Yang Liu

← Primer

HSC · Healing Source Code

同一个字母,三种改法——镰状细胞病的基因治疗路线之争 One Letter, Three Ways to Fix It — The Route Wars in Sickle Cell Gene Therapy

2026-07-12 · sickle-cell, 基因治疗, 血红蛋白病

镰状细胞病的全部祸根,只是基因组里一个字母的替换:β-globin 基因的第六个密码子,该是谷氨酸的地方错成了缬氨酸。就是这一个字母,让血红蛋白在缺氧时彼此串成僵硬的长纤维,把柔软的红细胞拧成镰刀的形状,堵住毛细血管,带来一生反复的剧痛与器官损伤。病因如此单纯,以至于它成了人类历史上第一个被称作”分子病”的疾病;可正因为它单纯,基因治疗才在同一个靶点上分出了三条互不相同的路——有人绕过这个坏字母,有人干脆不碰它、去唤醒另一套备份,还有人执意要把这个字母本身改回来。它们治的是同一种病,信奉的却是三种不同的哲学。

一个字母的病

早在 1949 年,Pauling 与 Itano 就在 Science 上用电泳把镰状细胞病人的血红蛋白和正常血红蛋白分了开来,第一次指出一种遗传病可以追溯到某个分子的异常,并给它起名”分子病”。八年后的 1957 年,Ingram 在 Nature 上把这个异常锁定到了一个氨基酸——正常与镰变血红蛋白之间,化学上只差这一处。这是分子医学的开篇,也为半个多世纪后的基因治疗预设好了靶心:所有人都知道要修的是哪里,分歧只在于”怎么修”。

那个致病的血红蛋白后来被记作 HbS。它在氧气充足时还算安分,一旦在组织深处卸下氧气,分子表面暴露的那个缬氨酸就像一个多出来的挂钩,把相邻的血红蛋白一个接一个勾连成纤维。红细胞被从内部撑变形、变脆,寿命骤减,又在血管里彼此挤堵——这就是那场反复发作、被称作血管闭塞危象的剧痛的由来。要终结它,理论上有三种下手的方式,而过去十年,这三种方式真的都走到了人体或接近人体的门口。

第一条路:再加一个好备份

最直觉的想法是:既然坏的 β-globin 改不掉，那就再塞进去一个好的。这条路不去触碰病人自己那份缺陷基因,而是借慢病毒(lentiviral)载体,把一段额外的、经过改造的 β-globin 基因永久地整合进造血干细胞的基因组里。改造的关键在于,这个外源基因编码的不是普通血红蛋白,而是一个带 T87Q 突变的抗镰变版本——它在血红蛋白纤维里插进来,就像往砖墙里掺进不合槽的砖,让 HbS 的长链搭不起来。

2017 年,Ribeil 等人在 New England Journal of Medicine 上报告了用这套方案治疗的第一位镰状细胞病患者:在清髓预处理之后回输经慢病毒转导的自体造血干细胞,治疗十五个月后,这个抗镰变血红蛋白稳定地占到了相当比例,不良反应主要来自清髓药物而非基因操作本身。到 2022 年,Kanter 等人在同一本杂志上给出了更成规模的证据:在更严格入组的一组 35 名患者里,回输后所有人都成功植入,总血红蛋白从治疗前的每分升 8.5 克回升到 6 个月起持续维持在 11 克以上,那个抗镰变血红蛋白贡献了至少四成、并且分布在平均约 85% 的红细胞里;在可评估的 25 名患者中,先前每年中位 3.5 次的严重血管闭塞危象全部消失。这套疗法就是后来的 lovo-cel(lovotibeglogene autotemcel,商品名 Lyfgenia)。

它的软肋恰恰藏在”整合”二字里。慢病毒把基因插进基因组的哪个位置并不完全受控,理论上存在扰动邻近基因、诱发恶性克隆的插入性风险。2021 年,一份令人揪心的报告出现:一位早期队列的患者在接受 LentiGlobin 治疗约 5 年半后患上了急性髓系白血病,白血病细胞里确实带着载体的插入位点。深入的溯源调查最终判断这次白血病不太可能由插入本身引起——插入位点的位置、极低的转基因表达、对周边基因表达的无影响都指向别处,而白血病细胞里另有若干本就易致白血病的体细胞突变;更可能的解释,是镰状细胞病本身、移植过程、以及治疗前控制不佳共同抬高了血液系统恶变的基础风险。无论归因如何,这个案例把慢病毒基因添加最需要长期盯防的那个问题,清清楚楚地摆到了台面上。

第二条路:不碰坏基因,唤醒旧备份

第二条路的思路截然不同:它根本不去修 HBB,也不加新的 β-globin,而是把每个人体内本就存在、却在出生后被关掉的另一套系统重新打开——胎儿血红蛋白(HbF)。胎儿时期我们用的是 γ-globin 组成的血红蛋白,它天生不参与 HbS 的聚合;人群里那些天生把较多 HbF 保留到成年的镰状细胞病人,病情往往轻得多。把这套沉睡的备份唤醒,就等于绕开了那个坏字母,却又不必替换它。

执行这一步的开关,是红系细胞里抑制 HbF 的转录因子 BCL11A——更准确地说,是它那段只在红系里起作用的enhancer。用 CRISPR-Cas9 在造血干细胞里精准剪断这段enhancer,就能在红细胞里把 BCL11A 调暗、让 γ-globin 重新表达。2021 年,Frangoul 等人在 New England Journal of Medicine 上首次报告这套编辑能让镰状细胞病和 β-地中海贫血患者摆脱输血依赖;到 2024 年,同一团队在该刊给出了镰状细胞病的关键试验结果:44 名患者接受治疗,在随访满一年以上、可评估的 30 人里,29 人(97%)连续至少 12 个月不再发作严重血管闭塞危象,全部 30 人都不再因此住院。这套疗法即 exa-cel(exagamglogene autotemcel,商品名 Casgevy),也是第一个获批上市的 CRISPR 基因编辑疗法。BCL11A 这个靶点本身的来龙去脉是另一篇的主角,这里要紧的是它代表的路线:换一套备份,而非修复原件。

这条路的代价写在它的工作原理里。CRISPR-Cas9 靠制造 DNA 的double-strand break来完成编辑,而断口的修复并不总是干净利落——大片段缺失、染色体层面的重排、以及被激活的 p53 反应,都是这类断裂潜在的副作用(这部分的安全账留到讲编辑安全性时再细算)。此外,唤醒的 HbF 到底够不够、在红细胞之间分布得是否均匀,直接决定了压制镰变的效果,而这既受个体生物学的限制,也不是一味把编辑效率调高就能线性解决的。

第三条路:把坏字母改回来

如果说前两条路一个是”绕开”、一个是”替换”,第三条路则最执拗:它要直接回到那个出错的字母,把它改掉。麻烦在于,经典的 CRISPR 要精确地把一个碱基换成另一个,得先制造断口再指望细胞按模板修复,效率低且风险高。base editing(碱基编辑)提供了另一种手法——它不切断双链,而是像化学橡皮擦一样,在原地把一个碱基直接转写成另一个。

真正巧妙的一步在于改的方向。把致病的缬氨酸密码子一步改回原本的谷氨酸,化学上并不顺手;但换个思路,只需一次腺嘌呤编辑,就能把 HbS 改写成一个叫 Makassar 的天然变体——这是一种在人群中存在、却不致病的良性 β-globin。2021 年,Newby 等人在 Nature 上用一个定制的腺嘌呤碱基编辑器(ABE8e-NRCH)做到了这件事:在镰状细胞病人的造血干细胞里,约八成的致病等位基因被就地改成了 Makassar 版本;把编辑后的人源细胞移植进小鼠,十六周后这一比例仍维持在近七成,缺氧诱导的镰变减少到原来的五分之一。在一个人源化的镰状细胞病小鼠模型里,Makassar 型 β-globin 占到了血液 β-globin 的约八成,镰变减少约三分之二,小鼠的血液学指标接近正常。更关键的是两个细节:只要有两成以上的等位基因被改成 Makassar,就足以逆转病理表型;而且因为全程不制造double-strand break,它避开了 Cas9 那种伴随的 p53 激活和大片段缺失。

不过,这条最优雅的路,眼下走得也最靠前又最靠后——优雅在机制,靠后在证据。前两条路都已经有获批上市的产品、有几十名随访多年的患者,而把 HbS 直接改成 Makassar 的这套方案,迄今扎实的疗效与安全数据仍主要来自细胞与小鼠,尚未走到同等成熟的人体阶段。它证明了原理,却还没证明它在人身上同样安全、持久、可扩展。

三条路,怎么摆在一起看

把三条路并排,分野就清楚了。慢病毒基因添加是”绕过病因”:坏基因原封不动,只是又加了一份抗镰变的好备份——它临床证据最厚,却把外源基因永久地掺进基因组,长期要盯防的是插入的不确定性。CRISPR 敲 BCL11A 是”换一套备份”:连坏基因都不碰,转而唤醒人体自带的胎儿血红蛋白,已成为第一个获批的 CRISPR 疗法——代价是它依赖双链断裂,以及唤醒的 HbF 够不够用这一生物学上限。base editing 改 Makassar 则是三者中最接近”纠正病因”的一条:它不留断口、直接把坏字母改成一个良性变体,理论上最干净——但它的人体证据链最短,前景与未知都最大。

值得一提的是,这三条路共享着同一套沉重的底盘:直到今天,它们仍都是一场完整的自体移植——需要动员、采集、体外操作、清髓预处理,费用高达百万美元级别,只有少数中心能做。技术上谁更”聪明”,并不直接等于谁更能抵达那些真正被镰状细胞病压垮的地区,而后者恰是全球绝大多数患者的所在(这笔”能治却治不到”的账,留到讲可及性时再摊开)。

从一个字母的错,到三种改法的分立,镰状细胞病这半个多世纪走过的,其实是基因治疗方法论的一次公开演武:绕过、替换、还是纠正,哪一种在什么代价下最值得。答案未必只有一个——不同的病情、不同的资源、不同的风险胃口,或许要的本就是不同的路。而当有一天,把坏字母改回来这件事能像今天的移植一样安全、普及,那个 1949 年就被看清的分子病,才算真正被从它出错的那一个字母上,治愈了。

参考文献

  1. Pauling L, Itano HA, Singer SJ, Wells IC. Sickle cell anemia, a molecular disease. Science. 1949;110(2865):543-8. DOI
  2. Ingram VM. Gene mutations in human haemoglobin: the chemical difference between normal and sickle cell haemoglobin. Nature. 1957;180(4581):326-8. DOI
  3. Ribeil JA, et al. Gene Therapy in a Patient with Sickle Cell Disease. N Engl J Med. 2017;376(9):848-855. DOI
  4. Kanter J, et al. Biologic and Clinical Efficacy of LentiGlobin for Sickle Cell Disease. N Engl J Med. 2022;386(7):617-628. DOI
  5. Goyal S, et al. Acute Myeloid Leukemia Case after Gene Therapy for Sickle Cell Disease. N Engl J Med. 2022;386(2):138-147. DOI
  6. Frangoul H, et al. CRISPR-Cas9 Gene Editing for Sickle Cell Disease and β-Thalassemia. N Engl J Med. 2021;384(3):252-260. DOI
  7. Frangoul H, et al. Exagamglogene Autotemcel for Severe Sickle Cell Disease. N Engl J Med. 2024;390(18):1649-1662. DOI
  8. Newby GA, et al. Base editing of haematopoietic stem cells rescues sickle cell disease in mice. Nature. 2021;595(7866):295-302. DOI

The entire calamity of sickle cell disease comes down to a single letter swapped in the genome: at the sixth codon of the β-globin gene, where there should be glutamate, there is valine instead. That one letter makes hemoglobin string itself into stiff, long fibers when oxygen runs low, twists soft red cells into the shape of a sickle, plugs the capillaries, and brings a lifetime of recurring agony and organ damage. The cause is so simple that it became the first disease in human history to be called a “molecular disease”; and precisely because it is so simple, gene therapy has split, at this very same target, into three mutually distinct roads — one bypasses the bad letter, one refuses to touch it and instead reawakens a backup system, and one insists on changing that letter itself back. They treat the same disease, but they profess three different philosophies.

A One-Letter Disease

As early as 1949, Pauling and Itano used electrophoresis in Science to separate the hemoglobin of sickle cell patients from normal hemoglobin, pointing out for the first time that a hereditary disease could be traced back to an abnormality in a single molecule, and giving it the name “molecular disease.” Eight years later, in 1957, Ingram in Nature pinned that abnormality down to a single amino acid — chemically, normal and sickle hemoglobin differ at just this one spot. This was the opening chapter of molecular medicine, and it set the bullseye for gene therapy half a century later: everyone knew where the repair had to be made; the only disagreement was over “how to make it.”

That pathogenic hemoglobin came to be written HbS. When oxygen is plentiful it stays fairly well-behaved, but the moment it unloads oxygen deep in the tissues, the valine exposed on the molecule’s surface acts like an extra hook, linking neighboring hemoglobin molecules one after another into fibers. Red cells are distended from within, turn brittle, see their lifespans plummet, and jam against one another in the vessels — and this is the origin of that recurring agony known as the vaso-occlusive crisis. To end it, there are, in theory, three ways to go about it, and over the past decade all three of these ways have genuinely arrived at, or come close to, the doorstep of human use.

The First Road: Add a Good Backup

The most intuitive idea is this: since the bad β-globin cannot be changed, then simply insert a good one. This road does not touch the patient’s own defective gene, but instead uses a lentiviral vector to permanently integrate an extra, engineered β-globin gene into the genome of the hematopoietic stem cells. The key to the engineering is that this foreign gene encodes not ordinary hemoglobin, but an anti-sickling version carrying the T87Q mutation — when it inserts itself into the hemoglobin fiber, it is like mixing an ill-fitting brick into a brick wall, so that the long chains of HbS cannot assemble.

In 2017, Ribeil and colleagues reported in the New England Journal of Medicine the first sickle cell patient treated with this approach: after myeloablative conditioning, autologous hematopoietic stem cells transduced by the lentivirus were reinfused, and fifteen months after treatment this anti-sickling hemoglobin stably accounted for a substantial proportion, with the adverse effects coming mainly from the myeloablative drugs rather than from the gene manipulation itself. By 2022, Kanter and colleagues in the same journal provided evidence at greater scale: in a group of 35 patients enrolled under stricter criteria, all engrafted successfully after reinfusion, total hemoglobin rose from 8.5 grams per deciliter before treatment to a sustained level above 11 grams from month 6 onward, that anti-sickling hemoglobin contributed at least 40% and was distributed across an average of roughly 85% of the red cells; and among the 25 evaluable patients, the previously median 3.5 severe vaso-occlusive crises per year vanished entirely. This therapy is what later became lovo-cel (lovotibeglogene autotemcel, trade name Lyfgenia).

Its weak spot is hidden precisely in that word “integration.” Which position in the genome the lentivirus inserts the gene into is not fully controlled, and there is, in theory, an insertional risk of perturbing neighboring genes and inducing a malignant clone. In 2021, a distressing report appeared: a patient from an early cohort developed acute myeloid leukemia about five and a half years after receiving LentiGlobin treatment, and the leukemic cells did indeed carry the vector’s insertion site. An in-depth investigation into the origin ultimately judged that this leukemia was unlikely to have been caused by the insertion itself — the location of the insertion site, the extremely low transgene expression, and the absence of any effect on the expression of surrounding genes all pointed elsewhere, while the leukemic cells separately carried several somatic mutations already known to predispose to leukemia; the more likely explanation is that sickle cell disease itself, the transplantation process, and poorly controlled disease before treatment together raised the baseline risk of malignant transformation in the blood system. However the attribution falls, this case laid squarely on the table the very problem that lentiviral gene addition most needs to keep watch on over the long term.

The Second Road: Don’t Touch the Bad Gene — Reawaken an Old Backup

The second road’s thinking is entirely different: it does not repair HBB at all, nor add a new β-globin, but instead reopens another system that exists in everyone from birth yet is switched off after birth — fetal hemoglobin (HbF). In fetal life we use hemoglobin made of γ-globin, which by nature does not take part in the polymerization of HbS; among the population, those sickle cell patients who naturally retain more HbF into adulthood tend to have much milder disease. To reawaken this dormant backup is to bypass the bad letter without having to replace it.

The switch that carries out this step is BCL11A, the transcription factor in erythroid cells that suppresses HbF — more precisely, the enhancer of BCL11A that acts only in the erythroid lineage. Using CRISPR-Cas9 to precisely cut this enhancer in hematopoietic stem cells dims BCL11A in the red cells and lets γ-globin be expressed again. In 2021, Frangoul and colleagues reported for the first time in the New England Journal of Medicine that this editing could free sickle cell disease and β-thalassemia patients from transfusion dependence; by 2024, the same team gave the pivotal-trial results for sickle cell disease in the same journal: 44 patients received the treatment, and among the 30 evaluable patients with at least one year of follow-up, 29 (97%) went at least 12 consecutive months without a severe vaso-occlusive crisis, and all 30 were no longer hospitalized for one. This therapy is exa-cel (exagamglogene autotemcel, trade name Casgevy), and it is also the first CRISPR gene-editing therapy to win marketing approval. The full backstory of the BCL11A target itself is the subject of another piece; what matters here is the route it represents: swap in a backup, rather than repair the original.

The price of this road is written into how it works. CRISPR-Cas9 accomplishes editing by creating a double-strand break in DNA, and the repair of that break is not always clean and neat — large deletions, chromosome-level rearrangements, and an activated p53 response are all potential side effects of this kind of break (the safety ledger on this part is left for the discussion of editing safety). Moreover, whether the reawakened HbF is enough, and whether it is evenly distributed among the red cells, directly determines the effect of suppressing sickling — and this is both limited by individual biology and not something that can be solved linearly simply by cranking up the editing efficiency.

The Third Road: Change the Bad Letter Back

If the first two roads are one of “bypassing” and one of “replacing,” the third road is the most stubborn: it wants to go straight back to that erroneous letter and change it. The trouble is that for classic CRISPR to precisely swap one base for another, it must first create a break and then rely on the cell repairing it against a template — low in efficiency and high in risk. Base editing offers another technique — it does not cut the double strand, but instead, like a chemical eraser, directly rewrites one base into another in place.

The truly ingenious step lies in the direction of the change. Changing the pathogenic valine codon in one step back to the original glutamate is not chemically convenient; but with a change of approach, a single adenine edit is enough to rewrite HbS into a natural variant called Makassar — a benign β-globin that exists in the population yet is not pathogenic. In 2021, Newby and colleagues in Nature accomplished this with a custom adenine base editor (ABE8e-NRCH): in the hematopoietic stem cells of sickle cell patients, about 80% of the pathogenic alleles were changed in place to the Makassar version; when the edited human cells were transplanted into mice, this proportion still held at nearly 70% sixteen weeks later, and hypoxia-induced sickling was reduced to one-fifth of the original. In a humanized mouse model of sickle cell disease, Makassar-type β-globin made up about 80% of the blood’s β-globin, sickling was reduced by about two-thirds, and the mice’s hematologic indices approached normal. More crucially, two details: as long as more than 20% of the alleles are changed to Makassar, it is enough to reverse the pathological phenotype; and because no double-strand break is created throughout, it avoids the kind of accompanying p53 activation and large deletions seen with Cas9.

Yet this most elegant road is, at present, both the most ahead and the most behind — elegant in mechanism, behind in evidence. The first two roads already have approved, marketed products and dozens of patients followed for years, whereas the approach of changing HbS directly into Makassar still has its solid efficacy and safety data coming mainly from cells and mice, and has not yet reached an equally mature human stage. It has proven the principle, but has not yet proven that it is equally safe, durable, and scalable in humans.

Three Roads, Laid Side by Side

Set the three roads side by side and the divisions become clear. Lentiviral gene addition is “bypassing the cause”: the bad gene is left untouched, only an anti-sickling good backup is added — it has the thickest clinical evidence, yet it permanently mixes a foreign gene into the genome, and over the long term it is the uncertainty of the insertion that must be watched. CRISPR knockout of BCL11A is “swapping in a backup”: it does not even touch the bad gene, but instead reawakens the body’s own fetal hemoglobin, and has become the first approved CRISPR therapy — the price being its reliance on double-strand breaks and the biological ceiling of whether the reawakened HbF is enough. Base editing to Makassar is the road closest to “correcting the cause” among the three: it leaves no break and changes the bad letter directly into a benign variant, cleanest in theory — but its human evidence chain is the shortest, its promise and its unknowns both the greatest.

It is worth noting that these three roads share the same heavy chassis: to this day, they are all still a complete autologous transplant — requiring mobilization, collection, ex vivo manipulation, and myeloablative conditioning, at a cost as high as the million-dollar level, and only a handful of centers can perform it. Which is technically more “clever” does not directly translate into which can better reach the regions truly crushed by sickle cell disease — and those regions are where the vast majority of the world’s patients are (this ledger of “curable yet unreached” is left to be unfolded in the discussion of accessibility).

From the error of one letter to the standoff of three ways to fix it, what sickle cell disease has traveled over this half century is, in fact, an open demonstration match of gene-therapy methodology: bypass, replace, or correct — which one, at what price, is most worth it. The answer need not be only one — different disease severity, different resources, different appetites for risk may, from the start, call for different roads. And when one day changing the bad letter back becomes as safe and widespread as transplantation is today, then that molecular disease, seen clearly as far back as 1949, will finally have been cured at the very letter where it went wrong.

References

  1. Pauling L, Itano HA, Singer SJ, Wells IC. Sickle cell anemia, a molecular disease. Science. 1949;110(2865):543-8. DOI
  2. Ingram VM. Gene mutations in human haemoglobin: the chemical difference between normal and sickle cell haemoglobin. Nature. 1957;180(4581):326-8. DOI
  3. Ribeil JA, et al. Gene Therapy in a Patient with Sickle Cell Disease. N Engl J Med. 2017;376(9):848-855. DOI
  4. Kanter J, et al. Biologic and Clinical Efficacy of LentiGlobin for Sickle Cell Disease. N Engl J Med. 2022;386(7):617-628. DOI
  5. Goyal S, et al. Acute Myeloid Leukemia Case after Gene Therapy for Sickle Cell Disease. N Engl J Med. 2022;386(2):138-147. DOI
  6. Frangoul H, et al. CRISPR-Cas9 Gene Editing for Sickle Cell Disease and β-Thalassemia. N Engl J Med. 2021;384(3):252-260. DOI
  7. Frangoul H, et al. Exagamglogene Autotemcel for Severe Sickle Cell Disease. N Engl J Med. 2024;390(18):1649-1662. DOI
  8. Newby GA, et al. Base editing of haematopoietic stem cells rescues sickle cell disease in mice. Nature. 2021;595(7866):295-302. DOI