generated and shared whenever possible. The THLHP is also committed to the FAIR Guiding Principles for scientific data management and stewardship (findable, accessible, interoperable and reusable). Requests for individual-level data must be submitted as an application detailing the exact uses of the data and the research questions to be addressed, the procedures that will be used for data security and privacy, potential benefits to the study communities and procedures used to assess and minimize stigmatized interpretations of the research results (the data-sharing policy and data-request forms can be found at https://tsimane.anth.ucsb.edu/data.html). Requests for individual-level data may require approval from the EGA under Data Access Committee EGAC50000000982; the Data Access Policy is accessible publicly from the EGA (EGAP50000000936). The authors and THLHP leadership are committed to open science and are available to assist interested investigators in preparing data-access requests. Processed data files for population genetics analysis and other data are available at Zenodo (https://doi.org/10.5281/zenodo.21765848)70. All Hadza data are publicly available at Zenodo as described previously (https://doi.org/10.5281/zenodo.8072245)3. Source data are provided with this paper.
References
Sommer, F. & Bäckhed, F. The gut microbiota — masters of host development and physiology. Nat. Rev. Microbiol. 11, 227–238 (2013).
Sanders, J. G. et al. Widespread extinctions of co-diversified primate gut bacterial symbionts from humans. Nat. Microbiol. 8, 1039–1050 (2023).
Carter, M. M. et al. Ultra-deep sequencing of Hadza hunter-gatherers recovers vanishing gut microbes. Cell 186, 3111–3124 (2023).
Sonnenburg, J. L. & Sonnenburg, E. D. Vulnerability of the industrialized microbiota. Science 366, eaaw9255 (2019).
Blaser, M. J. The theory of disappearing microbiota and the epidemics of chronic diseases. Nat. Rev. Immunol. 17, 461–463 (2017).
Sprockett, D. D. et al. Microbiota assembly, structure, and dynamics among Tsimane horticulturalists of the Bolivian Amazon. Nat. Commun. 11, 3772 (2020).
Suzuki, T. A. et al. Codiversification of gut microbiota with humans. Science 377, 1328–1332 (2022).
Moeller, A. H. et al. Cospeciation of gut microbiota with hominids. Science 353, 380–382 (2016).
Amato, K. R. & Carmody, R. N. Gut microbial intersections with human ecology and evolution. Annu. Rev. Anthropol. 52, 295–311 (2023).
Nyholm, S. V. & McFall-Ngai, M. J. The winnowing: establishing the squid–vibrio symbiosis. Nat. Rev. Microbiol. 2, 632–642 (2004).
Moran, N. A., McCutcheon, J. P. & Nakabachi, A. Genomics and evolution of heritable bacterial symbionts. Annu. Rev. Genet. 42, 165–190 (2008).
Wibowo, M. C. et al. Reconstruction of ancient microbial genomes from the human gut. Nature 594, 234–239 (2021).
Moodley, Y. et al. The peopling of the Pacific from a bacterial perspective. Science 323, 527–530 (2009).
Linz, B. et al. An African origin for the intimate association between humans and Helicobacter pylori. Nature 445, 915–918 (2007).
Falush, D. et al. Traces of human migrations in Helicobacter pylori populations. Science 299, 1582–1585 (2003).
Thorpe, H. A. et al. Repeated out-of-Africa expansions of Helicobacter pylori driven by replacement of deleterious mutations. Nat. Commun. 13, 6842 (2022).
Mah, J. C., Lohmueller, K. E. & Garud, N. R. Inference of the demographic histories and selective effects of human gut commensal microbiota over the course of human history. Mol. Biol. Evol. 42, msaf010 (2025).
Moeller, A. H., Sanders, J. G., Sprockett, D. D. & Landers, A. Assessing co-diversification in host-associated microbiomes. J. Evol. Biol. 36, 1659–1668 (2023).
Nishida, A. H. & Ochman, H. Captivity and the co-diversification of great ape microbiomes. Nat. Commun. 12, 5632 (2021).
Amato, K. R. et al. Evolutionary trends in host physiology outweigh dietary niche in structuring primate gut microbiomes. ISME J. 13, 576–587 (2019).
Ochman, H. et al. Evolutionary relationships of wild hominids recapitulated by gut microbial communities. PLoS Biol. 8, e1000546 (2010).
Perez-Lamarque, B. & Morlon, H. Distinguishing cophylogenetic signal from phylogenetic congruence clarifies the interplay between evolutionary history and species interactions. Syst. Biol. 73, 613–622 (2024).
Good, B. H. Limited codiversification of the gut microbiota with humans. mBio 17, e03727-25 (2026).
Washburne, A. D. et al. Methods for phylogenetic analysis of microbiome data. Nat. Microbiol. 3, 652–661 (2018).
Sakoparnig, T., Field, C. & van Nimwegen, E. Whole genome phylogenies reflect the distributions of recombination rates for many bacterial species. eLife 10, e65366 (2021).
Lawson, D. J., Hellenthal, G., Myers, S. & Falush, D. Inference of population structure using dense haplotype data. PLoS Genet. 8, e1002453 (2012).
Yahara, K. et al. Chromosome painting in silico in a bacterial species reveals fine population structure. Mol. Biol. Evol. 30, 1454–1464 (2013).
Liu, Z. & Good, B. H. Dynamics of bacterial recombination in the human gut microbiome. PLoS Biol. 22, e3002472 (2024).
Abdill, R. J., Adamowicz, E. M. & Blekhman, R. Public human microbiome data are dominated by highly developed countries. PLoS Biol. 20, e3001536 (2022).
Dixit, P. D., Pang, T. Y., Studier, F. W. & Maslov, S. Recombinant transfer in the basic genome of Escherichia coli. Proc. Natl Acad. Sci. USA 112, 9070–9075 (2015).
Milkman, R. & Bridges, M. M. Molecular evolution of the Escherichia coli chromosome. III. Clonal frames. Genetics 126, 505–517 (1990).
Qin, J. et al. A metagenome-wide association study of gut microbiota in type 2 diabetes. Nature 490, 55–60 (2012).
Li, J. et al. An integrated catalog of reference genes in the human gut microbiome. Nat. Biotechnol. 32, 834–841 (2014).
The Human Microbiome Project Consortium. Structure, function and diversity of the healthy human microbiome. Nature 486, 207–214 (2012).
Beghini, F. et al. Gut microbiome strain-sharing within isolated village social networks. Nature 637, 167–175 (2025).
Valles-Colomer, M. et al. The person-to-person transmission landscape of the gut and oral microbiomes. Nature 614, 125–135 (2023).
Yu, X. A. et al. Genome-wide sweeps create fundamental ecological units in the human gut microbiome. Nature 655, 202–209 (2026).
Arevalo, P., VanInsberghe, D., Elsherbini, J., Gore, J. & Polz, M. F. A reverse ecology approach based on a biological definition of microbial populations. Cell 178, 820–834 (2019).
Haber, M. et al. A rare deep-rooting D0 African Y-chromosomal haplogroup and its implications for the expansion of modern humans out of Africa. Genetics 212, 1421–1428 (2019).
Soares, P. et al. The expansion of mtDNA haplogroup L3 within and out of Africa. Mol. Biol. Evol. 29, 915–927 (2012).
Potter, B. A. et al. Current evidence allows multiple models for the peopling of the Americas. Sci. Adv. 4, eaat5473 (2018).
Bennett, M. R. et al. Evidence of humans in North America during the Last Glacial Maximum. Science 373, 1528–1531 (2021).
Jouganous, J., Long, W., Ragsdale, A. P. & Gravel, S. Inferring the joint demographic history of multiple populations: beyond the diffusion approximation. Genetics 206, 1549–1567 (2017).
Fragiadakis, G. K. et al. Links between environment, diet, and the hunter-gatherer microbiome. Gut Microbes 10, 216–227 (2019).
Smits, S. A. et al. Seasonal cycling in the gut microbiome of the Hadza hunter-gatherers of Tanzania. Science 357, 802–806 (2017).
Ragsdale, A. P. et al. A weakly structured stem for human origins in Africa. Nature 617, 755–763 (2023).
Li, F. et al. Cardiometabolic benefits of a non-industrialized-type diet are linked to gut microbiome modulation. Cell 188, 1226–1247 (2025).
Medina-Muñoz, S. G. et al. Demographic modeling of admixed Latin American populations from whole genomes. Am. J. Hum. Genet. 110, 1804–1816 (2023).
Ragsdale, A. P. & Gravel, S. Models of archaic admixture and recent history from two-locus statistics. PLoS Genet. 15, e1008204 (2019).
Pico, T., Mitrovica, J. X. & Mix, A. C. Sea level fingerprinting of the Bering Strait flooding history detects the source of the Younger Dryas climate event. Sci. Adv. 6, eaay2935 (2020).
Gurven, M. et al. The Tsimane Health and Life History Project: integrating anthropology and biomedicine. Evol. Anthropol. 26, 54–73 (2017).
Chen, S. Ultrafast one-pass FASTQ data preprocessing, quality control, and deduplication using fastp. iMeta 2, e107 (2023).
Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. Methods 9, 357–359 (2012).
Babraham Bioinformatics. FastQC: a quality control tool for high throughput sequence data. v.0.12.0 http://www.bioinformatics.babraham.ac.uk/projects/fastqc/(Babraham Institute, 2023).
Prjibelski, A., Antipov, D., Meleshko, D., Lapidus, A. & Korobeynikov, A. Using SPAdes DE Novo Assembler. Curr. Protoc. Bioinform. 70, e102 (2020).
Aroney, S. T. N. et al. CoverM: read alignment statistics for metagenomics. Bioinformatics 41, btaf147 https://doi.org/10.1093/bioinformatics/btaf147 (2025).
More, S. & More, A. Assessment the quality of genome assemblies by using QUAST tool for metagenomics. Int. J. Recent Technol. Eng. 8, 4253–4259 (2020).
Kang, D. D. et al. MetaBAT 2: an adaptive binning algorithm for robust and efficient genome reconstruction from metagenome assemblies. PeerJ 7, e7359 (2019).
Parks, D. H., Imelfort, M., Skennerton, C. T., Hugenholtz, P. & Tyson, G. W. CheckM: assessing the quality of microbial genomes recovered from isolates, single cells, and metagenomes. Genome Res. 25, 1043–1055 (2015).
Bowers, R. M. et al. Minimum information about a single amplified genome (MISAG) and a metagenome-assembled genome (MIMAG) of bacteria and archaea. Nat. Biotechnol. 35, 725–731 (2017).
Chaumeil, P.-A., Mussig, A. J., Hugenholtz, P. & Parks, D. H. GTDB-Tk: a toolkit to classify genomes with the Genome Taxonomy Database. Bioinformatics 36, 1925–1927 (2019).
Olm, M. R., Brown, C. T., Brooks, B. & Banfield, J. F. dRep: a tool for fast and accurate genomic comparisons that enables improved genome recovery from metagenomes through de-replication. ISME J. 11, 2864–2868 (2017).
Jain, C., Rodriguez-R, L. M., Phillippy, A. M., Konstantinidis, K. T. & Aluru, S. High throughput ANI analysis of 90K prokaryotic genomes reveals clear species boundaries. Nat. Commun. 9, 5114 (2018).
Varghese, N. J. et al. Microbial species delineation using whole genome sequences. Nucleic Acids Res. 43, 6761–6771 (2015).
Olm, M. R. et al. Consistent metagenome-derived metrics verify and delineate bacterial species boundaries. mSystems 5, e00731-19 (2020).
Hyatt, D. et al. Prodigal: prokaryotic gene recognition and translation initiation site identification. BMC Bioinform. 11, 119 (2010).
Novichkov, V., Kaznadzey, A., Alexandrova, N. & Kaznadzey, D. NSimScan: DNA comparison tool with increased speed, sensitivity and accuracy. Bioinformatics 32, 2380–2381 (2016).
Almeida, A. et al. A new genomic blueprint of the human gut microbiota. Nature 568, 499–504 (2019).
Steinegger, M. & Söding, J. MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets. Nat. Biotechnol. 35, 1026–1028 (2017).
Carter, M. M. Prehistoric global migration of vanishing gut microbes with humans [Data set]. Zenodo https://doi.org/10.5281/zenodo.21765848 (2026).
Carter, M. M. Prehistoric global migration of vanishing gut microbes with humans [Computer software]. Zenodo https://doi.org/10.5281/zenodo.21206858 (2026).
Acknowledgements
We thank the participants from the Hadza and Tsimane host villages and their families; the numerous people and organizations providing logistical support and collecting samples in Bolivia and Tanzania, including the THLHP staff members and researchers who provided assistance during field data collection; the Gran Consejo Tsimane; and D. Safaris, J. Changalucha, A. Manjurano, M. G. Domiguez-Bello, M. St. Onge, A. Weakley, B. Merrill, S. Smits, Y. Gautam, D. Bhandari, S. Tandukar, G. P. Gautam, and J. B. Sherchand.
Funding
This work was supported by Coefficient Giving, a Stanford Bio-X Bowes Fellowship to Z.L., the National Institutes of Health/National Institute on Aging (R01-AG054442 to H.K.), Alex and Susie Algard, the National Science Foundation (BCS 0422690 to M.G. and DDIG 1232370 to M.M.), the Wenner–Gren Foundation to M.M., the Thomas C. and Joan M. Merigan Endowment at Stanford University to D.A.R., the National Institutes of Health/National Institute of General Medical Sciences (R35-GM146949 to B.H.G.), and the French National Research Agency under the Investments for the Future (Investissements d’Avenir) programme (ANR-17-EURE-0010 to J.S.). J.L.S. and B.H.G. are Chan Zuckerberg Biohub – San Francisco Investigators.
Author information
Author notes
These authors contributed equally: Matthew M. Carter, Zhiru Liu, Matthew R. Olm
Authors and Affiliations
Department of Microbiology and Immunology, Stanford University School of Medicine, Stanford, CA, USA
Matthew M. Carter, Matthew R. Olm, David A. Relman, Erica D. Sonnenburg & Justin L. Sonnenburg
Department of Applied Physics, Stanford University, Stanford, CA, USA
Zhiru Liu & Benjamin H. Good
Department of Integrative Physiology, University of Colorado Boulder, Boulder, CO, USA
Matthew R. Olm & Parsa Ghadermazi
Department of Anthropology, University of California Santa Barbara, Santa Barbara, CA, USA
Melanie Martin & Michael Gurven
Department of Anthropology, University of Washington, Seattle, WA, USA
Melanie Martin
Department of Microbiology and Immunology, Wake Forest University School of Medicine, Winston-Salem, NC, USA
Daniel D. Sprockett
School of Human Evolution and Social Change, Arizona State University, Tempe, AZ, USA
Benjamin C. Trumble
Center for Evolution and Medicine, Arizona State University, Tempe, AZ, USA
Benjamin C. Trumble
Institute of Human Origins, Arizona State University, Tempe, AZ, USA
Benjamin C. Trumble
Economic Science Institute, Chapman University, Orange, CA, USA
Hillard Kaplan
Toulouse School of Economics, Toulouse, France
Jonathan Stieglitz
Universidad Mayor de San Simón, Cochabamba, Bolivia
Daniel Eid Rodriguez
Department of Medicine, Stanford University School of Medicine, Stanford, CA, USA
David A. Relman
Infectious Diseases Section, Veterans Affairs Palo Alto Health Care System, Palo Alto, CA, USA
David A. Relman
Department of Biology, Stanford University, Stanford, CA, USA
Benjamin H. Good
Chan Zuckerberg Biohub, San Francisco, CA, USA
Benjamin H. Good & Justin L. Sonnenburg
Center for Human Microbiome Studies, Stanford University School of Medicine, Stanford, CA, USA
Justin L. Sonnenburg
Authors
- Matthew M. Carter
- Zhiru Liu
- Matthew R. Olm
- Melanie Martin
- Daniel D. Sprockett
- Parsa Ghadermazi
- Benjamin C. Trumble
- Hillard Kaplan
- Jonathan Stieglitz
- Daniel Eid Rodriguez
- David A. Relman
- Erica D. Sonnenburg
- Michael Gurven
- Benjamin H. Good
- Justin L. Sonnenburg
Contributions
Conceptualization: M.M.C., Z.L., M.R.O., B.H.G. and J.L.S. Methodology: M.M.C., Z.L., M.R.O., P.G., B.H.G. and J.L.S. Software and analysis: M.M.C., Z.L., M.R.O., P.G. and B.H.G. Writing, review and editing: M.M.C., Z.L., M.R.O., M.M., D.D.S., B.C.T., H.K., J.S., D.E.R., D.A.R., E.D.S., M.G., B.H.G. and J.L.S. Visualization: M.M.C., Z.L., M.R.O., B.H.G. and J.L.S. Administration of the THLHP: B.C.T., H.K. and M.G. Project administration: B.H.G. and J.L.S. Supervision: B.H.G. and J.L.S. Funding acquisition: D.A.R., B.H.G. and J.L.S. All authors read and approved the manuscript.
Corresponding authors
Correspondence to Benjamin H. Good or Justin L. Sonnenburg.
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature thanks the anonymous reviewers for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data figures and tables
Extended Data Fig. 1 Metagenomics workflow and curation of Tsimane and Hadza data sets.
a, Overview of computational workflow, tools used and primary data generated from Tsimane stool samples. b, Overview of genome database curation for Tsimane and Hadza cohorts.
Extended Data Fig. 2 Prevalence of gut bacterial species in the Hadza, Tsimane and Industrial populations.
a, Scatter plot showing the prevalence of the 1,440 species recovered in the Tsimane (x axis) and in the Hadza (y-axis) as well as industrial populations (color scale). b, Violin plots summarize the prevalence of these gut microbial species in the Hadza, Tsimane and Industrial populations. c, Rarefaction curves showing the number of unique species detected in each population given some number of randomly selected subjects. We then fit a species accumulation curve [according to S = (Smax × N) / (k + N), where S is the number of species, N is the number of subjects and Smax is the maximum number of species] for each population’s rarefaction curve. Using these curves we can estimate the asymptote, or the maximal number of species present in the population. For the Hadza, the estimated number of species is 2,443 (95% C.I.: 2,385–2,503 species); for the Tsimane the estimated number of species is 1,590 (95% CI: 1,561–1,620). d, Depth-controlled comparison of species recovery between the two cohorts. Bars show the number of gut bacterial species detected only in the Hadza, shared between both cohorts, or detected only in the Tsimane, under two subsetting schemes: “No filter” (all samples) and “Depth-filtered”, in which the 4 Hadza samples (1.2%) sequenced more deeply than the deepest Tsimane sample were removed to match sequencing depth between cohorts. Matching depth changes the number of Hadza-only species by only 2.0% (1,109 to 1,087) and the number of shared species by 0.3% (1,231 to 1,227), indicating that the larger set of species recovered from the Hadza is not an artefact of their greater sequencing depth. n = 331 Hadza and 133 Tsimane biologically independent fecal samples.
Extended Data Fig. 3 Comparison of MAG recovery-based and read mapping-based prevalence metrics.
a, Comparison of species prevalence metrics using a MAG recovery based approach and a read mapping based approach (Methods). Pearson correlation: 0.74, p = 0.001, two-sided. b, Read-mapping based prevalence of species in industrial samples that were deemed to be absent from industrialized samples using MAG recovery based approach (703 of 759 species have a prevalence of less than 1%).
Extended Data Fig. 5
a, Schematic illustration of the bacterial molecular clock. A pair of strains descended from a clonal ancestor at time t = 0 accumulate genetic differences via (i) de novo mutations (red) and (ii) homologous recombination events from other strains (blue blocks), which can introduce multiple genetic differences at the same time (yellow lines). These differences accumulate as the divergence time increases, reducing the overall ANI as well as the fraction of identical genes (light blue). Neglecting recombination leads to an upper bound on the expected time to the most recent common ancestor of a pair of MAGs (Supplementary Methods 1). b, Implied upper bound of divergence times based on ANI. Shaded regions denote the TMRCA values that are consistent with a given ANI level. The solid red curve uses the midpoint estimate of the yearly mutation rate, while the dashed red curve denotes an approximate upper bound.
Extended Data Fig. 6
a, Distribution of average within-species ANI for each of the three population contrasts. b, Distribution of bottleneck scores for each of the 636 species. The bottleneck score was calculated for each species as: [Avg. T-T ANI / Avg. H-H ANI]. The median of the distribution is 1.00, represented by the vertical black line. (c) Distribution of population fixation indices (Fst) for each microbial species (Supplementary Methods 2). The vertical dashed line represents the approximate fixation index between host genomes for the Hadza and Tsimane (Fst ~ 0.25, Verdu et al., 2014).
Extended Data Fig. 7
a, Schematic illustration of the age boundary of clonal strains. A pair of strains descended from a clonal ancestor gradually lose their shared clonal background over time through (i) de novo mutations (red lines) and (ii) homologous recombination events from other strains (blue blocks). To determine a species-specific age boundary of clonal strains, recombination events and the clonal divergence times are inferred from closely related strain pairs (> 50% identical genes). The average trend across the aggregated data per species is then used to extrapolate the divergence time corresponding to the clonal strain boundary, defined as 10% identical genes (Supplementary Methods 3). b, Scatterplots showing the linear trend between the inferred clonal divergence time and the fraction of identical genes for six example species. Each blue dot denotes one pair of MAGs. Grey symbols denote the linearly extrapolated divergence time corresponding to 10% identical genes. c, 2-D histogram (density heatmap) analogous to panel a, showing the linear trend for data aggregated across species. Solid line shows the linear fit using the data of all 44,780 MAG pairs with more than 50% identical genes. Color bar indicates the number of MAG pairs within each bin. d, Histogram of the extrapolated clonal divergence times corresponding to 10% identical genes, for each microbial species. Results of recombination inference for each pair of closely related strains are included in Supplementary Table 4.
Extended Data Fig. 8 A larger collection of map plots, similar to the ones shown in Fig. 2e.
The top three rows (magenta lines) have Tsimane-Hadza (T-H) strain sharing rates of 0%, the fourth row (gray lines) shows species with T-H strain sharing rates between 0% and 5%, the bottom two rows (blue lines) show species that have T-H strain sharing rates >5%. Size of dots represent the number of MAGs of that species derived from each population (no dot indicates no MAGs from that population). Thickness of connecting lines indicates the recent lineage sharing rate.
Extended Data Fig. 9
Species-specific comparisons of the gut bacterial gene flow metric (L99 score, Fig. 3a) for species that are present in both the Tsimane and Hadza cohorts, as well as industrialized populations from Asia and North America. These comparisons show that for many individual species, the genetic isolation between the Tsimane and Hadza strains is stronger than that observed between comparable industrialized populations, even after controlling for differences in species composition (Wilcoxon test, P = 0.018, two-sided).
Extended Data Fig. 10
Scatter plot showing the correlation between the identity-by-state (IBS) inferred split time in years with the split time inferred by Moments demographic modeling for the 15 species for which we have data for both analyses (Supplementary Methods 5). Pearson correlation coefficient = 0.581, P = 0.022, two-sided.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
About this article
Cite this article
Carter, M.M., Liu, Z., Olm, M.R. et al. Prehistoric global migration of vanishing gut microbes with humans. Nature (2026). https://doi.org/10.1038/s41586-026-11106-1
Received: 22 May 2025
Accepted: 01 September 2026
Published: 07 October 2026
Version of record: 07 October 2026
DOI: https://doi.org/10.1038/s41586-026-11106-1
来源:Nature 综合主刊 · nature.com