Yang Liu

← Primer

HSC · Healing Source Code

静息不是关机——造血干细胞为「不工作」付出的代价,以及基因治疗为何在第一步就撞上它 Quiescence Is Not Power-Off — What Hematopoietic Stem Cells Pay to Do Nothing, and Why Gene Therapy Hits It on Step One

2026-07-25 · 造血干细胞, quiescence, 基因治疗, 基因编辑, 细胞周期

人体每一秒都在造出数以百万计的血细胞,这条生产线一天不停、一生不停。可位于最上游的那群细胞几乎什么都不做——它们安静地待在骨髓里,几个月才分裂一次,有的一整年不动。直觉会把这理解成待机:能量不够,先睡着。过去二十年的证据说的却是反面——那种”什么都不做”是被一整套分子机器主动按住的结果。而当我们想给这些细胞换一段基因,第一个要谈判的对象,恰恰就是这份被精心维持的静默。

先把”慢”数出来

要说清 quiescence 是什么,得先知道它到底有多慢。1999 年,Cheshier 与 Weissman 等人在 PNAS 上做了直接的计数:正常成年小鼠里,任一时刻约 75% 的长期造血干细胞处于 G0,只有约 5% 在 S/G2/M。但用 BrdU 追踪下去,池子并没有静止——每天约有 8% 异步地进入细胞周期,半年后 99% 都掺入了标记;换算下来,平均每 57 天分裂一次。

2008 年,Wilson 等人在 Cell 上把 BrdU 与 histone H2B-GFP 两种 label-retaining 方法叠在一起,在小鼠骨髓里分出一群 dormant 干细胞;计算模型给出的分裂间隔约 145 天,相当于一生只分裂五次左右,而它们恰恰承载了绝大部分多谱系长期自我更新能力。人身上没法打 BrdU,2011 年 Catlin 等人在 Blood 上改用女性血细胞里父源/母源 X 染色体表型比例随年龄的漂移反推,得到人类造血干细胞平均每约 40 周复制一次(区间 25–50 周)。那是推断而非直接观察,但量级与小鼠自洽:以年为单位,不是以天。

“慢”也不等于”固定不动的一小群”。Takizawa 等人 2011 年在 J Exp Med 上用 CFSE 追踪分裂史,发现稳态小鼠体内能重建终生造血的细胞,既出现在 7 周内分裂五次以上的快周期部分,也出现在 12–14 周一次都不分裂的静止部分;而大量增殖过的细胞倾向于重新回到 quiescence。静息因此不是身份,而是可进可出的状态——这一点后面会很关键。

不工作是要花力气的

如果静息只是能量不足导致的停摆,把它打断就不该付出什么。事实正相反:拆掉维持它的任何一个部件,这群细胞都会滑向增殖,然后被耗尽。

最直白的是刹车本身。2011 年,Matsumoto 等人在 Cell Stem Cell 上报告,Cip/Kip 家族里 p57(CDKN1C)是静息干细胞中最丰富的一个;敲掉它的小鼠,处于 G0 的比例下降、自我更新出现严重缺陷,而把 p27 基因敲入 p57 位点就能纠正这些异常。松开刹车的代价也被量过:2006 年 Yilmaz 等人在 Nature 上删掉小鼠的 Pten,干细胞被推入增殖,几天内出现骨髓增殖性疾病、几周内出现可移植的白血病,而正常干细胞被细胞自主的机制耗竭,不再能稳定重建受照射小鼠;次年 Tothova 等人在 Cell 上把 FoxO1、FoxO3、FoxO4 一起从成年小鼠的造血系统删掉,得到的同样是一个缩小、长期重建能力受损的干细胞区室。松开刹车换不来更旺盛的造血,只换来一个更快烧完的池子。

代谢与合成同样被主动压低。2013 年 Takubo 等人在 Cell Stem Cell 上证明,小鼠造血干细胞主要靠无氧糖酵解产能,而这依赖 pyruvate dehydrogenase kinase 主动抑制代谢物流入线粒体;敲掉 Pdk2 与 Pdk4,quiescence、糖酵解与移植能力一起下滑。次年 Signer 等人在 Nature 上量出另一个被压住的指标:小鼠体内的造血干细胞每小时合成的蛋白量低于大多数造血细胞;而把合成压得更低或抬得更高,功能都会受损。

连清除也是主动的。2013 年 Warr 等人在 Nature 上发现,面对 ex vivo 撤走细胞因子或在体内限制热量,小鼠造血干细胞能迅速启动 autophagy,而它们短命的髓系后代不能;FOXO3A 的作用是让干细胞预先上好膛。2017 年 Ho 等人在 Nature 上进一步指出,在小鼠里,autophagy 是靠清除仍然活跃的健康线粒体来压低代谢、维持 quiescence 与干性的。同年 Cabezas-Wallscheid 等人在 Cell 上用单细胞 RNA-seq 画出连续地图:从 dormancy 走向细胞周期不是一级级台阶,沿途生物合成程序整体上调;而喂食无维生素 A 饮食的小鼠会丢失干细胞,炎症刺激之后回归 dormancy 的过程也受损——连这份静默的原料,都得从外面供给。

静息因此不是缺省值,而是被 CDK inhibitor、转录因子、代谢通路与 autophagy 一层层按住的工作状态。细胞为此花掉的力气,买到的是一生的可用性。

第一个撞上这堵墙的是病毒

基因治疗最早接触到这份静默,是以一种很实际的方式:载体进不去。第一代临床载体是 γ-retroviral vector,而它需要细胞真正分裂。1993 年,Roe 等人在 EMBO J 上用同步化细胞证明,Moloney murine leukemia virus 的 DNA 只在细胞穿过有丝分裂之后才整合;他们提出这种依赖可能源于整合复合物要借核膜解体才能进核。1998 年,Uchida 等人在 PNAS 上把两类载体摆在一起比:对从动员外周血里新鲜分离、几乎全部处于 G0/G1 的人造血干细胞,HIV-1 来源的 lentiviral vector 能一步完成转导且细胞保持原有表型,MuLV 载体则做不到——除非先在细胞因子里养上三天。

但”不需要有丝分裂”和”在 G0 里效率一样高”,是两回事。次年,Sutton 等人在 J Virol 上把细胞按周期分选后逐一转导,发现 G1 或 S/G2/M 的细胞都比 G0 细胞更容易被转导;如果干脆把细胞因子从培养基里拿掉、让细胞留在 G0,转导率会下降最多十倍,阻断的位置大约就在 reverse transcription 起始那一步。lentiviral vector 绕开的是有丝分裂这道门槛,不是静息本身——它只把”必须分裂”放宽成了”至少要醒过来一点”。

叫醒它,就会失去它

于是整个领域走上一条注定要付代价的路:先用细胞因子唤醒细胞,再做基因操作。代价有多大,2000 年 Glimm、Oh 与 Eaves 在 Blood 上量得很清楚。他们把人脐带血细胞放进含多种细胞因子的无血清培养,到第 5 天检测——能在免疫缺陷小鼠体内重建造血的活性只保留在 G1 部分;而同一批培养里的集落形成细胞和长期培养起始细胞,在 G0/G1 与 S/G2/M 两部分中大致均分。更棘手的是那篇论文标题的后半句:在这套体外培养体系里,被推入周期的细胞不会重新回到 G0。

这和前面那条体内规律并不冲突,只是适用范围不同:在体内,被损伤或刺激活化的干细胞在稳态恢复后能回到 dormancy;而在五六天的 ex vivo 扩增培养里,这条退路是关着的。它几乎解释了 ex vivo 基因治疗工艺里所有看似琐碎的参数:培养时间买来的是可编辑性,花掉的是 engraftment。2017 年 Zonari 等人在 Stem Cell Reports 上的方案就照着这个矛盾设计——在人动员外周血来源的造血干细胞里用 PGE2 刺激,可以在不到 38 小时的培养里让 lentiviral vector 的转导接近完全,减轻标准培养对祖细胞功能的损害。工程上的进步,几乎都发生在”少叫醒一点、少叫醒一会儿”这个方向上。

编辑比转导更挑周期

到了基因编辑时代,这道关反而更紧,原因在于修复通路:homology-directed repair 主要限于 S/G2 期。2003 年,Rothkamm 等人在 Mol Cell Biol 上用 γ-H2AX foci,在仓鼠 CHO 突变株里逐期比较两条通路对电离辐射所致 DSB 的贡献:同源重组缺陷的细胞在 G1 只有轻微的修复缺陷,S 期受损更明显,到晚 S/G2 才出现严重缺陷——HR 在晚 S/G2 特别重要。它是主导,不是唯一;而静息细胞按定义就不在那里。

2014 年,Genovese 等人在 Nature 上把这件事说得很直接:在人造血干细胞里做定点整合,受制于两个障碍——细胞对基因转移的许可性低,HDR 通路的能力有限。他们靠重新搭配递送平台与培养条件跨过这两道坎,用移植小鼠的长期多谱系重建证明确有定点整合发生,并把一段矫正性 cDNA 敲进 SCID-X1 患者细胞的 IL2RG 位点。

跨过去之后,新的账单跟着来了。2019 年 Schiroli 等人在 Cell Stem Cell 上发现,哪怕只诱导一个 double-strand break,p53 介导的 DNA 损伤反应都是各造血干/祖细胞亚群里最主要的应答;DSB 负荷升高或用 AAV 递送修复模板时,激活会累加,限制编辑细胞的增殖、产量与 engraftment;好在负荷较低时损伤可逆。2020 年,Ferrari 等人在 Nat Biotechnol 上用条形码克隆追踪把后果量化:编辑激活 p53 会显著收缩移植后的干细胞克隆库,短暂抑制 p53 则能恢复多克隆性;而为了把 HDR 效率提上去,他们用腺病毒的 E4orf6/7 蛋白强行推动细胞周期并上调 HDR 机器,两者相加,在小鼠体内的长期人源移植物中把 HDR 效率做到了最高约 50%。

留意这套组合在做什么:一边强行把静息细胞推进周期,一边压住细胞因此拉响的警报——我们最想要的那种细胞,正是最不愿意配合这两件事的细胞。

另一头也卡在静息上

静息设的关卡还有第二道,在患者体内。要让改造过的细胞长期存活,骨髓里得先腾出位置,而占着位置的宿主干细胞同样静息、同样对针对增殖细胞的药物不敏感——这也是清髓方案至今仍要动用高剂量烷化剂的原因之一。绕开细胞毒性的思路很早有人试过:2007 年 Czechowicz 等人在 Science 上用阻断 c-kit 功能的抗体 ACK2 处理免疫缺陷小鼠,一过性清除了超过 98% 的内源干细胞,随后移植的嵌合率可达 90%;作者只说外推到人或许能带来温和而有效的预处理方案。

同一个生物学事实,于是在流程的两端各卡了一次:体外,静息让我们的细胞难以被改写;体内,静息让别人的细胞难以被请走。

一份不肯被打扰的工作

回到开头那个画面。骨髓里那些几个月才动一次的细胞,不是在等待被使用,而是在执行一项需要持续投入的任务:把自己保持在不被消耗的状态,以便在几十年的尺度上仍然可用。基因治疗要做的,本质上是在几十个小时里说服它临时放弃这份保守。

正在被推进的 in vivo 路线会把这个矛盾推到更极端的位置。这不是说体内的干细胞碰不得:Wilson 那篇论文里的 dormant 干细胞就能被骨髓损伤或 G-CSF 高效激活,Takizawa 那篇里注射细菌成分 lipopolysaccharide 也提高了干细胞的增殖与自我更新能力。真正失去的是控制力——ex vivo 那种逐个分选、定制细胞因子鸡尾酒、按小时计时的操作,在体内不复存在;递送系统再精准,最终仍要在一个多半处于 G0 的细胞内部完成工作。

所以要看懂造血干细胞基因治疗为什么这么难、这么贵,不必从载体或编辑器开始:它要修改的,是一群把”不动”当成本职工作、并且为此付了很多年薪水的细胞。

参考文献

  1. Cheshier SH, Morrison SJ, Liao X, Weissman IL. In vivo proliferation and cell cycle kinetics of long-term self-renewing hematopoietic stem cells. Proc Natl Acad Sci U S A. 1999;96(6):3120-5. DOI
  2. Wilson A, Laurenti E, Oser G, et al. Hematopoietic stem cells reversibly switch from dormancy to self-renewal during homeostasis and repair. Cell. 2008;135(6):1118-29. DOI
  3. Catlin SN, Busque L, Gale RE, Guttorp P, Abkowitz JL. The replication rate of human hematopoietic stem cells in vivo. Blood. 2011;117(17):4460-6. DOI
  4. Takizawa H, Regoes RR, Boddupalli CS, Bonhoeffer S, Manz MG. Dynamic variation in cycling of hematopoietic stem cells in steady state and inflammation. J Exp Med. 2011;208(2):273-84. DOI
  5. Matsumoto A, Takeishi S, Kanie T, et al. p57 is required for quiescence and maintenance of adult hematopoietic stem cells. Cell Stem Cell. 2011;9(3):262-71. DOI
  6. Yilmaz ÖH, Valdez R, Theisen BK, et al. Pten dependence distinguishes haematopoietic stem cells from leukaemia-initiating cells. Nature. 2006;441(7092):475-82. DOI
  7. Tothova Z, Kollipara R, Huntly BJ, et al. FoxOs are critical mediators of hematopoietic stem cell resistance to physiologic oxidative stress. Cell. 2007;128(2):325-39. DOI
  8. Takubo K, Nagamatsu G, Kobayashi CI, et al. Regulation of glycolysis by Pdk functions as a metabolic checkpoint for cell cycle quiescence in hematopoietic stem cells. Cell Stem Cell. 2013;12(1):49-61. DOI
  9. Signer RAJ, Magee JA, Salic A, Morrison SJ. Haematopoietic stem cells require a highly regulated protein synthesis rate. Nature. 2014;509(7498):49-54. DOI
  10. Warr MR, Binnewies M, Flach J, et al. FOXO3A directs a protective autophagy program in haematopoietic stem cells. Nature. 2013;494(7437):323-7. DOI
  11. Ho TT, Warr MR, Adelman ER, et al. Autophagy maintains the metabolism and function of young and old stem cells. Nature. 2017;543(7644):205-210. DOI
  12. Cabezas-Wallscheid N, Buettner F, Sommerkamp P, et al. Vitamin A-retinoic acid signaling regulates hematopoietic stem cell dormancy. Cell. 2017;169(5):807-823.e19. DOI
  13. Roe T, Reynolds TC, Yu G, Brown PO. Integration of murine leukemia virus DNA depends on mitosis. EMBO J. 1993;12(5):2099-108. DOI
  14. Uchida N, Sutton RE, Friera AM, et al. HIV, but not murine leukemia virus, vectors mediate high efficiency gene transfer into freshly isolated G0/G1 human hematopoietic stem cells. Proc Natl Acad Sci U S A. 1998;95(20):11939-44. DOI
  15. Sutton RE, Reitsma MJ, Uchida N, Brown PO. Transduction of human progenitor hematopoietic stem cells by human immunodeficiency virus type 1-based vectors is cell cycle dependent. J Virol. 1999;73(5):3649-60. DOI
  16. Glimm H, Oh IH, Eaves CJ. Human hematopoietic stem cells stimulated to proliferate in vitro lose engraftment potential during their S/G2/M transit and do not reenter G0. Blood. 2000;96(13):4185-93. DOI
  17. Zonari E, Desantis G, Petrillo C, et al. Efficient ex vivo engineering and expansion of highly purified human hematopoietic stem and progenitor cell populations for gene therapy. Stem Cell Reports. 2017;8(4):977-990. DOI
  18. Rothkamm K, Krüger I, Thompson LH, Löbrich M. Pathways of DNA double-strand break repair during the mammalian cell cycle. Mol Cell Biol. 2003;23(16):5706-15. DOI
  19. Genovese P, Schiroli G, Escobar G, et al. Targeted genome editing in human repopulating haematopoietic stem cells. Nature. 2014;510(7504):235-240. DOI
  20. Schiroli G, Conti A, Ferrari S, et al. Precise gene editing preserves hematopoietic stem cell function following transient p53-mediated DNA damage response. Cell Stem Cell. 2019;24(4):551-565.e8. DOI
  21. Ferrari S, Jacob A, Beretta S, et al. Efficient gene editing of human long-term hematopoietic stem cells validated by clonal tracking. Nat Biotechnol. 2020;38(11):1298-1308. DOI
  22. Czechowicz A, Kraft D, Weissman IL, Bhattacharya D. Efficient transplantation via antibody-based clearance of hematopoietic stem cell niches. Science. 2007;318(5854):1296-9. DOI

The human body makes millions of blood cells every second, a production line that never pauses for a day or for a lifetime. Yet the cells sitting at the very top of it do almost nothing — they wait quietly in the bone marrow, dividing once every few months, some of them not at all for a year. Intuition reads that as standby: low on energy, so sleep until needed. The evidence of the past two decades says the opposite — that doing nothing is the result of an entire molecular apparatus actively holding the cell down. And when we want to swap a piece of DNA into these cells, the first thing we have to negotiate with is exactly that carefully maintained silence.

First, count the slowness

To say what quiescence is, you first have to know how slow it actually is. In 1999, Cheshier, Weissman and colleagues published a direct count in PNAS: in normal adult mice, about 75% of long-term hematopoietic stem cells sit in G0 at any moment, and only about 5% are in S/G2/M. But BrdU tracking showed the pool was not frozen — roughly 8% enter the cell cycle asynchronously each day, and 99% had incorporated label by six months; which works out to a division on average every 57 days.

In 2008, Wilson and colleagues published in Cell a combination of two label-retaining approaches, BrdU and histone H2B-GFP, that isolated a population of dormant stem cells in mouse bone marrow; computational modeling suggested a division interval of about 145 days, roughly five divisions per lifetime, and these were precisely the cells harboring the vast majority of multilineage long-term self-renewal activity. You cannot give a person BrdU, so in 2011 Catlin and colleagues, writing in Blood, worked backwards instead from the age-related drift in the ratio of paternal to maternal X chromosome phenotypes in the blood cells of women, inferring that human hematopoietic stem cells replicate on average once every 40 weeks (range 25–50 weeks). That is inference rather than direct observation, but the order of magnitude is consistent with the mouse: years, not days.

Slow is also not the same as a small fixed group that never moves. Takizawa and colleagues, in J Exp Med in 2011, used CFSE to track division history and found that in steady-state mice the bone marrow cells capable of reconstituting lifelong hematopoiesis were found both in the fast-cycling fraction, undergoing five or more divisions in 7 weeks, and in the quiescent fraction, undergoing zero divisions in 12–14 weeks; and cells with extensive proliferative history were prone to return to quiescence. Quiescence is therefore not an identity but a state that can be entered and left — a point that matters a great deal later on.

Doing nothing takes effort

If quiescence were simply a shutdown caused by insufficient energy, interrupting it should cost nothing. The truth is the reverse: remove any one of the parts that maintain it and these cells slide into proliferation, and are then exhausted.

The most direct case is the brake itself. In 2011, Matsumoto and colleagues reported in Cell Stem Cell that among the Cip/Kip family, p57 (CDKN1C) is the most abundant in quiescent stem cells; mice lacking it showed a reduced fraction in G0 and a severe self-renewal defect, and knocking the p27 gene into the p57 locus corrected these abnormalities. The cost of releasing the brake has also been measured directly: in 2006, Yilmaz and colleagues deleted Pten in mice, reported in Nature, and the stem cells were driven into proliferation, producing a myeloproliferative disease within days and transplantable leukemia within weeks, while the normal stem cells were depleted through a cell-autonomous mechanism and could no longer stably reconstitute irradiated mice; the following year Tothova and colleagues, in Cell, deleted FoxO1, FoxO3 and FoxO4 together from the adult mouse hematopoietic system and likewise obtained a shrunken stem cell compartment with impaired long-term reconstituting ability. Releasing the brake does not buy more vigorous blood production, only a pool that burns out faster.

Metabolism and synthesis are held down just as actively. In 2013, Takubo and colleagues showed in Cell Stem Cell that mouse hematopoietic stem cells rely mainly on anaerobic glycolysis for energy, and that this depends on pyruvate dehydrogenase kinase actively restricting the flow of metabolites into mitochondria; deleting Pdk2 and Pdk4 caused quiescence, glycolysis and transplantation capacity to decline together. The following year, Signer and colleagues measured another suppressed variable in Nature: the amount of protein synthesized per hour by hematopoietic stem cells in vivo in mice was lower than in most other hematopoietic cells; and either lowering synthesis further or raising it impaired their function.

Clearance, too, is active. In 2013, Warr and colleagues found, in Nature, that in response to ex vivo cytokine withdrawal or in vivo calorie restriction, mouse hematopoietic stem cells robustly induce autophagy while their short-lived myeloid progeny do not; FOXO3A’s role is to keep the stem cells primed. In 2017, Ho and colleagues went further in Nature, showing that in mice autophagy suppresses metabolism by clearing active, healthy mitochondria, thereby maintaining quiescence and stemness. That same year, Cabezas-Wallscheid and colleagues used single-cell RNA-seq in Cell to draw a continuous map of the system: the transition from dormancy toward cell-cycle entry is a continuous path rather than a stepwise progression, associated with an upregulation of biosynthetic processes along the way; and mice maintained on a vitamin A-free diet lose stem cells and show a disrupted re-entry into dormancy after inflammatory stress — even the raw material of this silence has to be supplied from outside.

Quiescence, then, is not a default. It is a working state held down layer by layer by CDK inhibitors, transcription factors, metabolic pathways and autophagy. What the cell spends on it, it buys back as a lifetime of availability.

The first thing to hit this wall was a virus

Gene therapy first encountered this silence in a very practical way: the vector could not get in. The first generation of clinical vectors were γ-retroviral vectors, and they need the cell to actually divide. In 1993, Roe and colleagues showed in EMBO J, using synchronized cells, that Moloney murine leukemia virus DNA integrates only after cells traverse mitosis; they proposed that this dependence may be due to a requirement for nuclear envelope breakdown for the integration complex to enter the nucleus. In 1998, Uchida and colleagues put the two classes of vector side by side in PNAS: on human hematopoietic stem cells freshly isolated from mobilized peripheral blood, almost all of them in G0/G1, an HIV-1-derived lentiviral vector achieved transduction in a single step while the cells retained their original phenotype, whereas the MuLV vector could not — unless the cells were first cultured in cytokines for three days.

But not requiring mitosis and being just as efficient in G0 are two different things. The following year, Sutton and colleagues sorted cells by cycle phase and transduced them one fraction at a time in J Virol, finding that cells in G1 or in S/G2/M were both more readily transduced than G0 cells; and if the cytokines were removed from the medium altogether so that cells stayed in G0, transduction rates fell by up to 10-fold, with the block operating near the time of initiation of reverse transcription. What lentiviral vectors got around was the mitosis threshold, not quiescence itself — they only relaxed the requirement from having to divide to having to at least wake up a little.

Wake it up and you lose it

So the field walked down a road that was bound to cost something: use cytokines to wake the cells, then do the genetic work. Exactly how much it costs was measured cleanly by Glimm, Oh and Eaves in Blood in 2000. They placed human cord blood cells in serum-free culture containing multiple cytokines and assayed on day 5 — the activity capable of reconstituting hematopoiesis in immunodeficient mice was restricted to the G1 fraction, even though the colony-forming cells and long-term culture-initiating cells in the same cultures were approximately equally distributed between G0/G1 and S/G2/M. More awkward is the second half of that paper’s title: within this in vitro culture system, cells pushed into cycle do not reenter G0.

This does not conflict with the in vivo pattern described earlier; the scope is simply different. In vivo, stem cells activated by injury or stimulation can return to dormancy once homeostasis is re-established; in a five- or six-day ex vivo expansion culture, that exit is closed. It explains almost every seemingly trivial parameter in ex vivo gene therapy manufacturing: culture time buys you editability and spends your engraftment. The 2017 protocol from Zonari and colleagues in Stem Cell Reports was designed around exactly this tension — stimulating human mobilized peripheral blood hematopoietic stem cells with PGE2 allows near-complete lentiviral vector transduction during a culture time of less than 38 hours, mitigating the negative impact of standard culture on progenitor cell function. Almost all the engineering progress has happened in the direction of waking them less, and for less time.

Editing is fussier about the cycle than transduction is

In the gene editing era this barrier got tighter rather than looser, because of the repair pathway: homology-directed repair is largely confined to S/G2. In 2003, Rothkamm and colleagues used γ-H2AX foci in Mol Cell Biol to compare, in hamster CHO mutant cells, how the two pathways handle ionizing radiation-induced breaks phase by phase: homologous recombination-defective cells had a minor repair defect in G1, greater impairment in S, and a substantial defect in late S/G2 — HR is particularly important in late S/G2. It is dominant, not exclusive; and a quiescent cell is by definition not there.

In 2014, Genovese and colleagues put it plainly in Nature: targeted integration in human hematopoietic stem cells is constrained by two obstacles — the cells’ low permissiveness to gene transfer, and the limited capacity of the HDR pathway. By reconfiguring the delivery platform and culture conditions they crossed both, demonstrated genuine targeted integration through long-term multilineage reconstitution in transplanted mice, and knocked a corrective cDNA into the IL2RG locus of cells from SCID-X1 patients.

Having crossed it, a new bill arrived. In 2019, Schiroli and colleagues found in Cell Stem Cell that even inducing a single double-strand break made the p53-mediated DNA damage response the predominant response across hematopoietic stem and progenitor subpopulations; raising the DSB load or delivering the repair template with AAV made this activation additive, limiting the proliferation, yield and engraftment of edited cells; fortunately, at lower loads the damage is reversible. In 2020, Ferrari and colleagues quantified the consequences in Nat Biotechnol using barcoded clonal tracking: editing-induced p53 activation substantially contracted the stem cell clonal repertoire after transplantation, while transient p53 inhibition restored polyclonality; and to raise HDR efficiency they used the adenoviral E4orf6/7 protein to force cell-cycle progression and upregulate the HDR machinery, the two together achieving HDR editing efficiencies of up to 50% in the long-term human graft in mice.

Notice what this combination is doing: forcing quiescent cells into cycle on one side, and suppressing the alarm the cell raises in response on the other — the cells we want most are precisely the cells least willing to cooperate with either of those two things.

The other end is stuck on quiescence too

Quiescence sets a second checkpoint, inside the patient. For engineered cells to survive long term, room has to be cleared in the marrow first, and the host stem cells occupying that room are equally quiescent and equally insensitive to drugs aimed at dividing cells — one reason myeloablative regimens still call on high-dose alkylating agents. Attempts to get around the cytotoxicity came early: in 2007, Czechowicz and colleagues reported in Science that treating immunodeficient mice with ACK2, an antibody that blocks c-kit function, transiently removed more than 98% of endogenous stem cells, after which transplantation reached chimerism levels of up to 90%; the authors said only that extrapolation of these methods to humans may enable mild but effective conditioning regimens.

One biological fact thus blocks the pipeline at both ends: outside the body, quiescence makes our cells hard to rewrite; inside the body, it makes someone else’s cells hard to evict.

A job that refuses to be interrupted

Return to the opening image. Those cells in the marrow that stir once every few months are not waiting to be used; they are carrying out a task that requires continuous investment — keeping themselves in an unconsumed state so that they remain usable on a scale of decades. What gene therapy has to do is essentially to persuade them, over a few dozen hours, to set that conservatism aside.

The in vivo approaches now being pushed forward take this tension to a more extreme place. That is not to say stem cells in the body cannot be moved: the dormant stem cells in Wilson’s paper were efficiently activated to self-renew in response to bone marrow injury or G-CSF stimulation, and in Takizawa’s paper injection of the bacterial component lipopolysaccharide increased the proliferation and self-renewal capacity of HSCs. What is genuinely lost is control — the ex vivo practice of sorting cell by cell, tailoring a cytokine cocktail, timing everything by the hour, does not exist inside a body; and however precise the delivery system, the work still has to be finished inside a cell that is most likely in G0.

So to understand why hematopoietic stem cell gene therapy is this hard and this expensive, you do not have to start with vectors or editors: what it sets out to modify is a population of cells whose actual job is to stay still, and which have been drawing a salary for that for many years.

References

  1. Cheshier SH, Morrison SJ, Liao X, Weissman IL. In vivo proliferation and cell cycle kinetics of long-term self-renewing hematopoietic stem cells. Proc Natl Acad Sci U S A. 1999;96(6):3120-5. DOI
  2. Wilson A, Laurenti E, Oser G, et al. Hematopoietic stem cells reversibly switch from dormancy to self-renewal during homeostasis and repair. Cell. 2008;135(6):1118-29. DOI
  3. Catlin SN, Busque L, Gale RE, Guttorp P, Abkowitz JL. The replication rate of human hematopoietic stem cells in vivo. Blood. 2011;117(17):4460-6. DOI
  4. Takizawa H, Regoes RR, Boddupalli CS, Bonhoeffer S, Manz MG. Dynamic variation in cycling of hematopoietic stem cells in steady state and inflammation. J Exp Med. 2011;208(2):273-84. DOI
  5. Matsumoto A, Takeishi S, Kanie T, et al. p57 is required for quiescence and maintenance of adult hematopoietic stem cells. Cell Stem Cell. 2011;9(3):262-71. DOI
  6. Yilmaz ÖH, Valdez R, Theisen BK, et al. Pten dependence distinguishes haematopoietic stem cells from leukaemia-initiating cells. Nature. 2006;441(7092):475-82. DOI
  7. Tothova Z, Kollipara R, Huntly BJ, et al. FoxOs are critical mediators of hematopoietic stem cell resistance to physiologic oxidative stress. Cell. 2007;128(2):325-39. DOI
  8. Takubo K, Nagamatsu G, Kobayashi CI, et al. Regulation of glycolysis by Pdk functions as a metabolic checkpoint for cell cycle quiescence in hematopoietic stem cells. Cell Stem Cell. 2013;12(1):49-61. DOI
  9. Signer RAJ, Magee JA, Salic A, Morrison SJ. Haematopoietic stem cells require a highly regulated protein synthesis rate. Nature. 2014;509(7498):49-54. DOI
  10. Warr MR, Binnewies M, Flach J, et al. FOXO3A directs a protective autophagy program in haematopoietic stem cells. Nature. 2013;494(7437):323-7. DOI
  11. Ho TT, Warr MR, Adelman ER, et al. Autophagy maintains the metabolism and function of young and old stem cells. Nature. 2017;543(7644):205-210. DOI
  12. Cabezas-Wallscheid N, Buettner F, Sommerkamp P, et al. Vitamin A-retinoic acid signaling regulates hematopoietic stem cell dormancy. Cell. 2017;169(5):807-823.e19. DOI
  13. Roe T, Reynolds TC, Yu G, Brown PO. Integration of murine leukemia virus DNA depends on mitosis. EMBO J. 1993;12(5):2099-108. DOI
  14. Uchida N, Sutton RE, Friera AM, et al. HIV, but not murine leukemia virus, vectors mediate high efficiency gene transfer into freshly isolated G0/G1 human hematopoietic stem cells. Proc Natl Acad Sci U S A. 1998;95(20):11939-44. DOI
  15. Sutton RE, Reitsma MJ, Uchida N, Brown PO. Transduction of human progenitor hematopoietic stem cells by human immunodeficiency virus type 1-based vectors is cell cycle dependent. J Virol. 1999;73(5):3649-60. DOI
  16. Glimm H, Oh IH, Eaves CJ. Human hematopoietic stem cells stimulated to proliferate in vitro lose engraftment potential during their S/G2/M transit and do not reenter G0. Blood. 2000;96(13):4185-93. DOI
  17. Zonari E, Desantis G, Petrillo C, et al. Efficient ex vivo engineering and expansion of highly purified human hematopoietic stem and progenitor cell populations for gene therapy. Stem Cell Reports. 2017;8(4):977-990. DOI
  18. Rothkamm K, Krüger I, Thompson LH, Löbrich M. Pathways of DNA double-strand break repair during the mammalian cell cycle. Mol Cell Biol. 2003;23(16):5706-15. DOI
  19. Genovese P, Schiroli G, Escobar G, et al. Targeted genome editing in human repopulating haematopoietic stem cells. Nature. 2014;510(7504):235-240. DOI
  20. Schiroli G, Conti A, Ferrari S, et al. Precise gene editing preserves hematopoietic stem cell function following transient p53-mediated DNA damage response. Cell Stem Cell. 2019;24(4):551-565.e8. DOI
  21. Ferrari S, Jacob A, Beretta S, et al. Efficient gene editing of human long-term hematopoietic stem cells validated by clonal tracking. Nat Biotechnol. 2020;38(11):1298-1308. DOI
  22. Czechowicz A, Kraft D, Weissman IL, Bhattacharya D. Efficient transplantation via antibody-based clearance of hematopoietic stem cell niches. Science. 2007;318(5854):1296-9. DOI