HSC · Healing Source Code

血小板偏倚 HSC——为什么巨核命运可能直接挂在层级顶端 Platelet-biased HSCs — why megakaryocyte fate may be wired in at the very top

2026-09-27 · 造血干细胞, 巨核细胞, 血小板, 谱系偏倚, 单细胞移植, von Willebrand factor

血小板是血液里最不像细胞的细胞：没有细胞核，形状像被捏扁的碎片，寿命以天计，因此骨髓必须不停地补。它们的来源是巨核细胞——骨髓里体积最大、染色体倍数最高的那类细胞，成熟后把胞质伸进血窦，被血流剪切成一颗颗血小板。

在标准的造血树图上，这条生产线的起点离终点很远。造血干细胞（hematopoietic stem cell, HSC）先变成多能祖细胞，再分成髓系与淋系两大枝，髓系那枝里又分出红系／巨核共同祖细胞，然后才是巨核祖细胞、巨核细胞、血小板。从顶到底，六七步。这张图的美感在于它整齐：越往下潜能越窄，每一次分叉都是一次不可逆的取舍。

问题是，血小板的临床行为一点也不像一个末梢产物。化疗或移植之后，中性粒细胞通常先回来，血小板往往最后才回来，而且最容易卡住，成了输注依赖和出血风险的主要来源。感染时血小板被大量消耗，骨髓却能在很短的时间里把池子填回去——短到不太像是「从干细胞开始，逐级分化六七步」能解释的。这种时间上的不对称，是过去十几年里推动人们回头审视这条通路的第一个线索。

一个报告基因把干细胞群劈成了两半

真正撬动局面的实验来自 2013 年。Sanjuan-Pla 等人在《Nature》上报告，他们用一只 von Willebrand factor（vWF）报告小鼠——把 eGFP 挂在 vWF 基因上——去看表型上被定义为 HSC 的那群细胞。vWF 是一个和血小板黏附密切相关的蛋白，按经典图谱它不该在干细胞里表达。结果这群细胞被干净地劈成了 vWF-eGFP 阳性和阴性两部分。

阳性那部分不是杂质。它在分子上被血小板特异的基因程序预先启动，功能上在移植后无论短期还是长期都更擅长重建血小板；它常常同时带有长期的髓系偏倚；它能自我更新；而且——这是当时最引人注目的一点——它可以产生 vWF 阴性、淋系偏倚的 HSC，反向却不成立。方向性只有一种读法：这个偏向血小板的亚群不是层级下游的一个分支，它坐在上游。论文的标题就是这么下的：血小板偏倚的干细胞位于造血干细胞层级的顶端。

同一年，Nakauchi 实验室从另一个方向给出了呼应。Yamamoto 等人在《Cell》上发表的克隆分析，标题本身已经说清了结论：能自我更新的谱系限制型祖细胞可以直接由造血干细胞产生。也就是说，「潜能逐级收窄」不是唯一的出口——干细胞可以一步跨到一个窄得多、却仍能长期维持的命运上去。

单个细胞的命运能稳定到什么程度

报告基因只能说明群体被分层，说不清单个细胞的命运有多确定。真正把这件事钉住的是单细胞移植。2018 年，Carrelha 等人在《Nature》上报告，逐个移植单个 HSC 之后，可以看到一类干细胞稳定、有效地重建巨核／血小板这一支谱系树，而几乎不产出其他血系——但它并没有因此丢掉多能性。换句话说，「它只做血小板」不是因为它不会做别的，而是因为它选择性地不做。

更关键的是这种偏好不是移植当下的应激反应。这些模式在二次移植后依然保持——把第一轮受体骨髓里的细胞再取出来移植到第二个受体，命运偏好跟着一起过去了。能被二次移植传递的东西，只能是写在细胞里的、可随自我更新一起复制的状态，而不是环境临时给的指令。这篇论文把这些偏倚排成了一个有层次的谱系命运库：从只重建血小板，到血小板兼髓系，到再加上红系，直到完全多能。

关于整条造血树到底该画成离散分叉的层级还是连续的命运景观，这个连载另有一篇专门处理；这里只跟着巨核这一条支线走，它恰好是整场争论里证据最硬、也最早成形的一段。

应急通道：藏在干细胞表型里的备用军

如果偏倚从顶端就写下了，那么感染时血小板的快速补充就有了新的解释空间。2015 年，Haas 等人在《Cell Stem Cell》上报告，用表面标记定义出来的那个 HSC 区室里，藏着一群他们称为 SL-MkP（stem-like megakaryocyte-committed progenitor，干细胞样的巨核定向祖细胞）的细胞。它们和多能 HSC 共享许多特征，却已经是巨核限制型的。

它们的待机方式很有意思：稳态下保持静息，对日常的血小板生产贡献很小；巨核谱系特异的转录本已经转录好了，但蛋白合成被压住——像一份排好版却没有付印的稿子。急性炎症一来，这套程序被激活，细胞直接从现成的转录本开始生产蛋白，SL-MkP 和其他巨核祖细胞一起成熟，把炎症中损失的血小板补回来。作者把它称为一台应急机器，用来对抗急性炎症期间可能致命的血小板耗竭。

这条旁路的意义不只在速度。它同时说明了一件让所有基于表面标记的实验都必须面对的事：被我们叫做「HSC」的那个流式门里，本来就混着已经定向的细胞。

在没有被扰动的骨髓里，情况也一样

移植终究是一种极端扰动——清髓、归巢、重新定植，每一步都可能改变细胞的行为。所以还需要在原生造血里看一眼。2018 年，Rodriguez-Fraticelli 等人在《Nature》上用转座子条码给未受扰动的小鼠骨髓做克隆追踪，结论之一是：长期造血干细胞是巨核限制型祖细胞的一个重要来源，巨核这条命运在相当程度上独立于其他血系的命运生成。

不需要移植，不需要炎症，安静状态下的骨髓里就有这条相对独立的通路在运转。

偏倚会随年龄漂移

如果命运偏倚是干细胞的内在属性，它就应该像其他内在属性一样随年龄改变。2016 年，Grover 等人在《Nature Communications》上把单细胞转录组和单细胞功能测定放在一起做衰老 HSC，得到的结果比预期更干净：衰老最主要的变化不是髓系偏倚增强、淋系偏倚减弱这个老说法本身，而是血小板程序的分子预启动和功能性血小板偏倚整体上升，其中包括一类此前未被识别的、只产血小板的 HSC 明显增多。当他们通过敲除转录因子 FOG-1 来削弱 HSC 里的血小板程序时，淋系产出反而上去了。

这给「老年人免疫力下降」提供了一个不太直觉的解释路径：淋系产出的减少，可能部分不是淋系通路本身坏了，而是干细胞池被血小板偏倚的克隆挤占了份额。

2024 年的反转：顶端也许根本不是顶端

最有意思的转折发生在 2024 年，而且出自同一个实验室。Carrelha 等人在《Nature Immunology》上报告，重建所有血系的干细胞和几乎只重建血小板的干细胞之间，是非层级关系——不是一个产生另一个，而是两类分开的干细胞，各自使用在细胞学、分子和功能上都彼此独立的通路，去补充分子特征不同的巨核限制型祖细胞：一条是缓慢的稳态多能通路，另一条是被应急激活的、血小板限制的快通道。

2013 年那篇论文把偏倚 HSC 放在顶端，2024 年这篇把「顶端」这个说法本身拆了。这不是自我否认，而是分辨率提高之后的必然修正：当年能看到方向性，是因为群体层面的观察只能给出方向；能看到两条并行通路，需要的是单细胞分辨率下对中间祖细胞的分子刻画。这也顺带解释了那个最初的临床观察——为什么血小板在急需时能被快速补上，却在移植后恢复得最慢：如果快通道依赖的是一类和多能干细胞不同的细胞，那么一份移植物里这类细胞有多少、状态如何，就不是靠计数多能干细胞能预测的。

边界在哪里

需要说清楚的是，上面所有的单细胞移植、报告小鼠和条码追踪都是小鼠实验。vWF 报告小鼠这种手段在人身上没有对应物，人类 HSC 的谱系偏倚只能靠体外分选加移植到免疫缺陷小鼠、或者靠体细胞突变做自然条码来间接推断，分辨率和证据强度都不同。把小鼠里「只产血小板的 HSC」直接搬到人的骨髓里，目前还是一个有待验证的外推。

另一层麻烦更根本。「血小板偏倚的 HSC」和「藏在 HSC 表型里的巨核限制型祖细胞」在流式图上可能落在同一个位置——Haas 那篇讲的正是这件事。区分它们靠的不是标记，而是自我更新的证据：能不能长期维持，能不能经二次移植传递。这也意味着不同实验室因为分选门、报告品系和移植条件不同，得到的比例和结论可以有实质差别，而这类差别很难靠读摘要判断。

一个被默认了很久的假设

回到骨髓里那条生产线。如果巨核／血小板的命运在层级顶端就已经被部分写下，而且写下的方式可以随自我更新一起复制、随年龄漂移，那么一个在基因治疗里被默认了很久的假设就需要被重新检视：我们习惯把采集到的 CD34+ 细胞当成一群功能等价的细胞来处理——测一个群体平均的编辑效率，看一个群体平均的植入率，然后假定被修正的等位基因会按比例出现在所有血系里。

但如果这群细胞里本来就混着命运偏好不同的亚群，那么「哪些细胞被编辑到」和「修正最终出现在哪一系」就不再是同一个问题。一个被完美修正、却只产血小板的克隆，对一位需要红系产出的患者意义有限；反过来，一个在体外扩增条件下被系统性冷落的亚群，可能正是移植后最早需要它的那一支。群体平均值不会告诉你这些，因为它的前提就是每个细胞都一样。

那张整齐的造血树不是错的，它只是把一件事画得太干净了：从顶端出发的路，不止一条，而且有些细胞在出发之前就已经知道自己要去哪里。

参考文献

Sanjuan-Pla A, Macaulay IC, Jensen CT, Woll PS, Luis TC, Mead A, Moore S, Carella C, Matsuoka S, Bouriez Jones T, Chowdhury O, Stenson L, Lutteropp M, Green JCA, Facchini R, Boukarabila H, Grover A, Gambardella A, Thongjuea S, Carrelha J, Tarrant P, Atkinson D, Clark SA, Nerlov C, Jacobsen SEW. Platelet-biased stem cells reside at the apex of the haematopoietic stem-cell hierarchy. Nature. 2013;502(7470):232–236. DOI
Yamamoto R, Morita Y, Ooehara J, Hamanaka S, Onodera M, Rudolph KL, Ema H, Nakauchi H. Clonal analysis unveils self-renewing lineage-restricted progenitors generated directly from hematopoietic stem cells. Cell. 2013;154(5):1112–1126. DOI
Haas S, Hansson J, Klimmeck D, Loeffler D, Velten L, Uckelmann H, Wurzer S, Prendergast ÁM, Schnell A, Hexel K, Santarella-Mellwig R, Blaszkiewicz S, Kuck A, Geiger H, Milsom MD, Steinmetz LM, Schroeder T, Trumpp A, Krijgsveld J, Essers MAG. Inflammation-induced emergency megakaryopoiesis driven by hematopoietic stem cell-like megakaryocyte progenitors. Cell Stem Cell. 2015;17(4):422–434. DOI
Grover A, Sanjuan-Pla A, Thongjuea S, Carrelha J, Giustacchini A, Gambardella A, Macaulay I, Mancini E, Luis TC, Mead A, Jacobsen SEW, Nerlov C. Single-cell RNA sequencing reveals molecular and functional platelet bias of aged haematopoietic stem cells. Nature Communications. 2016;7:11075. DOI
Carrelha J, Meng Y, Kettyle LM, Luis TC, Norfo R, … Jacobsen SEW. Hierarchically related lineage-restricted fates of multipotent haematopoietic stem cells. Nature. 2018;554(7690):106–111. DOI
Rodriguez-Fraticelli AE, Wolock SL, Weinreb CS, Panero R, Patel SH, … Camargo FD. Clonal analysis of lineage fate in native haematopoiesis. Nature. 2018;553(7687):212–216. DOI
Carrelha J, Mazzi S, Winroth A, Hagemann-Jensen M, Ziegenhain C, Högstrand K, Seki M, Brennan MS, Lehander M, Wu B, Meng Y, Markljung E, Norfo R, Ishida H, Belander Strålin K, Grasso F, Simoglou Karali C, Aliouat A, Hillen A, Chari E, Siletti K, Thongjuea S, Mead AJ, Linnarsson S, Nerlov C, Sandberg R, Yoshizato T, Woll PS, Jacobsen SEW. Alternative platelet differentiation pathways initiated by nonhierarchically related hematopoietic stem cells. Nature Immunology. 2024;25(6):1007–1019. DOI

Platelets are the least cell-like cells in blood: no nucleus, shaped like flattened debris, a lifespan measured in days, which means the marrow has to keep replacing them. They come from megakaryocytes — the largest and most polyploid cells in the marrow, which extend their cytoplasm into the sinusoids and let the shear of blood flow tear it into platelets.

On the standard hematopoietic tree, the beginning of that production line is a long way from its end. A hematopoietic stem cell (HSC) becomes a multipotent progenitor, which splits into myeloid and lymphoid branches; the myeloid branch yields a megakaryocyte–erythroid progenitor, then a megakaryocyte progenitor, then a megakaryocyte, then platelets. Six or seven steps from top to bottom. The appeal of the diagram is its tidiness: potential narrows as you descend, and every fork is an irreversible commitment.

The trouble is that platelets do not behave clinically like a terminal product. After chemotherapy or transplantation, neutrophils usually return first and platelets last, and platelet recovery is the step most likely to stall, driving transfusion dependence and bleeding risk. During infection platelets are consumed in bulk, and yet the marrow can refill the pool remarkably fast — faster than “start at a stem cell and walk down six or seven steps” comfortably explains. That asymmetry in timing was the first reason to go back and re-examine the pathway.

A reporter gene splits the stem cell pool in two

The experiment that moved the field arrived in 2013. Sanjuan-Pla and colleagues reported in Nature that a von Willebrand factor (vWF) reporter mouse — eGFP knocked into the Vwf locus — cleanly divided the population defined phenotypically as HSCs into vWF-eGFP-positive and -negative fractions. vWF is a protein tied to platelet adhesion; on the classical map it has no business being expressed in a stem cell.

The positive fraction was not contamination. It was molecularly primed for platelet-specific gene expression, and functionally it reconstituted platelets better than its negative counterpart over both the short and the long term. It frequently carried a long-term myeloid bias as well. It could self-renew. And — the detail that made the paper — it could give rise to vWF-negative, lymphoid-biased HSCs, while the reverse did not hold. Directionality has only one reading: this platelet-leaning subset is not a branch hanging off the hierarchy, it sits upstream of it. Hence the title: platelet-biased stem cells reside at the apex of the hematopoietic stem cell hierarchy.

The same year, the Nakauchi lab supplied an answer from a different direction. The title of Yamamoto and colleagues’ clonal analysis in Cell states the finding outright: self-renewing lineage-restricted progenitors can be generated directly from hematopoietic stem cells. Gradual narrowing of potential is not the only available exit — a stem cell can step straight into a far narrower fate that nonetheless persists long-term.

How stable can one cell’s fate be

A reporter gene shows that a population is stratified; it cannot say how fixed an individual cell’s fate is. That required transplanting single cells. In 2018, Carrelha and colleagues reported in Nature that when HSCs were transplanted one at a time, a distinct class stably and efficiently replenished the megakaryocyte/platelet lineage tree and almost nothing else — while retaining multipotency. The restriction was not an inability to do anything else; it was a selective choice not to.

Crucially, the preference was not a transient response to the stress of transplantation. The patterns persisted through secondary transplantation: take cells back out of the first recipient’s marrow, put them into a second, and the fate preference travels with them. Anything transmissible that way has to be a state written into the cell and copied along with self-renewal, not an instruction the environment supplies on the spot. The paper arranged these behaviors into a hierarchically related repertoire of lineage-restricted fates, from platelet-only through platelet-plus-myeloid and platelet-myeloid-erythroid to fully multilineage.

Whether the blood tree as a whole is best drawn as discrete branch points or as a continuous fate landscape is the subject of another installment in this series; here we follow only the megakaryocyte branch, which happens to be where the evidence is hardest and took shape earliest.

An emergency lane hidden inside the stem cell phenotype

If bias is written at the top, the speed of platelet replacement during infection becomes easier to place. In 2015, Haas and colleagues reported in Cell Stem Cell that the compartment defined by HSC surface markers harbors what they called stem-like megakaryocyte-committed progenitors (SL-MkPs) — cells sharing many features with multipotent HSCs while already being lineage-restricted.

Their standby mode is the interesting part. At homeostasis they are kept primed but quiescent, contributing little to steady-state megakaryopoiesis. Megakaryocyte-specific transcripts are already made, but protein synthesis is suppressed — a manuscript typeset and held back from the press. Acute inflammation activates the program: protein is produced from the pre-existing transcripts, SL-MkPs and other megakaryocyte progenitors mature, and the platelets lost to the insult are replaced. The authors describe it as emergency machinery against life-threatening platelet depletion.

The significance is not only speed. It also states plainly something every marker-based experiment has to reckon with: the flow cytometry gate we call “HSC” already contains committed cells.

The same picture in undisturbed marrow

Transplantation is an extreme perturbation — conditioning, homing, re-engraftment, each step capable of changing how a cell behaves. So the question had to be asked in native hematopoiesis too. In 2018, Rodriguez-Fraticelli and colleagues used transposon barcoding to trace clones in unperturbed mouse marrow and reported in Nature, among other findings, that long-term hematopoietic stem cells are a significant source of megakaryocyte-restricted progenitors, with megakaryocyte fate arising substantially independently of the other lineages.

No transplant, no inflammation. The relatively independent route is running in quiet marrow.

The bias drifts with age

If fate bias is an intrinsic property of stem cells, it should shift with age like other intrinsic properties. In 2016, Grover and colleagues combined single-cell transcriptomes with single-cell functional assays on aged HSCs and reported in Nature Communications a result cleaner than expected: the predominant age-dependent change was not the familiar myeloid-up/lymphoid-down story per se, but an increase in molecular platelet priming and functional platelet bias — including a significant increase in a previously unrecognized class of HSCs that produce platelets exclusively. When they depleted the platelet program by removing the transcription factor FOG-1, lymphoid output went up.

That offers a counterintuitive route to the decline of adaptive immunity with age: the shrinking lymphoid output may be less about the lymphoid pathway breaking than about platelet-biased clones taking a larger share of the stem cell pool.

2024: perhaps the apex was never an apex

The most interesting turn came in 2024, from the same lab. Carrelha and colleagues reported in Nature Immunology that stem cells replenishing all blood lineages and stem cells replenishing almost exclusively platelets stand in a nonhierarchical relationship — not one giving rise to the other, but two distinct stem cell classes using cellularly, molecularly and functionally separate pathways to replenish molecularly distinct megakaryocyte-restricted progenitors: a slower steady-state multipotent pathway and a fast-track, emergency-activated platelet-restricted pathway.

The 2013 paper placed biased HSCs at the apex; the 2024 paper takes the apex apart. This is not a retraction so much as the correction that higher resolution forces. Population-level observation can establish directionality and little more; seeing two parallel routes requires molecular characterization of the intervening progenitors at single-cell resolution. It also loops back to the original clinical observation — why platelets can be topped up fast when urgently needed yet recover most slowly after transplantation. If the fast lane depends on a cell type distinct from multipotent stem cells, then how many of those cells a graft contains, and in what state, is not something a multipotent stem cell count can predict.

Where the limits are

All of the above — the single-cell transplants, the reporter mice, the barcode tracing — is mouse work. There is no human counterpart to a vWF reporter mouse. Lineage bias in human HSCs can only be inferred indirectly, by sorting and transplanting into immunodeficient mice or by using somatic mutations as natural barcodes, at different resolution and different evidential strength. Carrying “platelet-only HSCs” straight across into human marrow remains an extrapolation awaiting test.

A second limitation cuts deeper. A platelet-biased HSC and a megakaryocyte-restricted progenitor hiding inside the HSC phenotype may land in the same place on a flow plot — which is precisely what the Haas paper describes. Telling them apart depends not on markers but on evidence of self-renewal: does the behavior persist long-term, does it survive secondary transplantation? Which also means that differences in sorting gates, reporter strains and transplant conditions can produce substantively different proportions and conclusions between labs, and those differences are hard to judge from abstracts alone.

An assumption that has gone unexamined

Back to the production line in the marrow. If megakaryocyte/platelet fate is partly written at the top of the hierarchy, and written in a form that is copied along with self-renewal and drifts with age, then an assumption long taken for granted in gene therapy deserves re-examination: we habitually treat harvested CD34+ cells as one functionally interchangeable pool — measure a population-average editing efficiency, watch a population-average engraftment rate, and assume the corrected allele will show up proportionally across all lineages.

But if that pool already contains subsets with different fate preferences, then “which cells got edited” and “which lineage the correction ends up in” stop being the same question. A perfectly corrected clone that only makes platelets does limited good for a patient who needs erythroid output; conversely, a subset systematically disfavored by an ex vivo expansion protocol may be the very one needed soonest after infusion. A population average will not tell you any of this, because its premise is that every cell is the same.

The tidy tree is not wrong. It just draws one thing too cleanly: there is more than one road down from the top, and some cells know where they are going before they set out.

References

Sanjuan-Pla A, Macaulay IC, Jensen CT, Woll PS, Luis TC, Mead A, Moore S, Carella C, Matsuoka S, Bouriez Jones T, Chowdhury O, Stenson L, Lutteropp M, Green JCA, Facchini R, Boukarabila H, Grover A, Gambardella A, Thongjuea S, Carrelha J, Tarrant P, Atkinson D, Clark SA, Nerlov C, Jacobsen SEW. Platelet-biased stem cells reside at the apex of the haematopoietic stem-cell hierarchy. Nature. 2013;502(7470):232–236. DOI
Yamamoto R, Morita Y, Ooehara J, Hamanaka S, Onodera M, Rudolph KL, Ema H, Nakauchi H. Clonal analysis unveils self-renewing lineage-restricted progenitors generated directly from hematopoietic stem cells. Cell. 2013;154(5):1112–1126. DOI
Haas S, Hansson J, Klimmeck D, Loeffler D, Velten L, Uckelmann H, Wurzer S, Prendergast ÁM, Schnell A, Hexel K, Santarella-Mellwig R, Blaszkiewicz S, Kuck A, Geiger H, Milsom MD, Steinmetz LM, Schroeder T, Trumpp A, Krijgsveld J, Essers MAG. Inflammation-induced emergency megakaryopoiesis driven by hematopoietic stem cell-like megakaryocyte progenitors. Cell Stem Cell. 2015;17(4):422–434. DOI
Grover A, Sanjuan-Pla A, Thongjuea S, Carrelha J, Giustacchini A, Gambardella A, Macaulay I, Mancini E, Luis TC, Mead A, Jacobsen SEW, Nerlov C. Single-cell RNA sequencing reveals molecular and functional platelet bias of aged haematopoietic stem cells. Nature Communications. 2016;7:11075. DOI
Carrelha J, Meng Y, Kettyle LM, Luis TC, Norfo R, … Jacobsen SEW. Hierarchically related lineage-restricted fates of multipotent haematopoietic stem cells. Nature. 2018;554(7690):106–111. DOI
Rodriguez-Fraticelli AE, Wolock SL, Weinreb CS, Panero R, Patel SH, … Camargo FD. Clonal analysis of lineage fate in native haematopoiesis. Nature. 2018;553(7687):212–216. DOI
Carrelha J, Mazzi S, Winroth A, Hagemann-Jensen M, Ziegenhain C, Högstrand K, Seki M, Brennan MS, Lehander M, Wu B, Meng Y, Markljung E, Norfo R, Ishida H, Belander Strålin K, Grasso F, Simoglou Karali C, Aliouat A, Hillen A, Chari E, Siletti K, Thongjuea S, Mead AJ, Linnarsson S, Nerlov C, Sandberg R, Yoshizato T, Woll PS, Jacobsen SEW. Alternative platelet differentiation pathways initiated by nonhierarchically related hematopoietic stem cells. Nature Immunology. 2024;25(6):1007–1019. DOI