Yang Liu

← Primer

HSC · Healing Source Code

慢病毒/逆转录病毒——整合的力量与基因毒性的原罪 Integrating Vectors Cured Most of the Bubble Boys — and Then Gave Some of Them Leukemia

2026-07-31 · 慢病毒载体, γ逆转录病毒, 插入突变, LMO2, 自失活载体, X-SCID

1999 到 2002 年间,巴黎内克尔医院和伦敦大奥蒙德街医院的医生们,把一份份被病毒改写过的骨髓细胞,分别输回到近二十个患有 X 连锁重症联合免疫缺陷(X-SCID)的婴儿体内。这些孩子生来没有功能性免疫系统,通常活不过婴儿期,只能被养在无菌帐篷里,俗称”泡泡男孩”。这一次,他们大多数被治好了——免疫系统重建,能像普通孩子一样出门。几年后,仅巴黎这一队里,九名接受治疗的孩子中就有四人患上了白血病,一个孩子因此去世。治愈他们的那份礼物和伤害他们的那把刀,是同一件东西。

这就是整合型病毒载体绕不开的悖论。它之所以能治病,是因为它把治疗基因永久地写进了细胞的染色体——写进去之后,这份指令会随着细胞每一次分裂被复制、传给所有子代,一次注射管一辈子。但”永久写入”这四个字反过来意味着:病毒把自己安插进基因组的哪一个位置,不再是无关紧要的细节,而直接决定了这份礼物旁边坐着的,究竟是空地,还是一颗定时炸弹。

从慢到快:两种病毒,两套规则

在造血干细胞基因治疗的历史上,先出场的是 γ 逆转录病毒(gammaretrovirus)——以小鼠白血病病毒(MLV)为骨架改造而成。它结构简单、滴度容易做高,90 年代末就被用来给 ADA-SCID、X-SCID 患者的 CD34⁺ 造血干细胞转入正常基因。但它有一个致命的生物学限制:MLV 的核膜穿越能力很弱,必须等到细胞分裂、核膜短暂溶解时才能把病毒基因组塞进细胞核。而造血干细胞恰恰以静息著称——大多数时候按兵不动,这是它们维持长期重建能力的代价(这道难关在讲 HSC 静息生物学时已经细说过)。逼着静息的干细胞在培养皿里分裂,本身就会消耗它们最宝贵的资本。

1996 年,萨尔克研究所的 Naldini 等人在 Science 上给出了另一条路:把一种以 HIV 为骨架、去除了全部致病基因的慢病毒载体(lentiviral vector)递送进静止的 HeLa 细胞、周期阻滞的成纤维细胞,乃至终末分化的神经元,均实现了稳定转导。慢病毒之所以能做到这一点,是因为它的整合前复合物自带核定位信号,能主动穿过完整的核膜——不需要细胞分裂配合。这个特性,恰好精准地对上了造血干细胞的软肋:它让基因治疗第一次有机会在不强迫干细胞分裂的前提下,把基因稳稳地写进去。

一次警报,和它揭开的机制

X-SCID 试验最初看起来是全面成功。2000 年,Cavazzana-Calvo 等人在 Science 上报告,用 γ 逆转录病毒把正常的 γc 受体基因转入患儿自体 CD34⁺ 细胞后回输,T 细胞和 NK 细胞功能相继重建。但到 2002 年底,最年轻的两名患儿出现了失控的克隆性 T 细胞增殖。2003 年,Hacein-Bey-Abina 等人在 Science 上揭晓了原因:两个患儿的白血病克隆里,病毒载体都插在了原癌基因 LMO2 的启动子附近,导致它被异常激活。病毒本身没有携带任何致癌基因——它只是恰好落在了一个不该落的位置上,像一根导火索,接上了细胞自身的一个开关。

这不是运气不好那么简单。同年,Wu 等人在 Science 上系统比对了不同病毒的整合偏好,发现 MLV 基础的 γ 逆转录病毒载体格外偏爱落在基因的转录起始区附近——恰恰是启动子和增强子最密集、对基因表达影响最大的地带。而这份偏好背后还有一层更深的原罪:γ 逆转录病毒的长末端重复序列(LTR)里,U3 区自带一段完整的病毒增强子/启动子,专门用来驱动病毒基因表达。当整个病毒——连同这段增强子——恰好落进一个原癌基因附近,这段增强子就会反过来”跨界”点燃邻近的宿主基因。2008 年,Howe 等人在 J Clin Invest 上报告了伦敦(大奥蒙德街医院)队列十名患儿的随访数据,进一步描绘了具体的作案细节:在一名患儿的白血病克隆中,载体以反义方向插入 LMO2 上游约 35 千碱基处,足以让它被显著过表达,而完整的白血病转化还需要 NOTCH1 突变、CDKN2A 缺失等额外的体细胞事件叠加。而在最早报告的巴黎队列里,2010 年 Hacein-Bey-Abina 等人在 N Engl J Med 上给出的随访结果是:九名接受治疗的患儿中,四人发展为急性白血病,一人因此去世——这是整合型载体第一次让全世界看清:整合的力量和它的代价,是一体两面。

拆掉那根导火索

问题一旦被定位到 LTR 里那段病毒增强子,解法就变得直接:把它拆掉。1998 年,Zufferey 等人在 J Virol 上描述了自失活(self-inactivating,SIN)载体的设计——在 3′ LTR 的 U3 区删去约 400 个核苷酸,包括 TATA 盒本身。病毒颗粒在生产细胞里仍能正常包装,但转导进入靶细胞、完成逆转录之后,两端的 LTR 都会失去启动子/增强子活性,不再有能力去”点燃”周围的基因。这个改动没有牺牲滴度或表达效率,却从根子上拔掉了 LMO2 事故里那根导火索。

SIN 设计推广开之后,一个更细的问题浮现出来:慢病毒和 γ 逆转录病毒,即便都做成 SIN,安全性也并不对等。慢病毒偏爱整合进活跃转录基因的内部区域,而不是聚集在转录起始位点附近;2009 年,Modlich 等人在 Mol Ther 上用体外永生化实验直接比较两者,经载体拷贝数校正后,SIN 慢病毒载体的插入模式触发细胞转化的概率,大约只有 SIN γ 逆转录病毒载体的三分之一——换句话说,慢病毒的整合偏好本身,就自带一层额外的安全冗余。这也是为什么”整合的力量”这句话,从 90 年代到今天,武器换了一代,但规则始终没变:不是要不要整合,而是整合会落在哪、以及落进去之后旁边还挂不挂着一段病毒自己的启动子。

临床数据后来验证了这套修复逻辑,但验证得很克制。2014 年,Hacein-Bey-Abina 等人在 N Engl J Med 上报告了改用增强子缺失型 SIN γ 逆转录病毒载体的新一代 X-SCID 试验:九名患儿中位随访 29.1 个月(range 12.1–38.7 个月),论文付印期间随访进一步延长,更新为中位 33 个月(range 16–43 个月)。截至彼时,没有一例患儿发生白血病,插入位点在 LMO2、MECOM 等原癌基因附近的聚集,也显著少于前一代载体队列——U3 增强子,正是当年那个真正的祸首。但作者自己也留了一句克制的限定:这种疗法长期致白血病的效应,仍属未知。与此同时,慢病毒平台也迅速走向了脑病和血红蛋白病:2009 年,Cartier 等人在 Science 上首次证明,用慢病毒把功能性 ABCD1 基因转入自体造血干细胞,可以延缓脑型肾上腺脑白质营养不良(cerebral ALD)患儿的神经系统病变进展——这项工作后来发展成了获批疗法 Skysona(eli-cel)。

走到药架上,和一道没关上的门

今天,SIN 慢病毒载体已经是 ex vivo 造血干细胞基因治疗的主力平台:治疗 β-地中海贫血的 Zynteglo(beti-cel)、治疗镰状细胞病的 Lyfgenia(lovo-cel)、以及前面提到的 Skysona,都建立在这套设计之上;更早的 ADA-SCID 疗法 Strimvelis 则沿用了 γ 逆转录病毒路线,2016 年成为欧洲首个获批上市的离体基因疗法。整合型载体从一场几乎让整个领域停摆的事故里,走回了药架。

但这道门并没有被彻底关上。Skysona 的临床试验随访数据持续被 FDA 更新:2022 年获批时,67 名受试者中有 3 人(约 4%)出现血液系统恶性肿瘤;随着随访延长,这一数字后续增至 10 人(约 15%),发病时间从给药后 14 个月到 10 年不等。美国食品药品监督管理局(FDA)因此先后发出安全通告,并在标签中把适应症收紧为”无 HLA 相合供者可用”的患儿。相比之下,同样基于 SIN 慢病毒平台的 Zynteglo 和 Lyfgenia,目前尚未报告经证实由载体插入驱动的恶性肿瘤病例,尽管说明书上仍保留着相应的警示语言。同样是 SIN 慢病毒平台,为什么风险会在不同疾病、不同靶细胞群体之间显出差异——是清髓预处理的协同效应,还是脑白质营养不良患者本身的骨髓背景,抑或载体拷贝数与内部启动子强度的差别——目前还没有定论,是这类疗法长期安全性追踪里最需要盯住的一道未解之题。

回头看,整合型病毒载体教会这个领域的,其实是一件很朴素的事:能把基因永久焊进染色体,是它最大的资产,也是它唯一的原罪来源。从 γ 逆转录病毒到慢病毒,从带着完整增强子的 LTR 到自失活设计,每一步改进本质上都是同一个动作——不去否定整合本身,而是不断收窄它可能落错地方的概率。这条思路后来也启发了整个递送领域去追问一个更根本的问题:如果连”落在哪”都不必去赌,是不是可以干脆不整合、只让编辑器瞬时进出细胞——这条路,留到病毒样颗粒那一篇再细说。

参考文献

  1. Naldini L, et al. In vivo gene delivery and stable transduction of nondividing cells by a lentiviral vector. Science. 1996;272(5259):263-267. DOI
  2. Cavazzana-Calvo M, et al. Gene therapy of human severe combined immunodeficiency (SCID)-X1 disease. Science. 2000;288(5466):669-672. DOI
  3. Hacein-Bey-Abina S, et al. LMO2-associated clonal T cell proliferation in two patients after gene therapy for SCID-X1. Science. 2003;302(5644):415-419. DOI
  4. Wu X, Li Y, Crise B, Burgess SM. Transcription start regions in the human genome are favored targets for MLV integration. Science. 2003;300(5626):1749-1751. DOI
  5. Zufferey R, et al. Self-inactivating lentivirus vector for safe and efficient in vivo gene delivery. J Virol. 1998;72(12):9873-9880. DOI
  6. Howe SJ, et al. Insertional mutagenesis combined with acquired somatic mutations causes leukemogenesis following gene therapy of SCID-X1 patients. J Clin Invest. 2008;118(9):3143-3150. DOI
  7. Hacein-Bey-Abina S, et al. Efficacy of gene therapy for X-linked severe combined immunodeficiency. N Engl J Med. 2010;363(4):355-364. DOI
  8. Cartier N, et al. Hematopoietic stem cell gene therapy with a lentiviral vector in X-linked adrenoleukodystrophy. Science. 2009;326(5954):818-823. DOI
  9. Modlich U, et al. Insertional transformation of hematopoietic cells by self-inactivating lentiviral and gammaretroviral vectors. Mol Ther. 2009;17(11):1919-1928. DOI
  10. Hacein-Bey-Abina S, et al. A modified γ-retrovirus vector for X-linked severe combined immunodeficiency. N Engl J Med. 2014;371(15):1407-1417. DOI

Between 1999 and 2002, physicians at Necker Hospital in Paris and Great Ormond Street Hospital in London infused virus-rewritten bone marrow cells back into nearly twenty infants with X-linked severe combined immunodeficiency (X-SCID). These children were born without a functioning immune system and normally would not survive infancy; they were raised in sterile tents, popularly known as “bubble boys.” This time, most of them were cured — their immune systems reconstituted, and they could go outside like ordinary children. A few years later, in the Paris cohort alone, four of the nine treated children developed leukemia, and one of them died. The gift that cured them and the blade that hurt them were the same object.

This is the paradox that integrating viral vectors can never escape. They cure disease precisely because they write the therapeutic gene permanently into the cell’s chromosomes — once written in, the instruction is copied every time the cell divides and passed on to every daughter cell, one infusion for a lifetime. But those words, “permanently written in,” cut both ways: exactly where in the genome the virus inserts itself stops being an incidental detail and becomes the very thing that decides whether what sits next to that gift is empty ground or a ticking bomb.

From Slow to Fast: Two Viruses, Two Rulebooks

In the history of hematopoietic stem cell gene therapy, gammaretrovirus came first, built on the backbone of murine leukemia virus (MLV). It was structurally simple and easy to produce at high titer, and by the late 1990s it was already being used to transfer normal genes into the CD34+ hematopoietic stem cells of patients with ADA-SCID and X-SCID. But it carried one fatal biological limitation: MLV’s ability to cross the nuclear envelope is weak, and it can only smuggle its genome into the nucleus once the cell divides and the envelope briefly dissolves. Hematopoietic stem cells, notoriously, spend most of their time quiescent — standing down is the price they pay for keeping their long-term reconstituting capacity (a constraint already covered in the piece on HSC quiescence biology). Forcing a quiescent stem cell to divide in a dish already spends down its most precious capital.

In 1996, Naldini and colleagues at the Salk Institute, writing in Science, offered another route: they delivered a lentiviral vector built on an HIV backbone with all pathogenic genes removed into quiescent HeLa cells, cell-cycle-arrested fibroblasts, and even terminally differentiated neurons, achieving stable transduction in every case. What lets lentivirus do this is that its pre-integration complex carries its own nuclear localization signal and can actively cross an intact nuclear envelope, no cell division required. That property maps precisely onto hematopoietic stem cells’ weak point: for the first time, gene therapy had a way to write a gene stably into a stem cell without forcing it to divide.

An Alarm, and the Mechanism It Uncovered

The X-SCID trial looked, at first, like an unqualified success. In 2000, Cavazzana-Calvo and colleagues reported in Science that transferring the normal γc receptor gene into patients’ own CD34+ cells with a gammaretroviral vector, then reinfusing them, restored T-cell and NK-cell function in sequence. But by late 2002, the two youngest patients developed uncontrolled clonal T-cell proliferation. In 2003, Hacein-Bey-Abina and colleagues, in Science, revealed why: in both children’s leukemic clones, the vector had inserted near the promoter of the proto-oncogene LMO2, driving its abnormal activation. The virus itself carried no oncogene — it had simply landed in a place it should not have, like a fuse connecting to a switch the cell already had.

This was not simply bad luck. That same year, Wu and colleagues, in Science, systematically compared the integration preferences of different viruses and found that MLV-based gammaretroviral vectors disproportionately favored landing near transcription start sites — precisely the neighborhood most densely packed with promoters and enhancers, and most consequential for gene expression. Behind that preference sits a deeper original sin: the long terminal repeat (LTR) of a gammaretrovirus carries, in its U3 region, a complete viral enhancer/promoter dedicated to driving viral gene expression. When the whole virus — enhancer included — happens to land near a proto-oncogene, that enhancer can reach across and ignite the neighboring host gene. In 2008, Howe and colleagues, in the Journal of Clinical Investigation, reported follow-up on the London (Great Ormond Street) cohort of ten patients and filled in the mechanistic detail: in one patient’s leukemic clone, the vector had inserted in antisense orientation roughly 35 kilobases upstream of LMO2, enough to drive its significant overexpression, though full leukemic transformation still required additional somatic events layered on top — a NOTCH1 mutation, a CDKN2A deletion. And in the original Paris cohort, the follow-up that Hacein-Bey-Abina and colleagues reported in 2010 in the New England Journal of Medicine gave the number that made the whole field take notice: of nine treated children, four developed acute leukemia, and one died. This was the first time integrating vectors let the entire world see clearly that the power of integration and its cost are two sides of the same coin.

Pulling Out the Fuse

Once the problem had been localized to the viral enhancer sitting inside the LTR, the fix became direct: remove it. In 1998, Zufferey and colleagues, in the Journal of Virology, described the design of self-inactivating (SIN) vectors — deleting roughly 400 nucleotides from the U3 region of the 3′ LTR, including the TATA box itself. The viral particle still packages normally in producer cells, but once it transduces a target cell and completes reverse transcription, both resulting LTRs lose promoter/enhancer activity and can no longer “ignite” neighboring genes. The change cost nothing in titer or expression efficiency, yet it pulled out, at the root, the exact fuse behind the LMO2 accident.

Once the SIN design spread, a finer question surfaced: lentivirus and gammaretrovirus, even both made SIN, are not equally safe. Lentivirus prefers to integrate inside actively transcribed genes rather than clustering near transcription start sites; in 2009, Modlich and colleagues, in Molecular Therapy, compared the two directly with an in vitro immortalization assay and found that, after correcting for vector copy number, the insertion pattern of SIN lentiviral vectors triggered cellular transformation at roughly one-third the rate of SIN gammaretroviral vectors — in other words, lentivirus’s own integration preference carries a built-in margin of extra safety. This is also why “the power of integration” has meant the same thing from the 1990s to today even as the weapon changed generation: the question was never whether to integrate, but where it lands, and whether whatever lands next to it still drags along a viral promoter of its own.

Clinical data eventually validated this repair logic, though the validation came with restraint. In 2014, Hacein-Bey-Abina and colleagues, in the New England Journal of Medicine, reported a new generation of X-SCID trial using an enhancer-deleted SIN gammaretroviral vector: nine patients with a median follow-up of 29.1 months (range 12.1–38.7 months), extended by the time the paper went to press to a median of 33 months (range 16–43 months). By that point, no patient had developed leukemia, and insertion sites clustering near proto-oncogenes such as LMO2 and MECOM were also significantly reduced compared with the prior-generation vector cohort — the U3 enhancer had indeed been the real culprit. But the authors themselves left a careful qualifier: whether this therapy carries a long-term leukemogenic effect remains unknown. Around the same time, the lentiviral platform moved quickly toward cerebral disease and hemoglobinopathies: in 2009, Cartier and colleagues, in Science, first demonstrated that transferring a functional ABCD1 gene into autologous hematopoietic stem cells with a lentiviral vector could slow the progression of neurological disease in children with cerebral adrenoleukodystrophy (cerebral ALD) — work that later became the approved therapy Skysona (eli-cel).

Onto the Shelf, and a Door Left Ajar

Today, SIN lentiviral vectors are the workhorse platform of ex vivo hematopoietic stem cell gene therapy: Zynteglo (beti-cel) for beta-thalassemia, Lyfgenia (lovo-cel) for sickle cell disease, and Skysona, mentioned above, all rest on this design. The earlier ADA-SCID therapy Strimvelis instead followed the gammaretroviral route, becoming, in 2016, the first ex vivo gene therapy approved in Europe. Integrating vectors walked back onto the pharmacy shelf from an accident that had nearly stopped the entire field.

But that door was never fully shut. Skysona’s clinical trial follow-up data has kept being updated by the FDA: at approval in 2022, 3 of 67 subjects (about 4%) had developed hematologic malignancy; as follow-up extended, that number rose to 10 (about 15%), with onset ranging from 14 months to 10 years after dosing. The U.S. Food and Drug Administration (FDA) issued safety communications in response and tightened the label’s indication to boys with no available HLA-matched donor for allogeneic hematopoietic stem cell transplant. By comparison, Zynteglo and Lyfgenia, built on the same SIN lentiviral platform, have so far reported no confirmed cases of vector-insertion-driven malignancy, though their labels still carry the corresponding warning language. Why risk would diverge across diseases and target cell populations on the very same SIN lentiviral platform — whether it is a synergy with myeloablative conditioning, something in the marrow background specific to adrenoleukodystrophy patients, or a difference in vector copy number and internal promoter strength — has no settled answer yet, and remains the open question that most needs watching in the long-term safety tracking of this class of therapy.

Looking back, what integrating viral vectors taught this field is something quite plain: the ability to weld a gene permanently into the chromosome is their greatest asset, and also the sole source of their original sin. From gammaretrovirus to lentivirus, from an LTR carrying a complete enhancer to a self-inactivating design, every step of improvement has, at bottom, been the same move — not denying integration itself, but continually narrowing the odds that it lands in the wrong place. That line of thinking later pushed the whole delivery field to ask a more fundamental question: if you don’t have to gamble on “where it lands” at all, could you skip integration altogether and just let the editor pass in and out of the cell transiently — a road we’ll pick up in the piece on virus-like particles.

References

  1. Naldini L, et al. In vivo gene delivery and stable transduction of nondividing cells by a lentiviral vector. Science. 1996;272(5259):263-267. DOI
  2. Cavazzana-Calvo M, et al. Gene therapy of human severe combined immunodeficiency (SCID)-X1 disease. Science. 2000;288(5466):669-672. DOI
  3. Hacein-Bey-Abina S, et al. LMO2-associated clonal T cell proliferation in two patients after gene therapy for SCID-X1. Science. 2003;302(5644):415-419. DOI
  4. Wu X, Li Y, Crise B, Burgess SM. Transcription start regions in the human genome are favored targets for MLV integration. Science. 2003;300(5626):1749-1751. DOI
  5. Zufferey R, et al. Self-inactivating lentivirus vector for safe and efficient in vivo gene delivery. J Virol. 1998;72(12):9873-9880. DOI
  6. Howe SJ, et al. Insertional mutagenesis combined with acquired somatic mutations causes leukemogenesis following gene therapy of SCID-X1 patients. J Clin Invest. 2008;118(9):3143-3150. DOI
  7. Hacein-Bey-Abina S, et al. Efficacy of gene therapy for X-linked severe combined immunodeficiency. N Engl J Med. 2010;363(4):355-364. DOI
  8. Cartier N, et al. Hematopoietic stem cell gene therapy with a lentiviral vector in X-linked adrenoleukodystrophy. Science. 2009;326(5954):818-823. DOI
  9. Modlich U, et al. Insertional transformation of hematopoietic cells by self-inactivating lentiviral and gammaretroviral vectors. Mol Ther. 2009;17(11):1919-1928. DOI
  10. Hacein-Bey-Abina S, et al. A modified γ-retrovirus vector for X-linked severe combined immunodeficiency. N Engl J Med. 2014;371(15):1407-1417. DOI