Yang Liu

← Primer

HSC · Healing Source Code

自体 HSC 基因治疗为什么仍是一场完整的移植——动员到长期随访的全流程 Why Autologous HSC Gene Therapy Is Still a Full Transplant — From Mobilization to Long-Term Follow-Up

2026-07-31 · 造血干细胞移植, plerixafor, 清髓, engraftment, 长期随访, exa-cel

媒体把 exa-cel、lovo-cel 这些疗法叫作”一针治愈”,说的是最终结果,略去的是通往结果的那条路。真正落在病人身上的,是连续几天的动员注射、一次耗时数小时的血液分离、几个月里躺在冰柜里等待放行的细胞、四天连续输注的化疗药物,然后是三到五周几乎没有任何免疫力的空窗期——直到中性粒细胞计数一格一格地爬回来。把编辑好的 CD34+ 细胞写进基因组,只是这条链路里最短、也最常被拿出来讲的一环;剩下的每一步,都是移植医学写了几十年的老功课,一步都省不掉。

从血管里取出未来

自体基因治疗的第一步,不是编辑,而是把足够多的造血干细胞从骨髓里请出来。健康供者做这件事时,标准做法是打几天 G-CSF——它能让干细胞从骨髓龛位脱离,涌入外周血,方便用血细胞分离机采集。但对镰状细胞病患者,这条最常规的路是被明令禁止的:Fitzhugh 等人在 2009 年的 Cytotherapy 上系统梳理了多起病例,G-CSF 会诱发白细胞骤增,继而触发血管闭塞危象、急性胸部综合征乃至多器官衰竭,呼吁对这类患者暂停使用 G-CSF。

于是镰状细胞病的基因治疗试验大多改用另一条路:单用 plerixafor——一种通过阻断 CXCR4 受体、把干细胞从骨髓”推”进血液的小分子药物,不依赖 G-CSF 那种会诱发白细胞骤增的机制。但这条替代路径本身也谈不上万无一失。Boulad 等人 2018 年在 Haematologica 报告的剂量爬坡研究里,15 名镰状细胞病患者中有 9 人达到了 CD34+ >30 cells/μL 的采集阈值,且 CD34+ 动员效果与 plerixafor 剂量之间没有清晰的量效关系;与此同时,白细胞与中性粒细胞计数却随剂量呈上升趋势(分别 P=0.05、P=0.03)。研究团队给出的结论是:plerixafor 耐受性尚可,但在这组患者里未能稳定地实现足量 CD34+ 细胞动员。也是因为这一点,lovo-cel 的 HGB-206 试验把 plerixafor 动员和血细胞分离机采集结合起来使用,逐步取代了早期方案里更具创伤性的骨髓穿刺采集——但动员本身能否一次采到足量细胞,至今仍是自体基因治疗采集环节里一个因人而异、尚未被完全解决的现实问题。

工厂与冰柜

采集下来的细胞并不会立刻回到病人体内。它们要先被富集出 CD34+ 群体,再送进 GMP 车间接受编辑或转导——用 CRISPR-Cas9 切开 BCL11A 的红系 enhancer(这是 exa-cel 的路线,最早的概念验证见于 Frangoul 等人 2021 年在 N Engl J Med 上报告的两名先导患者的病例数据),或者用慢病毒载体写入一段抗镰变的 β-珠蛋白基因(lovo-cel、beti-cel 走的是这条整合式路线)。这部分工程细节,系列后面讲编辑工具和递送时会分别展开;在移植医学的视角里,重要的是这一步意味着什么:细胞离开身体的时间被拉长了。从采集到放行检测全部完成、可以安全回输,往往是以周甚至月计的等待,期间病人仍要用常规手段维持病情稳定。基因治疗常被想象成”当场改写、当场治愈”,而现实是这台”活体药厂”的生产周期,本身就是流程里一段沉默却真实的成本。

清髓,不是清创

细胞放行之后,真正决定这场治疗风险高低的一步才登场:conditioning。exa-cel 与 lovo-cel 的临床试验用的都是单药、按药代动力学个体化调整剂量的白消安(busulfan)清髓方案——连续几天的大剂量化疗,目的是把病人自己原有的骨髓尽可能清空,给回输的编辑细胞腾出生存位置。这不是一次”清创式”的辅助步骤,而是 myeloablative(清髓性)化疗本身:它会摧毁全部造血,带来脱发、恶心、粘膜炎等急性毒性,也带来长期不育风险——这些正是异体移植病人几十年来一直在承受的同一类代价。自体基因治疗省掉的,只是异体移植里为防排异而必需的免疫抑制和供者匹配;清髓这一步,几乎原封不动地被继承了下来。系列后面会专门拆开”清髓到底在做什么”这个问题——清病、免疫抑制、腾出龛位,其实是三件不同的事,而基因治疗只需要其中一件。

骨髓归零之后的三周

回输之后,病人要在几乎没有免疫力的状态下,等待编辑过的细胞在骨髓里重新定居、开始产血。这段”空窗期”有多长,exa-cel 的关键 3 期试验给出了具体数字:在镰状细胞病队列(n=44)中,中性粒细胞 engraftment 的中位时间是 27 天(范围 15–40 天);血小板 engraftment 的中位时间是 35 天(范围 23–126 天),定义是连续三次不同日期检测达到相应阈值、且未使用备用的未编辑细胞救援——这组数字同时见于 Frangoul 等人 2024 年发表在 N Engl J Med 上的原始报告,以及 Leonard 与 Kanter 2025 年在 Experimental Biology and Medicine 上发表的综述。这意味着,一名接受基因治疗的病人,通常要在医院或类似环境里熬过将近一个月严重的中性粒细胞减少期,期间感染是最直接的威胁。lovo-cel 的 HGB-206 试验走的是同一套骨架:plerixafor 动员采集,busulfan myeloablative conditioning,回输后同样要等待 engraftment,试验报告显示移植物顺利植入、未见明确的移植物失败。但这条路径并非没有代价:lovo-cel 现行的 FDA 处方信息带有关于血液肿瘤的黑框警告——早期方案(Study 1, Group A)里出现过两例急性髓系白血病,而即便是采用了有别于早期方案的动员与制造流程的队列(Study 1, Group C),也报告了一例骨髓增生异常综合征;监管方认为,动员、清髓与回输带来的造血应激本身可能增加这类风险,要求对接受 lovo-cel 治疗的患者终身监测血液肿瘤,具体的随访安排留到后面讨论安全性终点时再展开。这些数字听起来枯燥,却是判断一款”基因治疗”到底有多接近”移植”的最直接标尺:凡是要清髓、要等 engraftment、要熬过这几周空窗期的,骨子里就是一次自体移植,只是移植物被预先改写过。

出院不是终点

即便顺利出院,病人的观察也远没有结束。因为整合到基因组里的载体、或是留下永久性编辑的细胞,理论上都存在延迟出现不良反应的可能——包括插入突变带来的克隆异常。FDA 在 2020 年发布的行业指南《Long Term Follow-Up After Administration of Human Gene Therapy Products》要求,含整合载体或基因组编辑成分的疗法,长期随访最长可达 15 年:前 5 年需要每年一次的当面检查,之后最多再加 10 年的年度问询。换句话说,这场”一针”治疗真正意义上的终点,要在病床之外,再往后数十五年才能画上——这也是系列后面讨论安全性终点、克隆造血监测时会反复回到的时间尺度。

回到开头那幅被压缩掉的画面:动员针剂、采集椅、GMP 车间的冰柜、连续四天的白消安、三到五周的空窗期,以及此后长达十五年的随访问卷——这才是”基因治疗”完整的物理形态。之所以要把这条链路重新摆回台面,是因为它划出了这一代疗法的真实边界:只要治疗还需要先把细胞请出体外、再靠清髓换取生存空间,它就永远是一场移植,只是移植物出自病人自己、且被预先改写过。真正能甩掉这整套流程的,只有把编辑直接送进体内、在原地完成——那是系列另一条主线(in vivo 编辑)要回答的问题,而在那一天到来之前,动员、采集、清髓与漫长的随访,仍是每一款自体 HSC 基因疗法绕不开的底色。

参考文献

  1. Fitzhugh CD, Hsieh MM, Bolan CD, Saenz C, Tisdale JF. Granulocyte colony-stimulating factor (G-CSF) administration in individuals with sickle cell disease: time for a moratorium? Cytotherapy. 2009;11(4):464-471. DOI
  2. Boulad F, Shore T, van Besien K, et al. Safety and efficacy of plerixafor dose escalation for the mobilization of CD34+ hematopoietic progenitor cells in patients with sickle cell disease: interim results. Haematologica. 2018;103(5):770-777. DOI
  3. Frangoul H, Locatelli F, Sharma A, et al. Exagamglogene Autotemcel for Severe Sickle Cell Disease. N Engl J Med. 2024;390(18):1649-1662. DOI
  4. Kanter J, Walters MC, Krishnamurti L, et al. Biologic and Clinical Efficacy of LentiGlobin for Sickle Cell Disease. N Engl J Med. 2022;386(7):617-628. DOI
  5. Locatelli F, Thompson AA, Kwiatkowski JL, et al. Betibeglogene Autotemcel Gene Therapy for Non-β0/β0 Genotype β-Thalassemia. N Engl J Med. 2022;386(5):415-427. DOI
  6. U.S. Food and Drug Administration. Long Term Follow-Up After Administration of Human Gene Therapy Products: Guidance for Industry. January 2020. FDA
  7. Frangoul H, Ho TW, Corbacioglu S, et al. CRISPR-Cas9 Gene Editing for Sickle Cell Disease and β-Thalassemia. N Engl J Med. 2021;384(3):252-260. DOI
  8. Leonard A, Kanter J. Clinical data comparison for FDA-approved gene therapies in sickle cell disease. Exp Biol Med (Maywood). 2025;250:10806. DOI
  9. U.S. Food and Drug Administration. LYFGENIA (lovotibeglogene autotemcel) Prescribing Information, including Boxed Warning for hematologic malignancy. FDA

The press calls therapies like exa-cel and lovo-cel a “one-shot cure” — a label for the outcome that skips over the road to it. What actually happens to the patient is a run of consecutive days injecting a mobilizing agent, a several-hour apheresis session, months in a freezer while the cells wait for release testing, four straight days of chemotherapy infusions, and then three to five weeks with almost no immune function — until the neutrophil count climbs back, one point at a time. Writing an edited instruction into the genome of CD34+ cells is only the shortest link in that chain, and the one most often singled out for the story; every other step is homework hematopoietic transplant medicine has already done for decades, and none of it can be skipped.

Pulling a Future Out of the Bloodstream

The first step in autologous gene therapy is not editing — it’s coaxing enough hematopoietic stem cells out of the marrow in the first place. When a healthy donor does this, the standard approach is several days of G-CSF, which lets stem cells detach from their marrow niche and pour into peripheral blood, where an apheresis machine can collect them. For patients with sickle cell disease, though, that most routine of paths is explicitly off-limits: Fitzhugh and colleagues, reviewing multiple cases in a 2009 Cytotherapy paper, found that G-CSF can trigger a surge in white blood cells that in turn precipitates vaso-occlusive crisis, acute chest syndrome, and even multi-organ failure, and called for a moratorium on G-CSF in this population.

So sickle cell gene therapy trials mostly take a different route: plerixafor alone, a small molecule that blocks the CXCR4 receptor and “pushes” stem cells out of the marrow into the blood, without relying on the mechanism that makes G-CSF trigger a white-cell surge. But that alternative path is itself far from foolproof. In the 2018 dose-escalation study Boulad and colleagues reported in Haematologica, 9 of 15 sickle cell patients reached the collection threshold of CD34+ >30 cells/μL, and there was no clear dose-response relationship between plerixafor dose and CD34+ mobilization; at the same time, white blood cell and neutrophil counts rose with dose (P=0.05 and P=0.03, respectively). The team’s conclusion: plerixafor was reasonably well tolerated but did not reliably achieve adequate CD34+ mobilization in this cohort. Partly for that reason, lovo-cel’s HGB-206 trial combined plerixafor mobilization with apheresis collection, gradually replacing the more invasive bone-marrow harvest used in earlier protocols — but whether mobilization alone can collect enough cells in a single pass remains, to this day, a real and person-to-person problem in the collection step of autologous gene therapy that has not been fully solved.

The Factory and the Freezer

The collected cells don’t go straight back into the patient. First they’re enriched into a CD34+ population, then shipped into a GMP facility for editing or transduction — cutting the erythroid enhancer of BCL11A with CRISPR-Cas9 (exa-cel’s route, whose earliest proof of concept came from the two lead-in patients Frangoul and colleagues reported in N Engl J Med in 2021), or writing in an anti-sickling β-globin gene with a lentiviral vector (the integrating route taken by lovo-cel and beti-cel). Those engineering details will get their own treatment later in this series, when we cover editing tools and delivery; from the standpoint of transplant medicine, what matters here is what this step means: the time the cells spend outside the body stretches out. From collection to the completion of all release testing that clears the product for safe reinfusion is typically a wait measured in weeks or months, during which the patient still has to be kept stable by conventional means. Gene therapy is often pictured as rewriting and curing on the spot; the reality is that this “living drug factory’s” production cycle is itself a quiet but real cost built into the process.

Conditioning, Not a Quick Cleanup

Once the cells are released, the step that actually determines how risky this treatment is takes the stage: conditioning. The exa-cel and lovo-cel clinical trials both use a single-agent busulfan myeloablative regimen, with dosing individualized by pharmacokinetics — several straight days of high-dose chemotherapy meant to clear out as much of the patient’s own marrow as possible, opening up living space for the infused edited cells. This isn’t some ancillary cleanup step; it is myeloablative chemotherapy itself: it destroys the entire blood-forming system, brings acute toxicities like hair loss, nausea, and mucositis, and carries a long-term risk of infertility — the same category of cost that allogeneic transplant patients have been paying for decades. What autologous gene therapy dispenses with is only the immunosuppression and donor matching that allogeneic transplant needs to prevent rejection; conditioning itself is inherited almost intact. Later in the series we’ll take apart the question of what conditioning is actually doing — clearing disease, suppressing immunity, and opening up niche space are, in fact, three separate things, and gene therapy only needs one of them.

Three Weeks After the Marrow Hits Zero

After reinfusion, the patient waits — with almost no immune function — for the edited cells to resettle in the marrow and start making blood. Exa-cel’s pivotal phase 3 trial puts a number on how long that window lasts: in the sickle cell cohort (n=44), the median time to neutrophil engraftment was 27 days (range 15–40), and median time to platelet engraftment was 35 days (range 23–126), defined as three consecutive measurements on different days reaching the relevant threshold without rescue from the backup, unedited cells. Both numbers appear in Frangoul and colleagues’ original 2024 N Engl J Med report and in Leonard and Kanter’s 2025 review in Experimental Biology and Medicine. In practice, this means a patient receiving gene therapy typically has to get through nearly a month of severe neutropenia in a hospital or comparable setting, during which infection is the most immediate threat. Lovo-cel’s HGB-206 trial runs on the same skeleton: plerixafor mobilization and apheresis collection, busulfan myeloablative conditioning, and the same wait for engraftment after reinfusion — the trial reports show the graft took successfully, with no clear graft failure. But this path isn’t free of cost either: lovo-cel’s current FDA prescribing information carries a boxed warning for hematologic malignancy — the earlier protocol (Study 1, Group A) produced two cases of acute myeloid leukemia, and even the cohort using a mobilization and manufacturing process different from the earlier protocol (Study 1, Group C) reported one case of myelodysplastic syndrome. Regulators consider it possible that the hematopoietic stress from mobilization, conditioning, and reinfusion itself raises this risk, and require lifelong monitoring for hematologic malignancy in patients treated with lovo-cel — the specifics of that follow-up schedule are left for a later discussion of safety endpoints. These numbers sound dry, but they’re the most direct yardstick for how close a “gene therapy” really sits to a “transplant”: anything that requires myeloablation, a wait for engraftment, and getting through these weeks of the window is, at bottom, an autologous transplant — only the graft has been rewritten in advance.

Discharge Is Not the Finish Line

Even a smooth discharge doesn’t mean observation is over. Cells carrying an integrated vector, or cells that have been permanently edited, both carry a theoretical possibility of delayed adverse effects — including clonal abnormalities from insertional mutagenesis. The FDA’s 2020 industry guidance, Long Term Follow-Up After Administration of Human Gene Therapy Products, requires that therapies with an integrating vector or genome-editing component undergo long-term follow-up lasting up to 15 years: annual in-person exams for the first 5 years, followed by up to another 10 years of annual questionnaires. In other words, the real endpoint of this “one-shot” treatment, outside the hospital bed, doesn’t get drawn until another fifteen years have passed — a timescale the series will keep coming back to when it covers safety endpoints and clonal hematopoiesis surveillance.

Back to the picture compressed out of the opening line: the mobilization injections, the apheresis chair, the freezer in the GMP suite, four straight days of busulfan, three to five weeks in the window, and then up to fifteen years of follow-up questionnaires — that is the full physical shape of “gene therapy.” The reason to put this chain back on the table is that it marks the real boundary of this generation of therapies: as long as treatment still requires pulling cells out of the body first and then buying survival space with myeloablation, it will always be a transplant — only the graft comes from the patient’s own body, and has been rewritten in advance. The only way to truly shed this entire process is to send the edit directly into the body and complete it in place — that is the question for another main thread in this series, in vivo editing. Until that day arrives, mobilization, collection, conditioning, and the long follow-up remain the inescapable backdrop of every autologous HSC gene therapy.

References

  1. Fitzhugh CD, Hsieh MM, Bolan CD, Saenz C, Tisdale JF. Granulocyte colony-stimulating factor (G-CSF) administration in individuals with sickle cell disease: time for a moratorium? Cytotherapy. 2009;11(4):464-471. DOI
  2. Boulad F, Shore T, van Besien K, et al. Safety and efficacy of plerixafor dose escalation for the mobilization of CD34+ hematopoietic progenitor cells in patients with sickle cell disease: interim results. Haematologica. 2018;103(5):770-777. DOI
  3. Frangoul H, Locatelli F, Sharma A, et al. Exagamglogene Autotemcel for Severe Sickle Cell Disease. N Engl J Med. 2024;390(18):1649-1662. DOI
  4. Kanter J, Walters MC, Krishnamurti L, et al. Biologic and Clinical Efficacy of LentiGlobin for Sickle Cell Disease. N Engl J Med. 2022;386(7):617-628. DOI
  5. Locatelli F, Thompson AA, Kwiatkowski JL, et al. Betibeglogene Autotemcel Gene Therapy for Non-β0/β0 Genotype β-Thalassemia. N Engl J Med. 2022;386(5):415-427. DOI
  6. U.S. Food and Drug Administration. Long Term Follow-Up After Administration of Human Gene Therapy Products: Guidance for Industry. January 2020. FDA
  7. Frangoul H, Ho TW, Corbacioglu S, et al. CRISPR-Cas9 Gene Editing for Sickle Cell Disease and β-Thalassemia. N Engl J Med. 2021;384(3):252-260. DOI
  8. Leonard A, Kanter J. Clinical data comparison for FDA-approved gene therapies in sickle cell disease. Exp Biol Med (Maywood). 2025;250:10806. DOI
  9. U.S. Food and Drug Administration. LYFGENIA (lovotibeglogene autotemcel) Prescribing Information, including Boxed Warning for hematologic malignancy. FDA