Yang Liu

← Primer

HSC · Healing Source Code

同叫 CHIP,各有各的选择压力——DNMT3A、TET2、ASXL1、TP53 与 PPM1D 的分野 Same CHIP label, different selection pressures — how DNMT3A, TET2, ASXL1, TP53 and PPM1D diverge

2026-10-09 · CHIP, clonal hematopoiesis, DNMT3A, TET2, ASXL1, TP53, PPM1D, clonal dynamics, gene therapy safety

把一个人血液里的克隆统称为 CHIP,就像把所有”咳嗽”归为一种病。2014 年两篇 N Engl J Med 论文让这个现象被看见:它随年龄变得常见,多数突变落在几个相同的基因上。但此后的工作越来越清楚地显示,同一个标签底下,是生长方式不同、被不同力量选择、通向不同后果的克隆。对要把造血干细胞取出来、改造、再输回去的领域,这个分野与”有没有 CHIP”同样值得追问。

同一个筐,不同的生长速度

2014 年,Jaiswal 等人和 Genovese 等人各自分析了并非为此设计的外周血外显子测序数据(分别为 17,182 人和 12,380 人),发现随年龄增长而变得普遍的体细胞克隆,多数突变落在 DNMT3A、TET2 和 ASXL1 三个基因上。此后,TP53 与 PPM1D 也被反复报告为常见的 CHIP 基因。这五个基因常被放进同一张列表,克隆行为却不同。

最直接的证据来自 Fabre 等人 2022 年在 Nature 上的纵向研究:他们追踪了 385 名 55 岁及以上的个体、共 697 支克隆,中位随访 13 年。92.4% 的克隆以稳定的指数速率扩张,不同突变的速率差别很大:DNMT3A 与 TP53 突变克隆平均每年约增长 5%,TET2、ASXL1、PPM1D(以及 SF3B1)约为其两倍,即每年约 10%;SRSF2、PTPN11、U2AF1 为每年 15%–20%,SRSF2 P95H 这一个具体突变超过每年 50%。同一突变的克隆之间还有约 ±5% 每年的差异,所以这些数字是基因层面的平均,不是每支克隆的速度。

时间维度上也不同:DNMT3A 突变克隆倾向于在生命早期扩张,到老年因 oligoclonal 格局拥挤而放缓;TET2 突变克隆各年龄段都会出现;剪接因子突变则到生命后期才驱动扩张。Watson 等人 2020 年在 Science 上用约 50,000 人的血液测序数据得出,塑造克隆造血的主要力量是 positive selection 而非 genetic drift。

各自被什么推着长

速率的差别背后是选择压力的差别。以下多数线索来自小鼠或癌症患者队列。

DNMT3A 的线索来自环境。Hormaechea-Agulla 等人 2021 年在 Cell Stem Cell 上用 Dnmt3a 嵌合小鼠发现,慢性分枝杆菌感染期间 Dnmt3a 缺失的造血干细胞显著扩增,单独注射 IFNγ 即足以复现这一现象。

TET2 的特点是与炎症缠在一起。2018 年 Meisel 等人在 Nature 上报告,Tet2 缺失小鼠的 pre-leukaemic myeloproliferation 依赖肠道屏障失效后的细菌易位与 IL-6,germ-free 的 Tet2 缺失小鼠则不出现这一表型;Cai 等人同年在 Cell Stem Cell 上显示,炎症应激下 Tet2 缺失的造血干/祖细胞经 IL-6 获得抗凋亡优势。

ASXL1 的线索来自暴露史。Bolton 等人 2020 年在 Nat Genet 上分析接受过治疗的癌症患者,发现 ASXL1 突变在现吸烟者或曾吸烟者中富集,是一个相关性发现。

TP53 与 PPM1D 则与细胞毒治疗的选择联系在一起。2015 年 Wong 等人在 Nature 上对 22 例治疗相关 AML 做了基因组测序,其中 4 例在诊断时发现的那个 TP53 突变,3–6 年前就以 0.003%–0.7% 的低频存在于动员血或骨髓中(2 例在任何化疗之前即已检出);在小鼠嵌合体里,Tp53⁺ᐟ⁻ 造血干/祖细胞在化疗后优先扩张。作者的解读是,化疗并不直接诱发 TP53 突变,而是让本来就存在的克隆占了上风。Hsu 等人 2018 年在 Cell Stem Cell 上发现,PPM1D 突变见于约五分之一的治疗相关 AML/MDS 患者,并与顺铂暴露强相关;在小鼠体内,突变细胞在顺铂和阿霉素之后胜过野生型,而在细胞系实验里,主要机制是抵抗凋亡。Bolton 的研究则显示,放疗、铂类和 topoisomerase II 抑制剂偏好选择 DNA damage response 基因(TP53、PPM1D、CHEK2)的突变,连续取样提示这些克隆在特定治疗下胜出。

TP53 内部也不均一。Boettcher 等人 2019 年在 Science 上表明,DNA 结合域内的 missense 突变通过 dominant-negative effect 获得选择优势;Fabre 的数据与此一致:TP53 missense 克隆平均每年比 truncating 克隆快约 10 个百分点,而 truncating 克隆通常不扩张。

不同的后果

后果同样因基因而异。2017 年 Jaiswal 等人在 N Engl J Med 上指出,DNMT3A、TET2、ASXL1 和 JAK2 的突变各自都与冠心病相关;因果证据则主要来自 TET2:Tet2 缺失的骨髓移植进易发动脉粥样硬化的小鼠后,斑块更大。

关于血液肿瘤,Weeks 等人 2023 年在 NEJM Evid 上用 438,890 名 UK Biobank 参与者的外显子数据构建了 clonal hematopoiesis risk score(CHRS)。其中 DNMT3A 的位置颇有意味:与没有 CHIP/CCUS 的人相比,DNMT3A 突变者髓系肿瘤风险升高 4.25 倍;但在 CHIP/CCUS 内部比较,单个 DNMT3A 突变者的风险显著低于其他基因型(HR 0.188)。TP53 则需要照实写:该队列里没有观察到 TP53 相关突变与髓系肿瘤的关联,CHRS 的作者是基于既往数据显示的高外显率与不良预后,经验性地把它放进了高危清单。方向相近的是 Desai 等人 2018 年在 Nat Med 上的研究:在 Women’s Health Initiative 里回溯 212 名后来发生 AML 的女性,检出 TP53 突变的 21 人全部最终发展成了 AML——但这是从已发病的人倒推,不能读成携带者的普遍结局。

Fabre 还发现,在 CHIP 驱动基因之间,克隆生长越快,AML 风险越高(adjusted R² = 0.55)。DNMT3A 的慢速与较低风险符合这一趋势;TP53 在 Fabre 的数据里同样是慢速基因,却被上述两项研究以不同方式指向高危。并置来看,速率只是风险的一个维度。

落在基因治疗的随访上

这些差异落到基因治疗上,最具体的数据来自 Spencer Chapman 等人 2023 年在 Nat Med 上的研究。他们用 whole-genome sequencing 追踪了 6 名镰刀型贫血患者治疗前后的造血干细胞:治疗后没有发现克隆性扩张,但与髓系肿瘤或克隆造血相关的潜在 driver mutation 频率上升,以 DNMT3A 和 EZH2 突变克隆为著,并且在基因修饰和未修饰的细胞中都出现。集落中携带潜在 driver mutation 的比例从治疗前的 1/1,161(0.1%)升到治疗后的 12/1,431(0.8%)。在 9 支含 driver mutation 的克隆里,有 5 支(DNMT3A、PPM1D、TP53 各一,EZH2 两个)在治疗后显著增大:两个 EZH2 克隆至少增加 55 倍和 95 倍,DNMT3A、PPM1D、TP53 克隆则分别为 1.24 倍、2.7 倍和 4.5 倍。作者同时指出,这些增幅都低于随访所用临床靶向 panel 的检测灵敏度,临床意义尚不清楚。

这六名患者的数据至少提示两点。第一,不同基因的克隆在基因治疗后的扩张幅度相差悬殊,把所有 CHIP 克隆当作一个整体去追踪会抹平这种差别。第二,“哪种预处理偏好选哪种克隆”需要按具体方案回答:Bolton 的结果来自放疗、铂类和 topoisomerase II 抑制剂,Hsu 在小鼠里看到的 PPM1D 优势出现在顺铂和阿霉素之后,而不是骨髓移植后的恢复期,都不能直接外推到基因治疗所用的清髓方案。

还没有答案的部分

缺的恰恰是能把这些线索接在一起的数据:基因治疗人群里各基因克隆的基线分布、不同预处理方案各自偏好什么、克隆在多年随访里是继续扩张还是停下。已有的工作分散在癌症患者、一般人群和很小的基因治疗队列里,拼出的是一张有方向、但分辨率不够的地图。

CHIP 这个名字让人承认血液里的克隆是常态,局限在于把几种不同的生物学压进了同一个词。DNMT3A 克隆早早起步、后来慢下来,TET2 与炎症相关,ASXL1 与吸烟史相关,TP53 与 PPM1D 在化疗后占上风。把它们分开看,长期随访里的每一条曲线才有可以被解读的坐标。

参考文献

  1. Jaiswal S, et al. Age-related clonal hematopoiesis associated with adverse outcomes. N Engl J Med. 2014;371(26):2488-2498. DOI
  2. Genovese G, et al. Clonal hematopoiesis and blood-cancer risk inferred from blood DNA sequence. N Engl J Med. 2014;371(26):2477-2487. DOI
  3. Fabre MA, et al. The longitudinal dynamics and natural history of clonal haematopoiesis. Nature. 2022;606(7913):335-342. DOI
  4. Watson CJ, et al. The evolutionary dynamics and fitness landscape of clonal hematopoiesis. Science. 2020;367(6485):1449-1454. DOI
  5. Hormaechea-Agulla D, et al. Chronic infection drives Dnmt3a-loss-of-function clonal hematopoiesis via IFNγ signaling. Cell Stem Cell. 2021;28(8):1428-1442.e6. DOI
  6. Meisel M, et al. Microbial signals drive pre-leukaemic myeloproliferation in a Tet2-deficient host. Nature. 2018;557(7706):580-584. DOI
  7. Cai Z, et al. Inhibition of inflammatory signaling in Tet2 mutant preleukemic cells mitigates stress-induced abnormalities and clonal hematopoiesis. Cell Stem Cell. 2018;23(6):833-849.e5. DOI
  8. Bolton KL, et al. Cancer therapy shapes the fitness landscape of clonal hematopoiesis. Nat Genet. 2020;52(11):1219-1226. DOI
  9. Wong TN, et al. Role of TP53 mutations in the origin and evolution of therapy-related acute myeloid leukaemia. Nature. 2015;518(7540):552-555. DOI
  10. Hsu JI, et al. PPM1D mutations drive clonal hematopoiesis in response to cytotoxic chemotherapy. Cell Stem Cell. 2018;23(5):700-713.e6. DOI
  11. Boettcher S, et al. A dominant-negative effect drives selection of TP53 missense mutations in myeloid malignancies. Science. 2019;365(6453):599-604. DOI
  12. Jaiswal S, et al. Clonal hematopoiesis and risk of atherosclerotic cardiovascular disease. N Engl J Med. 2017;377(2):111-121. DOI
  13. Weeks LD, et al. Prediction of risk for myeloid malignancy in clonal hematopoiesis. NEJM Evid. 2023;2(5):EVIDoa2200310. DOI
  14. Desai P, et al. Somatic mutations precede acute myeloid leukemia years before diagnosis. Nat Med. 2018;24(7):1015-1023. DOI
  15. Spencer Chapman M, et al. Clonal selection of hematopoietic stem cells after gene therapy for sickle cell disease. Nat Med. 2023;29(12):3175-3183. DOI

Calling every clone in a person’s blood CHIP is like filing every cough under one disease. Two N Engl J Med papers in 2014 made the phenomenon visible: it becomes common with age, and most of the mutations fall in a handful of the same genes. But later work has shown, more and more clearly, that under this one label sit clones that grow differently, are selected by different forces and lead to different outcomes. For a field that takes hematopoietic stem cells out, engineers them and puts them back, this divide deserves as much attention as whether CHIP is present at all.

One basket, different growth rates

In 2014, Jaiswal et al. and Genovese et al. each analyzed peripheral-blood exome sequencing data that had not been generated for this purpose (17,182 and 12,380 people respectively) and found somatic clones that become common with age, with most of the mutations falling in three genes: DNMT3A, TET2 and ASXL1. Since then, TP53 and PPM1D have also been repeatedly reported as common CHIP genes. These five genes are often put on one list, yet the clones behave differently.

The most direct evidence comes from the 2022 longitudinal study by Fabre et al. in Nature. They followed 385 individuals aged 55 and older, 697 clones in all, for a median of 13 years. 92.4% of clones expanded at a stable exponential rate, and the rates differed widely between mutations: DNMT3A and TP53 mutant clones grew by about 5% a year on average; TET2, ASXL1 and PPM1D (and SF3B1) grew at roughly twice that, about 10% a year; SRSF2, PTPN11 and U2AF1 grew at 15%–20% a year, and one specific mutation, SRSF2 P95H, at more than 50% a year. Clones carrying the same mutation also differed from one another by about ±5% a year, so these figures are gene-level averages, not the speed of any individual clone.

The time dimension differs as well: DNMT3A mutant clones tend to expand early in life and slow down in old age as the oligoclonal landscape becomes crowded; TET2 mutant clones arise at all ages; splice-factor mutations drive expansion only late in life. Watson et al., in Science in 2020, drew on blood sequencing data from about 50,000 people and concluded that the main force shaping clonal hematopoiesis is positive selection rather than genetic drift.

What pushes each one

Behind the differences in speed are differences in selection pressure. Most of the clues below come from mice or from cohorts of cancer patients.

The clue for DNMT3A comes from the environment. Hormaechea-Agulla et al., in Cell Stem Cell in 2021, used Dnmt3a chimeric mice and found that Dnmt3a-deficient hematopoietic stem cells expanded markedly during chronic mycobacterial infection, and that injecting IFNγ alone was enough to reproduce this.

TET2 is characterized by its entanglement with inflammation. In 2018, Meisel et al. reported in Nature that the pre-leukaemic myeloproliferation of Tet2-deficient mice depends on bacterial translocation after gut barrier failure and on IL-6, and that germ-free Tet2-deficient mice do not show the phenotype; in the same year, Cai et al. showed in Cell Stem Cell that under inflammatory stress, Tet2-deficient hematopoietic stem/progenitor cells gain an anti-apoptotic advantage through IL-6.

The clue for ASXL1 comes from exposure history. Bolton et al., in Nat Genet in 2020, analyzed patients who had received treatment for cancer and found ASXL1 mutations enriched among current or former smokers, a correlative finding.

TP53 and PPM1D are tied to selection under cytotoxic therapy. In 2015, Wong et al. sequenced the genomes of 22 cases of therapy-related AML in Nature; in 4 of them, the TP53 mutation found at diagnosis had been present 3–6 years earlier at a low frequency of 0.003%–0.7% in mobilized blood or bone marrow (in 2 cases it was detected before any chemotherapy). In mouse chimeras, Tp53+/- hematopoietic stem/progenitor cells expanded preferentially after chemotherapy. The authors’ reading was that chemotherapy does not directly induce TP53 mutations but lets clones that were already there gain the upper hand. Hsu et al., in Cell Stem Cell in 2018, found PPM1D mutations in about one fifth of patients with therapy-related AML/MDS, strongly associated with cisplatin exposure; in mice, the mutant cells outcompeted wild-type cells after cisplatin and doxorubicin, while in cell-line experiments the main mechanism was resistance to apoptosis. Bolton’s study showed that radiotherapy, platinum agents and topoisomerase II inhibitors favor selection of mutations in DNA damage response genes (TP53, PPM1D, CHEK2), and sequential sampling suggested that these clones win out under particular treatments.

TP53 is not uniform within itself either. Boettcher et al., in Science in 2019, showed that missense mutations in the DNA-binding domain gain a selective advantage through a dominant-negative effect; Fabre’s data are consistent with this: TP53 missense clones grew on average about 10 percentage points a year faster than truncating clones, while truncating clones generally did not expand.

Different consequences

The consequences also differ by gene. In 2017, Jaiswal et al. reported in N Engl J Med that mutations in DNMT3A, TET2, ASXL1 and JAK2 are each associated with coronary heart disease; the causal evidence comes mainly from TET2, where bone marrow transplanted from Tet2-deficient mice into atherosclerosis-prone mice produced larger plaques.

For hematologic malignancy, Weeks et al., in NEJM Evid in 2023, used exome data from 438,890 UK Biobank participants to build a clonal hematopoiesis risk score (CHRS). The position of DNMT3A in it is telling: compared with people without CHIP/CCUS, those with DNMT3A mutations had a 4.25-fold higher risk of myeloid malignancy; but within CHIP/CCUS, people with a single DNMT3A mutation had a significantly lower risk than other genotypes (HR 0.188). TP53 has to be reported as it stands: in that cohort no association between TP53-related mutations and myeloid malignancy was observed, and the CHRS authors placed it on the high-risk list empirically, on the basis of the high penetrance and poor prognosis shown in earlier data. Similar in direction is Desai et al., in Nat Med in 2018: in the Women’s Health Initiative they looked back at 212 women who later developed AML, and all 21 in whom a TP53 mutation was detected eventually developed AML — but this reasons backward from people who had already fallen ill, and cannot be read as the typical outcome for carriers.

Fabre also found that, across CHIP driver genes, the faster the clonal growth, the higher the AML risk (adjusted R² = 0.55). The slow growth of DNMT3A fits this trend with its lower risk; TP53 is likewise a slow gene in Fabre’s data, yet the two studies above point to it as high-risk in different ways. Set side by side, growth rate is only one dimension of risk.

Where this lands in gene therapy follow-up

The most concrete data on how these differences bear on gene therapy come from Spencer Chapman et al., in Nat Med in 2023. Using whole-genome sequencing, they tracked the hematopoietic stem cells of 6 patients with sickle cell disease before and after treatment: no clonal expansion was found after treatment, but the frequency of potential driver mutations linked to myeloid malignancy or clonal hematopoiesis rose, most notably in DNMT3A and EZH2 mutant clones, and this occurred in both gene-modified and unmodified cells. The proportion of colonies carrying a potential driver mutation rose from 1/1,161 (0.1%) before treatment to 12/1,431 (0.8%) after. Of 9 clones carrying driver mutations, 5 (one each of DNMT3A, PPM1D and TP53, and two of EZH2) grew markedly after treatment: the two EZH2 clones increased at least 55-fold and 95-fold, and the DNMT3A, PPM1D and TP53 clones by 1.24-, 2.7- and 4.5-fold respectively. The authors also noted that these increases were all below the detection sensitivity of the clinical targeted panel used in follow-up, and that their clinical significance is unclear.

The data from these six patients suggest at least two things. First, clones of different genes expand by very different amounts after gene therapy, and tracking all CHIP clones as one group would flatten this difference. Second, which clone a given conditioning regimen favors has to be answered regimen by regimen: Bolton’s results come from radiotherapy, platinum agents and topoisomerase II inhibitors, and the PPM1D advantage Hsu saw in mice appeared after cisplatin and doxorubicin, not in the recovery period after bone marrow transplantation; neither can be extrapolated directly to the myeloablative regimens used in gene therapy.

What remains unanswered

What is missing is precisely the data that could join these clues together: the baseline distribution of clones by gene in the gene therapy population, what each conditioning regimen favors, and whether clones keep expanding or stop over years of follow-up. The existing work is scattered across cancer patients, the general population and very small gene therapy cohorts, and what it adds up to is a map with a direction but not enough resolution.

The name CHIP makes us accept that clones in the blood are the norm; its limitation is that it presses several different biologies into one word. DNMT3A clones start early and later slow down, TET2 is tied to inflammation, ASXL1 to smoking history, and TP53 and PPM1D gain the upper hand after chemotherapy. Look at them separately, and every curve in long-term follow-up gains coordinates by which it can be read.

References

  1. Jaiswal S, et al. Age-related clonal hematopoiesis associated with adverse outcomes. N Engl J Med. 2014;371(26):2488-2498. DOI
  2. Genovese G, et al. Clonal hematopoiesis and blood-cancer risk inferred from blood DNA sequence. N Engl J Med. 2014;371(26):2477-2487. DOI
  3. Fabre MA, et al. The longitudinal dynamics and natural history of clonal haematopoiesis. Nature. 2022;606(7913):335-342. DOI
  4. Watson CJ, et al. The evolutionary dynamics and fitness landscape of clonal hematopoiesis. Science. 2020;367(6485):1449-1454. DOI
  5. Hormaechea-Agulla D, et al. Chronic infection drives Dnmt3a-loss-of-function clonal hematopoiesis via IFNγ signaling. Cell Stem Cell. 2021;28(8):1428-1442.e6. DOI
  6. Meisel M, et al. Microbial signals drive pre-leukaemic myeloproliferation in a Tet2-deficient host. Nature. 2018;557(7706):580-584. DOI
  7. Cai Z, et al. Inhibition of inflammatory signaling in Tet2 mutant preleukemic cells mitigates stress-induced abnormalities and clonal hematopoiesis. Cell Stem Cell. 2018;23(6):833-849.e5. DOI
  8. Bolton KL, et al. Cancer therapy shapes the fitness landscape of clonal hematopoiesis. Nat Genet. 2020;52(11):1219-1226. DOI
  9. Wong TN, et al. Role of TP53 mutations in the origin and evolution of therapy-related acute myeloid leukaemia. Nature. 2015;518(7540):552-555. DOI
  10. Hsu JI, et al. PPM1D mutations drive clonal hematopoiesis in response to cytotoxic chemotherapy. Cell Stem Cell. 2018;23(5):700-713.e6. DOI
  11. Boettcher S, et al. A dominant-negative effect drives selection of TP53 missense mutations in myeloid malignancies. Science. 2019;365(6453):599-604. DOI
  12. Jaiswal S, et al. Clonal hematopoiesis and risk of atherosclerotic cardiovascular disease. N Engl J Med. 2017;377(2):111-121. DOI
  13. Weeks LD, et al. Prediction of risk for myeloid malignancy in clonal hematopoiesis. NEJM Evid. 2023;2(5):EVIDoa2200310. DOI
  14. Desai P, et al. Somatic mutations precede acute myeloid leukemia years before diagnosis. Nat Med. 2018;24(7):1015-1023. DOI
  15. Spencer Chapman M, et al. Clonal selection of hematopoietic stem cells after gene therapy for sickle cell disease. Nat Med. 2023;29(12):3175-3183. DOI