Yang Liu

← Primer

HSC · Healing Source Code

现货的血——免疫隐身工程已经走进人体,还没轮到造血干细胞 Blood off the Shelf — Immune Cloaking Has Reached Human Patients, but Not Hematopoietic Stem Cells

2026-07-25 · 造血干细胞, 现货疗法, 免疫隐身, iPSC, 异体移植

在血液病房里,时间常常是按”等一个配型”来计的。全相合的供者可能几周就找到,也可能永远找不到。而接受自体基因治疗的患者,等待只是换了个形式——他要等自己的细胞被采集、运走,在千里之外的车间里被编辑、放行,再运回他所在的那张病床。治疗总得围着某一个人重新造一遍。于是有了一个几乎不需要解释的愿望:如果血液也能像药品一样,一批做好、冻在货架上,谁需要谁取一袋。挡在前面的不是制造,而是免疫。

三层看门人

免疫这一关不是一堵墙,是三道门。最外面一道是 T 细胞:它们通过 HLA class I 与 class II 读取一个细胞的身份,读到的不是自己的就发动排斥。最直接的对策是敲掉 B2M——HLA-A、B、C 都要靠它才装配得上膜,B2M 一去,class I 从表面整体消失。

第二道门的逻辑恰好相反。NK 细胞盯的不是”陌生的 HLA”,而是”HLA 不见了”;一个不表达 class I 的细胞在它眼里是典型的 missing self,该杀。2017 年,Gornalusse 等人在 Nat Biotechnol 上给出对策:用 AAV 介导的基因编辑把 HLA-E 单链分子敲进 B2M 位点,让细胞在不表达 HLA-A、B、C 的前提下,仍挂出一个几乎不带多态性的抑制性配体。这样的细胞既不被 CD8⁺ T 细胞当作异体,也不结合抗 HLA 抗体,还抵抗 NK 裂解。

第三道门是巨噬细胞。2019 年,Deuse 等人在 Nat Biotechnol 上把三件事合成一套配方:敲除 MHC class I、敲除 class II 的总开关 CIITA,再过表达 CD47——那个被称作”别吃我”的信号。由这种 hypoimmunogenic iPSC 分化出的内皮细胞、平滑肌细胞和心肌细胞,在完全 MHC 错配的小鼠受体里不用免疫抑制就能长期存活;其中免疫原性最强的人源内皮细胞,还被放进重建有人免疫系统的 BLT 小鼠里检验过。两年后同一批人发现,CD47 的保护并不只针对巨噬细胞:NK 细胞自己也表达 SIRPα,CD47 通过它直接压住 NK 的杀伤——而这条轴种属特异性极强,换个物种,配方未必还成立。

隐身走到了人体

接下来几年,这套配方一级一级往上爬。Hu 等人把同样编辑过的恒河猴细胞肌肉注射进四只完全免疫健全的异体猴,细胞在 16 周内不受限制地存活并分化,未编辑的对照则被迅速排斥;同样编辑过的猴原代胰岛在一只异体猴体内维持了 40 周(2023 年在线、2024 年刊于 Nat Biotechnol)。

然后是人。2025 年,Carlsson 等人在 N Engl J Med 上报告了一名长期 1 型糖尿病男性患者:异体供者胰岛细胞经 CRISPR-Cas12b 编辑与慢病毒转导后注入他的前臂肌肉,全程不用免疫抑制药物,12 周时未见针对这些细胞的免疫反应,基线检测不到的 C-peptide 变得稳定且对葡萄糖有反应(NCT06239636)。他更长时间的随访在 2026 年以通讯形式发回同一本杂志。同一套编辑也上了 CAR-T:2025 年 Cell Stem Cell 上的一份分析显示,两项早期试验里,患者对产品中未被完全编辑、仍带 HLA 的那部分细胞产生了同种免疫反应,而对完全编辑的细胞,所有患者都没有出现新生免疫应答。

把这些成果排成一排,会看到一个容易被忽略的共同点:内皮、平滑肌、心肌、胰岛 β 细胞、T 细胞——没有一个是造血干细胞。

为什么偏偏是造血干细胞

原因藏在造血干细胞的定义里:它的后代就是免疫系统本身。给一块心肌装上隐身外衣,受影响的只有那块心肌;给造血干细胞装上同一件外衣,长出来的每一个淋巴细胞都穿着它。

这不是一句修辞。1990 年,Zijlstra 等人和 Koller 等人分别在 Nature 与 Science 上独立报道:B2M 缺失的小鼠表面几乎检不到 MHC class I,也因此几乎没有 CD8⁺ 细胞毒性 T 细胞——但两篇都强调这些小鼠看上去健康(Zijlstra 写的是可育、看似健康),Koller 那篇的标题本身就写着”发育正常”。把同一个缺失放到人身上,分量却不一样:2015 年 Ardeniz 等人在 J Allergy Clin Immunol 上描述了两名 β2m 缺陷的同胞,论文题目直接把它定性为一种累及固有与适应性免疫的复合免疫缺陷——消失的不只是多态性的 MHC class I,连 CD1a、CD1b、CD1c 与新生儿 Fc 受体也一起从细胞表面不见了,尽管这两名同胞的临床轻重相差很大。CIITA 同样不是可以随手关掉的开关:1993 年 Steimle 等人在 Cell 上通过互补克隆发现,人类遗传性 MHC class II 缺陷——即 bare lymphocyte syndrome——正是 CIITA 突变所致,那是一种严重的原发性免疫缺陷。把这套方案原封不动搬到造血干细胞上,等于在患者身上把这两条通路一起关掉。局部组织可以隐形,一整套免疫系统不能。

其实边界在最早那篇论文里就写着。Gornalusse 等人给自己方案划的适用范围是:分化产物不表达 HLA class II 的场景。而造血干细胞的后代——B 细胞、单核-巨噬细胞、树突状细胞——恰恰是 class II 表达最活跃的那一群。

还有第二重不对称。让细胞对免疫系统隐形,用的是肿瘤早就在用的那套语言:丢掉 MHC class I 是经典的免疫逃逸手段,而升高的 CD47 同样能替肿瘤挡刀——在 Deuse 等人 2021 年那项工作里,过表达 CD47 就保护了 K562 肿瘤细胞不被 SIRPα⁺ NK 细胞杀死。所以现货类细胞产品往往要配一个 kill switch,而这类装置在人体里确实管用——2011 年 Di Stasi 等人在 N Engl J Med 上报告,五名接受单倍体相合移植后回输 iCasp9 修饰供者 T 细胞的患儿中,四名出现 GVHD,单次给予二聚化小分子后 30 分钟内超过九成的修饰细胞被清除,GVHD 终止且未复发。可同样的装置装在造血干细胞上,含义就变了:按下开关意味着摧毁患者的整个造血。

何况隐身只处理了一半方向。它对付的是宿主排斥移植物,而造血干细胞移植还有反方向的 graft-versus-host。让移植物更不容易被清除,并不会让它更不容易伤人。

细胞从哪来,仍然没有答案

即便免疫这一关有解,还有更前面的一关:现货意味着细胞要能被无限地、标准化地造出来,而这几乎只能指望 iPSC。可从多能干细胞造出真正能长期重建造血的 HSC,至今没有稳定实现。

路标一直很清晰:2013 年 Doulatov 等人用转录因子把多能干细胞(hPSC)来源的髓系前体”回改”成多系祖细胞;2014 年 Sturgeon 等人厘清 WNT 信号决定分化走向 definitive 还是 primitive,而只有 definitive 那一支通向真正的 HSC;2016 年 Ng 等人做出带 HOXA 特征、类似主动脉-性腺-中肾区的 hemogenic 血管内皮。2017 年,Sugimura 等人在 Nature 上从 26 个候选转录因子里筛出 7 个,把 hemogenic endothelium 转化成能在初代和二代小鼠受体中重建髓系、B 与 T 细胞的造血干与祖细胞——代价是这 7 个因子由 doxycycline 诱导的慢病毒载体提供,离给人用还隔着一整个安全性问题。

最接近的一次尝试来自 Ng 等人(2024 年在线、2025 年刊于 Nat Biotechnol)。他们不用转基因,靠 retinyl acetate 把中胚层导向 HOXA 特征,再由 BMP4 与 VEGF 指定 hemogenic endothelium、撤掉 VEGF 促成 endothelial-to-hematopoietic transition,收集自发释放进培养基的 CD34⁺ 细胞冻存。两百万个复苏细胞输入免疫缺陷小鼠后,有 25% 到 50% 的受体实现多系骨髓重建,水平与脐带血移植相当。再移植给二代受体时,重建水平普遍很低、多数只剩髓系,只有一只重现了 B、T 与髓系俱全的造血;但同一实验里脐带血对照的二次重建同样低——能传出二次重建的初代受体,iHSC 组是 12 只中 6 只,脐带血组是 5 只中 2 只——作者因此认为,这更像是这套再移植体系本身的限度,而不是 iPSC 来源细胞独有的短板。他们给自己留的余地是:这些细胞在骨髓里建起的干细胞区室功能与脐带血相当,但可能缺少真正 AGM 来源 HSC 那种大幅扩增的能力。

已经在货架上的那部分

在这些都还没解决之前,现实里起作用的是两种折中。一是脐带血——它天然是”半现货”:早已采集冻存、可即时调取。经烟酰胺体外扩增的脐带血产品 omidubicel 在一项 125 例的随机三期试验中把中性粒细胞植入的中位时间从 22 天缩短到 12 天(Horwitz 等人,2021 年,Blood),并于 2023 年 4 月 17 日以 Omisirge 之名获 FDA 批准。它证明”预制、冻存、可调度”在造血里做得到,但仍然需要匹配,仍然需要清髓。

二是绕过而不是取消免疫:2012 年 Taylor 等人在 Cell Stem Cell 上算过,150 名精选的 HLA 纯合志愿者建成的 iPSC 库可以覆盖英国人群的 93%,而所需的免疫抑制降到最低。这是用统计学换掉一部分工程学。

回到开头那间病房。免疫隐身工程这些年真正证明的,是一小块组织可以被做成对免疫系统透明的东西,而且这件事已经在猴子和人身上发生过。它还没有证明的是:当这块组织本身就要长成一套免疫系统时,透明还是不是一个可以要的属性。货架上的血,眼下缺的并不是货架。

参考文献

  1. Gornalusse GG, Hirata RK, Funk SE, et al. HLA-E-expressing pluripotent stem cells escape allogeneic responses and lysis by NK cells. Nat Biotechnol. 2017;35(8):765-772. DOI
  2. Deuse T, Hu X, Gravina A, et al. Hypoimmunogenic derivatives of induced pluripotent stem cells evade immune rejection in fully immunocompetent allogeneic recipients. Nat Biotechnol. 2019;37(3):252-258. DOI(另见 Author Correction. Nat Biotechnol. 2022;40(11):1690. DOI)
  3. Deuse T, Hu X, Agbor-Enoh S, et al. The SIRPα-CD47 immune checkpoint in NK cells. J Exp Med. 2021;218(3):e20200839. DOI
  4. Hu X, White K, Olroyd AG, et al. Hypoimmune induced pluripotent stem cells survive long term in fully immunocompetent, allogeneic rhesus macaques. Nat Biotechnol. 2024;42(3):413-423. DOI
  5. Carlsson PO, Hu X, Scholz H, et al. Survival of Transplanted Allogeneic Beta Cells with No Immunosuppression. N Engl J Med. 2025;393(9):887-894. DOI
  6. Carlsson PO, Hu X, Scholz H, et al. Long-Term Survival of Hypoimmune Allogeneic Islets without Immunosuppression. N Engl J Med. 2026 Jul 10 (online ahead of print). DOI
  7. Hu X, Beauchesne P, Wang C, et al. Hypoimmune CD19 CAR T cells evade allorejection in patients with cancer and autoimmune disease. Cell Stem Cell. 2025;32(9):1356-1368.e4. DOI
  8. Zijlstra M, Bix M, Simister NE, et al. Beta 2-microglobulin deficient mice lack CD4-8+ cytolytic T cells. Nature. 1990;344(6268):742-746. DOI
  9. Koller BH, Marrack P, Kappler JW, Smithies O. Normal development of mice deficient in beta 2M, MHC class I proteins, and CD8+ T cells. Science. 1990;248(4960):1227-1230. DOI
  10. Ardeniz Ö, Unger S, Onay H, et al. β2-Microglobulin deficiency causes a complex immunodeficiency of the innate and adaptive immune system. J Allergy Clin Immunol. 2015;136(2):392-401. DOI
  11. Steimle V, Otten LA, Zufferey M, Mach B. Complementation cloning of an MHC class II transactivator mutated in hereditary MHC class II deficiency (or bare lymphocyte syndrome). Cell. 1993;75(1):135-146. DOI
  12. Di Stasi A, Tey SK, Dotti G, et al. Inducible apoptosis as a safety switch for adoptive cell therapy. N Engl J Med. 2011;365(18):1673-1683. DOI
  13. Doulatov S, Vo LT, Chou SS, et al. Induction of multipotential hematopoietic progenitors from human pluripotent stem cells via respecification of lineage-restricted precursors. Cell Stem Cell. 2013;13(4):459-470. DOI(另见 Erratum. Cell Stem Cell. 2025;32(6):1025. DOI)
  14. Sturgeon CM, Ditadi A, Awong G, et al. Wnt signaling controls the specification of definitive and primitive hematopoiesis from human pluripotent stem cells. Nat Biotechnol. 2014;32(6):554-561. DOI
  15. Ng ES, Azzola L, Bruveris FF, et al. Differentiation of human embryonic stem cells to HOXA+ hemogenic vasculature that resembles the aorta-gonad-mesonephros. Nat Biotechnol. 2016;34(11):1168-1179. DOI
  16. Sugimura R, Jha DK, Han A, et al. Haematopoietic stem and progenitor cells from human pluripotent stem cells. Nature. 2017;545(7655):432-438. DOI
  17. Ng ES, Sarila G, Li JY, et al. Long-term engrafting multilineage hematopoietic cells differentiated from human induced pluripotent stem cells. Nat Biotechnol. 2025;43(8):1274-1287. DOI
  18. Horwitz ME, Stiff PJ, Cutler C, et al. Omidubicel vs standard myeloablative umbilical cord blood transplantation: results of a phase 3 randomized study. Blood. 2021;138(16):1429-1440. DOI
  19. U.S. Food and Drug Administration. FDA approves omidubicel to reduce time to neutrophil recovery and infection in patients with hematologic malignancies(2023 年 4 月 17 日批准). 链接
  20. Taylor CJ, Peacock S, Chaudhry AN, et al. Generating an iPSC bank for HLA-matched tissue transplantation based on known donor and recipient HLA types. Cell Stem Cell. 2012;11(2):147-152. DOI

On a hematology ward, time is often measured in waiting for a match. A fully matched donor may turn up in a few weeks, or never. And for a patient receiving autologous gene therapy, the waiting only changes shape: he waits for his own cells to be collected, shipped away, edited and released in a facility a thousand miles off, and shipped back to the bed he is lying in. The treatment always has to be built again around one particular person. Hence a wish that hardly needs explaining: if blood could be made like a drug — one batch, frozen on a shelf, a bag for whoever needs it. What stands in the way is not manufacturing. It is immunity.

Three gatekeepers

Immunity here is not one wall but three doors. The outermost is the T cell: it reads a cell’s identity through HLA class I and class II, and rejects whatever it does not read as self. The most direct countermeasure is to knock out B2M — HLA-A, B and C all need it to assemble at the membrane, so once B2M is gone, class I disappears from the surface altogether.

The second door works by the opposite logic. NK cells watch not for unfamiliar HLA but for HLA that has gone missing; a cell with no class I is textbook missing self, and should be killed. In 2017, Gornalusse and colleagues offered a countermeasure in Nat Biotechnol: adeno-associated virus (AAV)-mediated gene editing to knock HLA-E single-chain molecules into the B2M locus, so that a cell displays a minimally polymorphic inhibitory ligand while expressing no HLA-A, B or C. Such cells are not recognized as allogeneic by CD8⁺ T cells, do not bind anti-HLA antibodies, and resist NK-mediated lysis.

The third door is the macrophage. In 2019, Deuse and colleagues combined three moves into one recipe in Nat Biotechnol: inactivate MHC class I, inactivate CIITA — the master switch for class II — and overexpress CD47, the signal that tells a macrophage not to eat the cell. Endothelial cells, smooth muscle cells and cardiomyocytes derived from such hypoimmunogenic iPSCs survived long term in fully MHC-mismatched mouse recipients without immunosuppression; the human endothelial cells, the most immunogenic of these derivatives, were further tested in BLT mice reconstituted with a human immune system. Two years later the same group found that CD47’s protection was not aimed only at macrophages: NK cells themselves express SIRPα, and CD47 acts through it to suppress NK killing directly — and that axis turned out to be highly species specific, so the recipe does not necessarily carry across species.

Cloaking reaches human patients

Over the next few years the recipe climbed the ladder. Hu and colleagues injected identically edited rhesus macaque cells into the muscle of four fully immunocompetent allogeneic macaques; the cells survived unrestricted for 16 weeks and differentiated, while unedited controls were vigorously rejected. Edited primary macaque islets survived 40 weeks in one allogeneic macaque recipient (online in 2023, in print in Nat Biotechnol in 2024).

Then came humans. In 2025, Carlsson and colleagues reported in N Engl J Med on a man with long-standing type 1 diabetes: allogeneic donor islet cells, edited with CRISPR-Cas12b and lentivirally transduced, were transplanted into his forearm muscle. He received no immunosuppressive drugs at all; at 12 weeks there was no immune response against the cells, and C-peptide, undetectable at baseline, became stable and glucose-responsive (NCT06239636). Longer follow-up was returned to the same journal as correspondence in 2026. The same editing has also been applied to CAR-T: an analysis published in Cell Stem Cell in 2025 showed that across two early-phase trials, patients mounted an alloimmune response against the HLA-replete subpopulations of the product, while against fully edited cells no de novo immune response was seen in any patient.

Line these results up and a shared feature is easy to miss: endothelium, smooth muscle, cardiomyocytes, pancreatic β cells, T cells — not one of them is a hematopoietic stem cell.

Why hematopoietic stem cells are the hard case

The reason is buried in the definition of a hematopoietic stem cell: its progeny are the immune system. Put a cloak on a patch of heart muscle and only that patch is affected; put the same cloak on a hematopoietic stem cell and every lymphocyte it produces is wearing it.

This is not a figure of speech. In 1990, Zijlstra and colleagues and Koller and colleagues reported independently, in Nature and in Science, that B2M-deficient mice have almost no detectable surface MHC class I and consequently almost no CD8⁺ cytotoxic T cells — yet both papers stressed that the animals looked healthy: Zijlstra’s describes them as “fertile and apparently healthy”, and the title of Koller’s paper begins with the words “Normal development”. In humans, the same deficiency does not weigh the same. In 2015, Ardeniz and colleagues described two siblings with β2m deficiency in J Allergy Clin Immunol, and the title of that paper states it plainly: β2-microglobulin deficiency causes a complex immunodeficiency of the innate and adaptive immune system. What vanished from the cell surface was not only polymorphic MHC class I but also CD1a, CD1b, CD1c and the neonatal Fc receptor — though the two siblings differed widely in how ill they were. CIITA is likewise not a switch to be flipped casually: in 1993, Steimle and colleagues reported in Cell, via complementation cloning, that hereditary MHC class II deficiency — bare lymphocyte syndrome — is caused by mutations in CIITA, and described it as a form of severe primary immunodeficiency. Transplanting this recipe unchanged into hematopoietic stem cells would mean shutting both pathways down in a patient. A local tissue can be made invisible; an entire immune system cannot.

The boundary was in fact written into the earliest of these papers. Gornalusse and colleagues framed their own approach as a potential source of universal donor cells for applications in which the differentiated derivatives lack HLA class II expression. And the progeny of a hematopoietic stem cell — B cells, monocytes and macrophages, dendritic cells — are precisely the most active expressers of class II.

There is a second asymmetry. Hiding a cell from the immune system uses the language tumors have long spoken: losing MHC class I is a classic route of immune escape, and elevated CD47 likewise shields tumor cells — in the 2021 work by Deuse and colleagues, raised CD47 expression protected K562 tumor cells against SIRPα⁺ primary NK cells. So off-the-shelf cell products usually come with a kill switch, and such devices do work in human beings. In 2011, Di Stasi and colleagues reported in N Engl J Med that among five children who received iCasp9-modified donor T cells after haploidentical transplantation, four developed GVHD; a single dose of a dimerizing drug eliminated more than 90% of the modified cells within 30 minutes and ended the GVHD without recurrence. Install the same device in a hematopoietic stem cell and its meaning changes: pressing the switch means destroying the patient’s entire hematopoiesis.

And cloaking only handles one direction. It addresses host rejection of the graft, whereas hematopoietic stem cell transplantation also has graft-versus-host running the other way. Making a graft harder to clear does not make it less able to do harm.

Where the cells come from, still unanswered

Even if immunity were solved, an earlier problem sits upstream: off-the-shelf means cells must be manufacturable without limit and to a standard, and for that there is essentially only iPSCs to hope for. Yet making, from pluripotent stem cells, HSCs that truly reconstitute hematopoiesis long term has still not been reliably achieved.

The waypoints have been clear enough. In 2013, Doulatov and colleagues used transcription factors to respecify myeloid precursors derived from human pluripotent stem cells (hPSCs) into multilineage progenitors. In 2014, Sturgeon and colleagues established that WNT signaling determines whether differentiation goes definitive or primitive, and that only the definitive branch leads to true HSCs. In 2016, Ng and colleagues generated HOXA⁺ hemogenic vasculature resembling the aorta-gonad-mesonephros region. In 2017, Sugimura and colleagues reported in Nature a screen of 26 candidate transcription factors that recovered seven, converting hemogenic endothelium into hematopoietic stem and progenitor cells that engrafted myeloid, B and T cells in primary and secondary mouse recipients — at the cost that those seven factors were delivered by a doxycycline-inducible lentiviral vector, which leaves an entire safety problem between this and human use.

The closest attempt so far comes from Ng and colleagues (online in 2024, in print in Nat Biotechnol in 2025). Without transgenes, they used retinyl acetate to pattern mesoderm toward a HOXA signature, specified hemogenic endothelium with BMP4 and VEGF, removed VEGF to drive the endothelial-to-hematopoietic transition, and cryopreserved the CD34⁺ cells released spontaneously into the medium. Two million thawed cells transplanted into immune-deficient mice produced multilineage bone marrow engraftment in 25% to 50% of recipients, at levels similar to those achieved with umbilical cord blood transplantation. On transplantation into secondary recipients, engraftment was generally at a low level and in most cases restricted to myeloid lineages, with only one recipient displaying B, T and myeloid lineages; but secondary engraftment from the cord blood controls in the same experiment was equally low — of the primary recipients, 6 of 12 in the iHSC arm and 2 of 5 in the cord blood arm went on to yield secondary engraftment — which led the authors to conclude that this was not a finding restricted to iPS cell-derived hematopoietic cells but rather a limit of the serial transplantation system itself. The hedge they left themselves: these cells generate a bone marrow stem cell compartment that functions similarly to cord blood, but may lack the capacity for the substantial expansion that marks true AGM-derived HSCs.

The part already on the shelf

Until all of that is solved, what works in practice are two compromises. One is cord blood, which is naturally half off-the-shelf: already collected and frozen, available on demand. Omidubicel, a cord blood product expanded ex vivo with nicotinamide, shortened the median time to neutrophil engraftment from 22 days to 12 days in a randomized phase 3 study of 125 patients (Horwitz and colleagues, 2021, Blood), and was approved by the FDA on April 17, 2023, as Omisirge. It proves that prefabricated, frozen and schedulable is achievable in hematopoiesis — but it still requires matching, and it still requires myeloablative conditioning.

The other is to bypass immunity rather than abolish it. In 2012, Taylor and colleagues calculated in Cell Stem Cell that an iPSC bank built from 150 selected HLA-homozygous volunteers could match 93% of the UK population with a minimal requirement for immunosuppression. That trades a piece of engineering for a piece of statistics.

Back to the ward we started in. What immune cloaking has genuinely demonstrated over these years is that a small piece of tissue can be made transparent to the immune system, and that this has already happened in macaques and in human beings. What it has not demonstrated is whether transparency is still a property you can ask for when the tissue in question is itself going to grow into an immune system. Blood on the shelf is not, at the moment, short of a shelf.

References

  1. Gornalusse GG, Hirata RK, Funk SE, et al. HLA-E-expressing pluripotent stem cells escape allogeneic responses and lysis by NK cells. Nat Biotechnol. 2017;35(8):765-772. DOI
  2. Deuse T, Hu X, Gravina A, et al. Hypoimmunogenic derivatives of induced pluripotent stem cells evade immune rejection in fully immunocompetent allogeneic recipients. Nat Biotechnol. 2019;37(3):252-258. DOI (see also Author Correction. Nat Biotechnol. 2022;40(11):1690. DOI)
  3. Deuse T, Hu X, Agbor-Enoh S, et al. The SIRPα-CD47 immune checkpoint in NK cells. J Exp Med. 2021;218(3):e20200839. DOI
  4. Hu X, White K, Olroyd AG, et al. Hypoimmune induced pluripotent stem cells survive long term in fully immunocompetent, allogeneic rhesus macaques. Nat Biotechnol. 2024;42(3):413-423. DOI
  5. Carlsson PO, Hu X, Scholz H, et al. Survival of Transplanted Allogeneic Beta Cells with No Immunosuppression. N Engl J Med. 2025;393(9):887-894. DOI
  6. Carlsson PO, Hu X, Scholz H, et al. Long-Term Survival of Hypoimmune Allogeneic Islets without Immunosuppression. N Engl J Med. 2026 Jul 10 (online ahead of print). DOI
  7. Hu X, Beauchesne P, Wang C, et al. Hypoimmune CD19 CAR T cells evade allorejection in patients with cancer and autoimmune disease. Cell Stem Cell. 2025;32(9):1356-1368.e4. DOI
  8. Zijlstra M, Bix M, Simister NE, et al. Beta 2-microglobulin deficient mice lack CD4-8+ cytolytic T cells. Nature. 1990;344(6268):742-746. DOI
  9. Koller BH, Marrack P, Kappler JW, Smithies O. Normal development of mice deficient in beta 2M, MHC class I proteins, and CD8+ T cells. Science. 1990;248(4960):1227-1230. DOI
  10. Ardeniz Ö, Unger S, Onay H, et al. β2-Microglobulin deficiency causes a complex immunodeficiency of the innate and adaptive immune system. J Allergy Clin Immunol. 2015;136(2):392-401. DOI
  11. Steimle V, Otten LA, Zufferey M, Mach B. Complementation cloning of an MHC class II transactivator mutated in hereditary MHC class II deficiency (or bare lymphocyte syndrome). Cell. 1993;75(1):135-146. DOI
  12. Di Stasi A, Tey SK, Dotti G, et al. Inducible apoptosis as a safety switch for adoptive cell therapy. N Engl J Med. 2011;365(18):1673-1683. DOI
  13. Doulatov S, Vo LT, Chou SS, et al. Induction of multipotential hematopoietic progenitors from human pluripotent stem cells via respecification of lineage-restricted precursors. Cell Stem Cell. 2013;13(4):459-470. DOI (see also Erratum. Cell Stem Cell. 2025;32(6):1025. DOI)
  14. Sturgeon CM, Ditadi A, Awong G, et al. Wnt signaling controls the specification of definitive and primitive hematopoiesis from human pluripotent stem cells. Nat Biotechnol. 2014;32(6):554-561. DOI
  15. Ng ES, Azzola L, Bruveris FF, et al. Differentiation of human embryonic stem cells to HOXA+ hemogenic vasculature that resembles the aorta-gonad-mesonephros. Nat Biotechnol. 2016;34(11):1168-1179. DOI
  16. Sugimura R, Jha DK, Han A, et al. Haematopoietic stem and progenitor cells from human pluripotent stem cells. Nature. 2017;545(7655):432-438. DOI
  17. Ng ES, Sarila G, Li JY, et al. Long-term engrafting multilineage hematopoietic cells differentiated from human induced pluripotent stem cells. Nat Biotechnol. 2025;43(8):1274-1287. DOI
  18. Horwitz ME, Stiff PJ, Cutler C, et al. Omidubicel vs standard myeloablative umbilical cord blood transplantation: results of a phase 3 randomized study. Blood. 2021;138(16):1429-1440. DOI
  19. U.S. Food and Drug Administration. FDA approves omidubicel to reduce time to neutrophil recovery and infection in patients with hematologic malignancies (approved April 17, 2023). Link
  20. Taylor CJ, Peacock S, Chaudhry AN, et al. Generating an iPSC bank for HLA-matched tissue transplantation based on known donor and recipient HLA types. Cell Stem Cell. 2012;11(2):147-152. DOI