生命的起源

让一个分子同时当说明书和工人，圈就断了

1960 年代，几个人几乎同时想到同一条出路：如果有一种分子，既能像说明书那样把信息按顺序记下来，又能像工人那样折成形状去干活，这个圈就不必是圈——它可以自己撑住自己。

RNA 正好长这样。它跟 DNA 是近亲，同样由一串小零件接成，顺序本身就是信息；但它一般是单股的，不像 DNA 那样两股绞着绷得笔直，所以能自己折起来，折出的形状可以卡住别的分子、把某个反应加快很多倍。存信息和干活，在它身上不冲突。

这本来只是个猜想，直到有人真抓到了会干活的 RNA：1982 年 Cech 发现有一段 RNA 能把自己中间一截剪下来、再把两头接上，全程不用蛋白质帮忙，他和 Altman 为此拿了 1989 年的诺贝尔化学奖。更像化石的证据在 2000 年到手——核糖体是每个活细胞里造蛋白质的那台机器，细菌、蘑菇还是你用的都是它，说明它极其古老；那年它的结构被解出来，人们看清真正把两个氨基酸接在一起的位置周围一圈全是 RNA。整条蛋白质生产线的核心工序，是 RNA 在做。Gilbert 1986 年给这套图景起的名字沿用至今：RNA 世界。

这条线远没走完，两个坑敞着。一是零件：RNA 的小零件在没有生命的环境里怎么攒出来？2009 年 Sutherland 组发表了一条绕开教科书拆法的路线，做出了其中两种——这证明可以做出来，不等于地球上就是这么来的。二是复制：至今没人做出能完整复制自己的核酶，最好的成绩是把短片段扩增一万倍以上，可它一碰到复杂的折叠结构就走不动——而它自己恰恰就是那样一个结构。

EN In every living cell the instructions sit in DNA, but reading and copying them requires proteins, which are themselves built from those instructions. Around 1960 several researchers saw the way out: a molecule that both stores information in its sequence and folds into a working shape would close the loop from inside. RNA is such a molecule. The guess became evidence when Cech (1982) found an RNA that splices itself without protein help and Altman showed the same for another machine — a Nobel in 1989 — and again in 2000, when the ribosome's structure revealed that the site where amino acids are actually joined is surrounded entirely by RNA. Gilbert named the picture the RNA world in 1986. Two gaps stay open: how RNA's building blocks form without life (Sutherland's group published a plausible route to two of them in 2009, showing it is possible, not that it happened), and self-replication — the best ribozymes amplify short templates but stall on the folded structures they themselves are made of.

另一派说：先转起来的不是复制，是一台还没有说明书的发动机

反对意见很直接：RNA 太精巧了，不像开局那一步——一根能存信息的长链要靠现成的能量和足够浓的零件才攒得出来，而一锅稀汤里什么都在往外扩散。所以另一派主张：先转起来的应该是一组互相喂料的化学反应。甲的产物正好是乙的原料，乙的产物又喂回甲；只要外面持续供料，圈里的东西就越转越多。它不需要说明书，化学自己就闭合了。

那场地在哪？1977 年在深海发现的黑烟囱几百度、酸性，很壮观但太烈。对这一派真正有用的是 2000 年在大西洋中脊发现的「失落之城」：温度低得多，流体是碱性的，堆起来的矿物塔像凝固的泡沫，内部密密麻麻全是几十微米的小孔。

为什么这个形状要紧，得先说一件今天的事。地球上每个活细胞产能的方式都一样：把质子（氢原子丢掉电子剩下的那个小东西）泵到膜的一边，再让它们从一个小通道漏回来，漏的过程顺手带动一台分子马达，转出细胞能直接花的能量货币（Mitchell 1961 年提出，后来拿了诺贝尔化学奖）。而那些小孔的孔壁两边天然就有这个落差：孔里是碱性流体，质子少；孔外是偏酸的海水，质子多。一层薄壁两边，现成摆着一个免费的、几万年不断的质子梯度。Russell 和 Martin 由此主张：细胞不是发明了这套产能方式，是先蹭着用，后来才学会自己造膜把它随身带走。

代谢先行自己最硬的那道坎在别处：没有可遗传的变异，就没有选择。一个化学循环再能自持，它偶然转得好一点也传不给「下一个」循环——「自然选择的逻辑」那期讲的三个条件里，它缺了最要命的一条。所以两派现在没当年那么对立：争的是交接点在哪，不是谁全对。还有第三条线在补另一块——脂肪酸这类分子在水里会自己围成小泡，而把东西围起来，稀释就不再是绝症。

EN The competing view is that RNA is too intricate to be step one, and that what turned over first was a set of chemical reactions feeding each other — an autocatalytic cycle needing no instructions. Its favoured setting is not the scalding acidic black smokers found in 1977 but the alkaline field found in 2000 on the Mid-Atlantic Ridge: cooler, alkaline fluid, mineral towers riddled with pores tens of micrometres across. The point is the pore wall. Every living cell today makes energy by pumping protons across a membrane and letting them leak back through a molecular turbine — Mitchell's 1961 idea, a Nobel in 1978 — and a pore wall with alkaline fluid inside and acidic early ocean outside already carries that gradient for free. Russell and Martin argue cells first borrowed it and later learned to build their own membrane. Weiss et al. (2016) reconstructed a common ancestor that was anaerobic, thermophilic and hydrogen-dependent, though lateral gene transfer makes such reconstructions contested. The hard objection to metabolism-first is that a cycle without heritable variation cannot be improved by selection. The camps have converged: geochemistry supplies energy and concentration, an informational molecule joins later, and the argument is about the handover.

能做出来，不等于当初就是这么来的

1953 年，研究生米勒在导师尤里的实验室搭了一套玻璃装置：底下烧水，上面灌进几种简单气体，两根电极隔着放电模拟闪电。放了一星期，他把水抽出来一化验，里面有氨基酸——蛋白质的零件。这是第一次有人在瓶子里造出「生命的材料」。

但这个经典实验今天得反过来读：米勒假设的那种大气后来被认为不对，早期地球的空气现在多认为以二氧化碳和氮气为主，那种条件下产率低得多。不过 2008 年 Bada 组重测他当年封存没发表的样品，模拟火山羽流的那一组产物反而更丰富——整体大气不合适，不代表火山口那一小片不合适。零件本身确实不稀有，1969 年落在澳大利亚的 Murchison 陨石里就检出了几十种氨基酸。

但零件从来不是这道题的难处。这类实验能证明的是「这条路走得通」，不是「当初走的就是这条」。而这个差距填不上，有两个独立的原因。第一，化学不留化石——最早那些东西没有骨头没有壳，连形状都没有，不像三叶虫会在石头里压出一个印子。第二更狠：一个自己撑起自己的系统，最早那一版一旦生出更好的替代品就没理由继续存在。它被抹掉不是意外，是常规结局。

时间窗还在收紧。地球大约 45.4 亿岁；争议最小的生命痕迹是西澳大利亚约 34.8 亿年前的叠层石（一层层微生物垫压出来的岩石纹理），连它也有人主张可能是非生物过程做的。而 2024 年一项分子钟推算把所有现存生命的共同祖先放到了约 42 亿年前，还给它配了一个跟今天细菌相当的基因组。这个年代还会被继续检验，但如果它扛得住，留给「从化学到细胞」的时间就窄得吓人。

最后一个诱人的问题也得留在半空。所有已知生命共用同一套遗传密码，用的氨基酸清一色是左手那一版（同一种分子可以有互为镜像的两个版本，像左右手，生命只用其中一只）。这可以读成「只成功过一次」，也可以读成「成功过好多次，只有一支留了下来」，两种读法目前分不开。最诚实的总结是：我们有好几条各自说得通的路线，还没有一条能从化学一路接到细胞。

EN In 1953 Miller, in Urey's lab, ran a spark through a flask of simple gases above warm water and found amino acids a week later — the first time life's raw material had been made from scratch. Read it in reverse today: the reducing atmosphere he assumed is now thought wrong, though Bada's group re-analysed Miller's sealed unpublished samples in 2008 and found the volcanic-plume apparatus gave a richer yield, so local vent gases may still qualify. Building blocks are not rare anyway — dozens of amino acids turned up in the Murchison meteorite of 1969. But the blocks were never the hard part. Such experiments establish that a route is possible, never that it was taken, for two independent reasons: chemistry leaves no fossils, and in any self-bootstrapping system the earliest version loses its reason to exist once it produces something better. The window is tightening: Earth is about 4.54 billion years old, the least contested traces of life are roughly 3.48-billion-year-old Australian stromatolites, and a 2024 molecular-clock study puts the common ancestor of all living things near 4.2 billion years ago with a genome the size of a modern bacterium's. All known life shares one genetic code and one handedness of amino acids, which reads equally well as one success or as many successes with a single survivor. Those readings cannot yet be told apart.