Yang Liu

← Primer

HSC · Healing Source Code

采不出细胞,就没有药——HSC 基因治疗的第一公里 No cells, no drug—the first mile of HSC gene therapy

2026-07-25 · 造血干细胞, apheresis, 动员, 镰刀型贫血, CASGEVY

一台血细胞分离机在病房里低声运转。血从一条管路出去,在离心腔里分成几层,含造血干细胞的那薄薄一层被引到采集袋,剩下的顺着另一条管路回到身体。几个小时后,袋子里那点浑浊的液体就是这场两百多万美元治疗的全部原料。基因治疗所有的精巧——载体、编辑器、无菌车间、放行检定——都发生在这袋细胞之后;如果这袋细胞不够,后面的一切都不会开始。这段路叫”第一公里”。

药的原料是病人自己

先把量级说清楚。CASGEVY 的美国说明书写得毫不含糊:为制造这一剂药,建议的总采集目标是每公斤体重至少 20×10⁶ 个 CD34⁺ 细胞,而最终回输的最低剂量只有 3×10⁶/kg;此外还必须另外采集并冻存至少 2×10⁶/kg 未经修饰的备份细胞,以备 engraftment 失败或产品出问题时救急。也就是说,理想情况下要从病人身上取走最终用量好几倍的细胞——损耗发生在纯化、培养、编辑和放行的每一环。万一成品达不到最低剂量,说明书给的处理方式只有一条:再来一轮 mobilization 和 apheresis,两轮之间至少隔 14 天。

问题是,造血干细胞并不天然在血里流动。它们靠骨髓基质分泌的趋化因子 CXCL12 与自身表面的受体 CXCR4 结合被锚在原位,要把它们弄出来,就得解开这个锚。传统做法是打 G-CSF:2003 年 Lévesque 等人在 J Clin Invest 上报告,在小鼠体内,G-CSF 动员伴随着骨髓中性粒细胞蛋白酶的蓄积,CXCR4 的 N 端被切断、CXCL12 本身也被降解——锚是被”腐蚀”掉的,这也与 G-CSF 需要连打数天才见效的用法相符。看清这条通路,更直接的想法就浮了出来:不如直接堵住受体。2005 年,Broxmeyer 等人在 J Exp Med 上报告,CXCR4 拮抗剂 AMD3100(即后来的 plerixafor)能在数小时内动员小鼠和人的造血干祖细胞;2009 年 DiPersio 等人的两项 III 期随机双盲试验把它送进临床,在非霍奇金淋巴瘤中,plerixafor 加 G-CSF 组有 59% 的患者在四次以内的 apheresis 里采足 5×10⁶ CD34⁺/kg,安慰剂组只有 20%。

唯独在镰刀型贫血里,标准工具是禁忌

可这套成熟工具,偏偏在最需要基因治疗的那个病里用不了。1998 年,Abboud、Laver 和 Blau 在 Lancet 上发出一则简短警告:他们试图用 G-CSF 为一名镰刀型贫血患者动员自体造血干细胞以供基因治疗研究,结果引发严重的疼痛危象,作者认定这条路不安全。2001 年,Adler 等人在 Blood 上报告了更坏的结局:一名血红蛋白 SC 病患者(镰刀型贫血中较轻的一型)作为同胞供者接受 G-CSF 动员——这位供者的病轻到在供者筛查之前从未被发现——最终死于多器官功能衰竭。到 2009 年,Fitzhugh 等人在 Cytotherapy 上把当时文献中全部 11 例记录汇总起来:7 人因并发症住院,2 人发生多器官功能衰竭,其中 1 人死亡。更令人不安的是,他们找不到任何安全阈值——一名已被输血把 HbS 压到 14.4% 的患者仍发作了需要住院的危象,而发生多器官衰竭的一位,白细胞峰值只有 6.4×10⁹/L;另有一位患者在每天仅 2.5 μg/kg 的剂量下就被送进了医院。那篇文章的标题是个问句:是时候暂停了吗。后来的答案是肯定的——今天 CASGEVY 说明书里写着一句不留余地的话:镰刀型贫血患者不得使用 G-CSF 动员。

于是只剩 plerixafor,而且要讲究

2018 年,三组人几乎同时回答了”单用 plerixafor 行不行”。Boulad 等人在 Haematologica 上报告了一项剂量递增研究:15 名镰刀型贫血患者中只有 8 人外周血 CD34⁺ 浓度超过 30 个/μL,而且把剂量往上加并没有带来一致的提升——不过这批人里有 10 人仍在服 hydroxyurea、只有 1 人长期输血,作者认为 hydroxyurea 可能是动员不足的原因之一。两名患者分别在 80 μg/kg 和 240 μg/kg 剂量下发作了血管闭塞危象——plerixafor 本身也不是全无代价。同一期上,Lagresle-Peyrou 等人给出了另一半答案:三名重症患者先经红细胞置换把 HbS 降到 30% 以下,再打一针 0.24 mg/kg plerixafor,除枸橼酸相关的中度低钾外没有不良反应,三人分别采到 4.6、5.8 和 4.5×10⁶ CD34⁺/kg;更关键的是,这些细胞高表达干性相关基因,移植进免疫缺陷小鼠后重建效率很高。而 Esrick 等人在 Blood Adv 上指出,真正把产量做上去的往往不是药,而是采集本身的手艺:在六名接受输血的成人中,他们把分离机的采集界面调深、并在给药后四小时内开始 apheresis,标准剂量下最高拿到 24.5×10⁶ CD34⁺/kg,产物还富集了免疫表型上的 long-term HSC。

mobilization 方案不只决定采到多少细胞,也决定采到的是什么细胞。

病越重,越采不出来

细胞产量并不是运气。2021 年,Leonard 等人在 Blood Adv 上汇总了两个中心 23 名镰刀型贫血受试者的数据:两轮以内采到的 CD34⁺ 中位数是 4.0×10⁶/kg,范围从 1.5 到 12.0——八倍的跨度。产量与年龄负相关、与采集前外周血 CD34 计数正相关,这些在别的人群里也一样;但有两条是这个病独有的:hydroxyurea 停药天数越长产量越高(在五周之内如此),以及产量与疾病严重程度负相关——过去一年住院次数越多、日常服用的慢性疼痛药物越多,采到的细胞越少。这条相关性读起来令人不适,因为它意味着:病得最重、最该被治愈的那些人,恰恰是最难为他们造出药来的人。

这不是纸面推演。在 12 岁及以上镰刀型贫血患者的注册试验 CLIMB SCD-121(疗效结果由 Frangoul 等人 2024 年发表于 N Engl J Med)中,按 CASGEVY 说明书记录的采集数据:63 人入组,58 人开始动员,其中 44 人(76%)最终接受了输注;制造成品外加备份细胞所需的周期数中位为 2 轮、范围 1 到 6 轮;六名患者(10%)因为细胞采集不足而没能接受治疗,儿童试验里 13 人中也有 1 人因同一原因出局。作为对照,同一款药用于输血依赖型 β-地中海贫血时可以用 G-CSF 联合 plerixafor,59 名入组者全部进入动员,所需周期的中位数是 1 轮,最终 52 人(88%)接受了输注。同样的编辑、同样的生产线,只因为上游的病不同,第一公里的中位长度就翻了一倍;而在镰刀型贫血这一侧,有十分之一的人根本走不完。

数出来的 CD34,不等于药效

还有一层更根本的问题:用来验收第一公里的那个指标本身并不完美。CD34 是一个便于用流式和磁珠抓取的表面标志,而不是长期重建能力的度量。Fanconi 贫血把这一点暴露得最清楚。2021 年 Sevilla 等人报告,11 名接受 filgrastim 加 plerixafor 动员的患者中,9 人达到了启动 apheresis 所需的外周血 5 个 CD34⁺/μL 阈值,没达到的两人恰好是年龄最大的两位(15 岁和 16 岁);两到三次 apheresis 采到的 CD34⁺ 中位数为 4.27×10⁶/kg,可经免疫磁珠纯化后骤降到 1.1×10⁶/kg——作者推测是因为这些细胞的 CD34 抗原表达偏弱。同一群细胞,换一种抓取方式,数出来的”剂量”就掉了四分之三——而作者同时指出,这个缩水后的数字仍足以让改造过的细胞在未做 conditioning 的患者体内 engraft。这道缝隙后面还会再遇到:体外扩增能不能把数量补上、效力究竟该用什么定义,是另外两篇的事。

回到那台低声运转的分离机。它看上去是整条流程里技术含量最低的一环——没有酶,没有载体,没有测序仪。可正是在这一环,一部分患者被筛了出去,而筛掉他们的理由与基因型无关,只与年龄、病情轻重、用药史,以及那一天骨髓愿意放出多少细胞有关。ex vivo 基因治疗把治愈变成了一件可以被制造的事,而制造总得先有原料;当原料只能从病人自己身上取,病人的状态就成了产能的一部分。这大概也是 in vivo 路线最诱人的地方之一:它要取消的,恰恰是这第一公里。

参考文献

  1. CASGEVY (exagamglogene autotemcel) US Prescribing Information. Vertex Pharmaceuticals; revised 07/2026. FDA 版本 · 厂商 PI
  2. Lévesque JP, Hendy J, Takamatsu Y, Simmons PJ, Bendall LJ. Disruption of the CXCR4/CXCL12 chemotactic interaction during hematopoietic stem cell mobilization induced by GCSF or cyclophosphamide. J Clin Invest. 2003;111(2):187-196. DOI
  3. Broxmeyer HE, Orschell CM, Clapp DW, et al. Rapid mobilization of murine and human hematopoietic stem and progenitor cells with AMD3100, a CXCR4 antagonist. J Exp Med. 2005;201(8):1307-1318. DOI
  4. DiPersio JF, Micallef IN, Stiff PJ, et al. Phase III prospective randomized double-blind placebo-controlled trial of plerixafor plus granulocyte colony-stimulating factor compared with placebo plus granulocyte colony-stimulating factor for autologous stem-cell mobilization and transplantation for patients with non-Hodgkin’s lymphoma. J Clin Oncol. 2009;27(28):4767-4773. DOI
  5. DiPersio JF, Stadtmauer EA, Nademanee A, et al. Plerixafor and G-CSF versus placebo and G-CSF to mobilize hematopoietic stem cells for autologous stem cell transplantation in patients with multiple myeloma. Blood. 2009;113(23):5720-5726. DOI
  6. Abboud M, Laver J, Blau CA. Granulocytosis causing sickle-cell crisis. Lancet. 1998;351(9107):959. DOI
  7. Adler BK, Salzman DE, Carabasi MH, Vaughan WP, Reddy VV, Prchal JT. Fatal sickle cell crisis after granulocyte colony-stimulating factor administration. Blood. 2001;97(10):3313-3314. DOI
  8. 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
  9. 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
  10. Lagresle-Peyrou C, Lefrère F, Magrin E, et al. Plerixafor enables safe, rapid, efficient mobilization of hematopoietic stem cells in sickle cell disease patients after exchange transfusion. Haematologica. 2018;103(5):778-786. DOI
  11. Esrick EB, Manis JP, Daley H, et al. Successful hematopoietic stem cell mobilization and apheresis collection using plerixafor alone in sickle cell patients. Blood Adv. 2018;2(19):2505-2512. DOI
  12. Leonard A, Sharma A, Uchida N, et al. Disease severity impacts plerixafor-mobilized stem cell collection in patients with sickle cell disease. Blood Adv. 2021;5(9):2403-2411. DOI
  13. Sevilla J, Navarro S, Rio P, et al. Improved collection of hematopoietic stem cells and progenitors from Fanconi anemia patients for gene therapy purposes. Mol Ther Methods Clin Dev. 2021;22:66-75. DOI
  14. 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

An apheresis machine hums quietly in a hospital room. Blood leaves through one line, separates into layers in the centrifuge chamber, and the thin stratum holding hematopoietic stem cells is drawn off into a collection bag while the rest returns to the body through another line. A few hours later, the cloudy fluid in that bag is the entire raw material for a treatment costing over two million dollars. All the ingenuity of gene therapy—the vectors, the editors, the clean rooms, the release assays—happens downstream of this bag of cells; if the bag falls short, none of it begins. This stretch is the first mile.

The raw material is the patient

Start with the magnitudes. The US label for CASGEVY is unambiguous: to manufacture one dose, the recommended total collection target is at least 20×10⁶ CD34⁺ cells per kilogram of body weight, while the minimum infused dose is only 3×10⁶/kg; on top of that, at least 2×10⁶/kg of unmodified back-up rescue cells must be collected and cryopreserved separately, in case of engraftment failure or product compromise. In other words, the ideal is to take from the patient several times what will ultimately be given back—losses accumulate at every step of selection, culture, editing, and release. If the finished product fails to reach the minimum dose, the label offers exactly one remedy: another cycle of mobilization and apheresis, with at least 14 days between cycles.

The problem is that hematopoietic stem cells do not naturally circulate in the blood. They are anchored in place by the chemokine CXCL12 secreted by the bone marrow stroma binding their surface receptor CXCR4, and to get them out that anchor has to be released. The traditional approach is G-CSF: in 2003, Lévesque and colleagues reported in J Clin Invest that in mice, G-CSF mobilization coincides with an accumulation of neutrophil proteases in the marrow, cleavage of the N-terminus of CXCR4, and degradation of CXCL12 itself—the anchor is eroded away, which fits the fact that G-CSF has to be given for several days before it works. Seeing the pathway clearly, a more direct idea suggested itself: block the receptor outright. In 2005, Broxmeyer and colleagues reported in J Exp Med that the CXCR4 antagonist AMD3100—later known as plerixafor—could mobilize hematopoietic stem and progenitor cells in mice and in humans within hours; in 2009, two phase III randomized double-blind trials from DiPersio and colleagues carried it into practice, showing that in non-Hodgkin’s lymphoma, 59% of patients in the plerixafor-plus-G-CSF arm collected at least 5×10⁶ CD34⁺/kg in four or fewer apheresis days, against 20% on placebo.

In sickle cell disease alone, the standard tool is forbidden

Yet this mature toolkit cannot be used in the very disease that most needs gene therapy. In 1998, Abboud, Laver, and Blau published a brief warning in Lancet: they had tried to use G-CSF to mobilize autologous hematopoietic stem cells from a patient with sickle cell disease for gene therapy studies, it triggered a severe pain crisis, and they concluded the approach was not safe. In 2001, Adler and colleagues reported a worse outcome in Blood: a patient with hemoglobin SC disease—a milder form of sickle cell disease—was serving as a stem cell donor for a sibling, the disease so mild it had gone undetected until donor screening, and died of multi-organ failure during G-CSF mobilization. By 2009, Fitzhugh and colleagues had pulled together all 11 cases then in the literature in Cytotherapy: 7 required hospitalization for complications, 2 developed multi-organ failure, and 1 of those died. More unsettling still, they could find no safety threshold—one patient whose HbS had been transfused down to 14.4% still had a crisis requiring hospitalization, while one of those who developed multi-organ failure had a peak leukocyte count of only 6.4×10⁹/L, and a separate patient was hospitalized on a dose as low as 2.5 μg/kg per day. The paper’s title ended in a question: time for a moratorium? The answer, in the end, was yes—today the CASGEVY label carries a line that leaves no room: G-CSF must not be used for mobilization in patients with sickle cell disease.

That leaves plerixafor, and the details matter

In 2018, three groups answered almost simultaneously whether plerixafor alone would do. Boulad and colleagues reported a dose-escalation study in Haematologica: of 15 patients with sickle cell disease, only 8 reached a peripheral blood CD34⁺ concentration above 30 cells/μL, and pushing the dose higher brought no consistent improvement—though 10 of these patients were still on hydroxyurea and only one was chronically transfused, and the authors suggest hydroxyurea may have contributed to the limited mobilization. Two patients developed vaso-occlusive crises, at doses of 80 μg/kg and 240 μg/kg—plerixafor is not free of cost either. In the same issue, Lagresle-Peyrou and colleagues supplied the other half of the answer: three patients with severe disease first underwent exchange transfusion to bring HbS below 30%, then received a single 0.24 mg/kg injection of plerixafor, with no adverse effects other than moderate hypokalemia related to the citrate anticoagulant; they yielded 4.6, 5.8, and 4.5×10⁶ CD34⁺/kg respectively, and—more importantly—those cells expressed high levels of stemness genes and engrafted very efficiently in immunodeficient mice. And Esrick and colleagues pointed out in Blood Adv that what really raises the yield is often not the drug but the craft of the collection itself: in six transfused adults, they used a deep collection interface and began apheresis within four hours of dosing, reaching yields of up to 24.5×10⁶ CD34⁺/kg at a single standard dose, with the product enriched in immunophenotypically defined long-term HSCs.

A mobilization regimen determines not only how many cells you get, but which cells you get.

The sicker the patient, the harder the harvest

Cell yield is not a matter of luck. In 2021, Leonard and colleagues pooled data in Blood Adv from 23 participants with sickle cell disease across two institutions: the median CD34⁺ yield after two or fewer cycles was 4.0×10⁶/kg, ranging from 1.5 to 12.0—an eightfold spread. Yield correlated negatively with age and positively with pre-apheresis peripheral blood CD34 counts, as it does in other populations; but two correlations were unique to this disease: the longer hydroxyurea had been held, the higher the yield (up to five weeks), and yield correlated negatively with markers of disease severity—the more hospitalizations in the preceding year, and the more medications taken for chronic pain, the fewer cells collected. That correlation is uncomfortable to read, because it means the people who are sickest, and who most need a cure, are the hardest to manufacture a drug for.

This is not a paper exercise. In CLIMB SCD-121, the registration trial in patients 12 years and older with sickle cell disease—whose efficacy results were published by Frangoul and colleagues in N Engl J Med in 2024—the collection data recorded in the CASGEVY label read as follows: 63 patients enrolled, 58 started mobilization, and 44 of them (76%) were ultimately infused; the median number of cycles required to manufacture the product plus collect the back-up cells was 2, with a range of 1 to 6; six patients (10%) were unable to receive the therapy because of inadequate cell collection, and in the pediatric trial 1 of 13 dropped out for the same reason. By contrast, when the same drug is used for transfusion-dependent β-thalassemia, G-CSF can be combined with plerixafor: all 59 enrolled patients started mobilization, the median number of cycles was 1, and 52 (88%) were ultimately infused. The same edit, the same production line—and simply because the disease upstream is different, the median length of the first mile doubles; and on the sickle cell side, one in ten never finishes it at all.

A CD34 count is not a measure of potency

There is a more fundamental problem underneath: the very metric used to sign off on the first mile is imperfect. CD34 is a surface marker convenient for flow cytometry and magnetic bead capture, not a measure of long-term reconstituting capacity. Fanconi anemia exposes this most clearly. In 2021, Sevilla and colleagues reported that among 11 patients mobilized with filgrastim plus plerixafor, 9 reached the threshold of 5 CD34⁺ cells/μL in peripheral blood required to initiate apheresis, and the two who did not happened to be the two oldest (15 and 16 years); a median of 4.27×10⁶ CD34⁺/kg was collected in two or three apheresis sessions, which fell sharply to 1.1×10⁶/kg after immunoselection—probably, the authors suggest, because of weak expression of the CD34 antigen on these cells. The same population of cells, captured a different way, and the counted dose drops by three quarters—though the authors also note that these reduced numbers were still sufficient to facilitate the engraftment of corrected cells in non-conditioned patients. This gap will come up again: whether ex vivo expansion can make up the numbers, and how potency should be defined at all, are matters for two other pieces.

Back to the machine humming quietly in the room. It looks like the least technical link in the whole chain—no enzymes, no vectors, no sequencers. Yet it is precisely here that some patients are filtered out, and the reasons for filtering them have nothing to do with genotype: only age, disease severity, medication history, and how many cells the marrow is willing to release that day. Ex vivo gene therapy turned cure into something that can be manufactured, and manufacturing requires raw material; when the raw material can only come from the patient, the patient’s condition becomes part of the production capacity. That is probably one of the most alluring things about the in vivo route: what it aims to abolish is precisely this first mile.

References

  1. CASGEVY (exagamglogene autotemcel) US Prescribing Information. Vertex Pharmaceuticals; revised 07/2026. FDA version · Manufacturer PI
  2. Lévesque JP, Hendy J, Takamatsu Y, Simmons PJ, Bendall LJ. Disruption of the CXCR4/CXCL12 chemotactic interaction during hematopoietic stem cell mobilization induced by GCSF or cyclophosphamide. J Clin Invest. 2003;111(2):187-196. DOI
  3. Broxmeyer HE, Orschell CM, Clapp DW, et al. Rapid mobilization of murine and human hematopoietic stem and progenitor cells with AMD3100, a CXCR4 antagonist. J Exp Med. 2005;201(8):1307-1318. DOI
  4. DiPersio JF, Micallef IN, Stiff PJ, et al. Phase III prospective randomized double-blind placebo-controlled trial of plerixafor plus granulocyte colony-stimulating factor compared with placebo plus granulocyte colony-stimulating factor for autologous stem-cell mobilization and transplantation for patients with non-Hodgkin’s lymphoma. J Clin Oncol. 2009;27(28):4767-4773. DOI
  5. DiPersio JF, Stadtmauer EA, Nademanee A, et al. Plerixafor and G-CSF versus placebo and G-CSF to mobilize hematopoietic stem cells for autologous stem cell transplantation in patients with multiple myeloma. Blood. 2009;113(23):5720-5726. DOI
  6. Abboud M, Laver J, Blau CA. Granulocytosis causing sickle-cell crisis. Lancet. 1998;351(9107):959. DOI
  7. Adler BK, Salzman DE, Carabasi MH, Vaughan WP, Reddy VV, Prchal JT. Fatal sickle cell crisis after granulocyte colony-stimulating factor administration. Blood. 2001;97(10):3313-3314. DOI
  8. 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
  9. 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
  10. Lagresle-Peyrou C, Lefrère F, Magrin E, et al. Plerixafor enables safe, rapid, efficient mobilization of hematopoietic stem cells in sickle cell disease patients after exchange transfusion. Haematologica. 2018;103(5):778-786. DOI
  11. Esrick EB, Manis JP, Daley H, et al. Successful hematopoietic stem cell mobilization and apheresis collection using plerixafor alone in sickle cell patients. Blood Adv. 2018;2(19):2505-2512. DOI
  12. Leonard A, Sharma A, Uchida N, et al. Disease severity impacts plerixafor-mobilized stem cell collection in patients with sickle cell disease. Blood Adv. 2021;5(9):2403-2411. DOI
  13. Sevilla J, Navarro S, Rio P, et al. Improved collection of hematopoietic stem cells and progenitors from Fanconi anemia patients for gene therapy purposes. Mol Ther Methods Clin Dev. 2021;22:66-75. DOI
  14. 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