Hematopoietic stem cells (HSCs) continuously produce blood cells while maintaining their self-renewal, proliferation, and differentiation potential. Normal blood cell production is balanced between myeloid and lymphoid progeny. With aging, the number of HSCs increases but their differentiation potential declines.¹ One of the hallmarks of aged HSCs is a myeloid differentiation bias, with less capability of differentiation toward the lymphoid lineage, resulting in age-related myeloid skewing. Another common feature of the aging hematopoietic system is the increased prevalence of somatic driver mutations within the HSC compartment. Clonal outgrowth of a subpopulation of cells sharing a mutation in a hematological malignancy-associated driver gene is called clonal hematopoiesis (CH).² Since the prevalence of both conditions increase with age, we questioned whether there is an association between myeloid skewing and CH.

To gain insight into the changes in myeloid and lymphoid progeny upon aging, we analyzed all individuals from the Dutch population-based Lifelines cohort ≥18 years with available myeloid and lymphoid peripheral blood counts (n = 144 676). In males, the percentage of myeloid cells from the total leukocytes increased significantly with aging (p < .001; Figures S1 and S2), while in females, the myeloid cells showed a periodic pattern with an initial increase, followed by a decrease during menopause and finally increased again from the age of 70 (Figures S1 and S2). A clear difference was observed between males and females for the changes in myeloid cell counts. This may be explained by changes in sex hormone levels, as the number of neutrophils decreases significantly during menopause in females. However, we observed a clear shift in the mean percentage of myeloid cells upon aging (Figure S1).

To investigate whether there is an association between the myeloid cell percentage and CH, we evaluated all individuals ≥60 years from the Lifelines cohort (n = 21 727) with available myeloid and lymphoid blood cell counts from whom we had generated CH data previously (n = 4607; Figures S3 and S4, Supplemental Methods; Data S1, Table S1). The percentage of myeloid cells was normally distributed in this cohort with a mean of 67.8% myeloid cells (Figure S4). From these individuals, n = 1899 (41.2%) carried at least one driver gene mutation with a variant allele frequency (VAF) ≥1%. A significant association was observed between the percentage of myeloid cells and mutations in JAK2 (OR 1.06, 95% CI 1.03–1.09; p < .001), SF3B1 (OR 1.03, 95% CI 1.00–1.07; p = .034), and TET2 (OR 1.01, 95% CI 1.00–1.02; p = .020; Figure S4). Overall, no significant correlation was observed between the percentage of myeloid cells and the clone size in the cohort with available myeloid cell counts and CH (n = 1899; p = .891; Figures 1A and S5). However, we observed a positive correlation between the percentage of myeloid cells and the clone size of JAK2 (Spearman's rank correlation coefficient 0.319; p = .012; Figure 1B) and ASXL1 (Spearman's rank correlation coefficient 0.279; p = .002; Figure 1C). In line with this, it has been shown that homozygous JAK2-V617F mutations associate with increased white blood cell counts compared with heterozygous mutations.³ Our data suggest that this dosage effect may already be present in a premalignant heterozygous state.

Subsequently, we investigated the association between CH and aberrant myeloid cell counts upon aging. In the absence of a generally accepted, fixed threshold to describe myeloid skewing upon aging, we selected individuals ≥60 years with the highest (99th) percentile (n = 218, 138 males, 80 females, myeloid percentage ≥ 83.82%) and lowest (1st) percentile (n = 218, 47 males, 171 females, myeloid percentage ≤ 46.08%) of myeloid cells, as well as 1:1 age- and sex-matched controls (Tables S2 and S3; Figures S6 and S7). The prevalence of CH was not significantly different in high myeloid cases compared with their matched controls (42.8% vs. 38.9%; p = .431; Figure S7), and low myeloid cases compared with their matched controls (33.3% vs. 36.4%; p = .543). Despite their low frequency, we observed significantly more mutations in spliceosome-associated genes (SF3B1 and SRSF2) compared with their matched controls (4.2% vs. 0.5%; p = .020; Figures 1D,G, S8, and S9). All spliceosome mutations were detected in cases with concurrent anemia (n = 9). We subsequently investigated the clone size in cases with low or high myeloid cell percentages and observed a significantly larger clone size in TET2 mutant low myeloid cases compared with their matched controls (median VAF 13.0% vs. 2.0%; p = .005; Figure 1H), while the clone sizes of high myeloid cases were not significantly different from controls (Figure 1E). We hypothesize that low myeloid cases carrying TET2 mutations represent cases with an underlying lymphoid malignancy, as TET2 mutations are commonly identified in, for example, lymphoma or diffuse large B-cell lymphoma.⁴

To determine whether CH-associated mutations are restricted to cells derived from the myeloid or lymphoid lineage in high or low myeloid cases, we sorted mature myeloid and lymphoid cells from selected cases and established the presence and clone size of the mutations (Supplemental Methods; Data S1). We selected samples that carried mutations in DNMT3A or TET2, as these are the most commonly mutated genes. Overall, the VAF was increased in cells derived from the myeloid lineage (granulocytes, monocytes) compared with cells derived from the lymphoid lineage (T- and B-lymphocytes; Figure S10). The VAF of the mutated NK-cell clones were comparable to the myeloid lineage (monocytes, granulocytes).⁵ Although we observed a significant increase in the TET2 mutant clone size of low myeloid cases compared with their matched controls, no specific enrichment of TET2 mutations in the lymphoid fraction compared with the myeloid fraction was revealed (Tables S4 and S5).

Follow-up data were available from a subset of cases with a high or low myeloid percentage but showed no significant difference in the proportion of CH between cases with a consistently high or low myeloid percentage at follow-up or cases who corrected their myeloid cell percentage (Figures S11 and S12). Furthermore, the changes of myeloid cell percentages over time did not depend on the presence of CH at baseline (Figure S11) nor the clone size (Figure S13). Cases with a high myeloid percentage had a significantly higher number of platelets (p = .023) and the marker for inflammation, high-sensitive C-reactive protein (hsCRP), was significantly increased in cases with a high myeloid cell percentage compared with the matched controls (p < .001, Table S2), while in cases with a low myeloid percentage the platelet counts (p = .035) and hsCRP levels (p = .015) were significantly decreased. The elevated hsCRP level could be a contributing factor to myeloid skewing.

To evaluate the consequences of aberrant myeloid cell counts, we determined whether the presence of CH impacted the survival of cases with high or low myeloid skewing. High myeloid skewing cases have inferior survival compared with their age- and sex-matched controls (p < .001; HR 2.47, 95% CI 1.54–3.97), which is in line with studies showing that a high Neutrophil-to-Lymphocyte Ratio (NLR) is associated with all-cause mortality.⁶ However, the presence of CH did not significantly affect survival of high myeloid skewing cases (p = .427; HR 1.25, 95% CI 0.72–2.19; Figure 1F; Table S6A). Furthermore, low myeloid skewing cases showed a trend toward inferior survival compared with their matched controls (p = .051; HR 2.19, 95% CI 1.00–4.82), but the presence of CH did not have an effect on survival in these cases (p = .345; HR 1.53, 95% CI 0.63–3.72; Figure 1I; Table S6B). When stratifying the analysis for specific mutated genes, cases with low myeloid skewing showed a higher all-cause mortality for individuals carrying TET2 mutations (p = .020; HR 3.42, 95% CI 1.22–9.64; Figure S4; Table S7).

By establishing linkage of the Lifelines cohort to the nationwide population-based National Cancer Registry database, we investigated the incident diagnoses of hematological neoplasms. First, we evaluated the cumulative incidence of hematological malignancies in all individuals ≥60 years from the Lifelines cohort with available myeloid cell counts (n = 21 599) with correction for age and sex. A higher cumulative incidence was observed in individuals with the highest myeloid cell percentage (>80%, n = 739; p < .001; HR 3.13, 95% CI 1.94–5.03) and lowest myeloid cell percentage (≤50%, n = 621; p < .001; HR 3.06, 95% CI 1.76–5.32; Figure S14). Furthermore, a higher cumulative incidence was observed in individuals with a myeloid cell percentage of 51–60% (p = .011; HR 1.55, 95% CI 1.11–2.17; Figure S14). Thereafter, we evaluated the cumulative incidence of hematological malignancies in high or low myeloid cases compared with their matched controls. Due to the low number of incidences, we could not stratify our analysis for the presence of CH. After a follow-up period of 10 years, 6 high myeloid cases were diagnosed with hematological malignancies (2.8% of the cases, n = 3 developed myeloid malignancies and n = 3 developed lymphoid malignancies), which showed a higher cumulative incidence compare with their matched controls, although this did not reach statistical significance (p = .052; Figure 1J). The incidence of hematological malignancies was comparable among low myeloid cases (n = 6 incident diagnoses, 2.9%, n = 2 developed myeloid malignancies and n = 4 developed lymphoid malignancies) and their matched controls (p = .316; Figure 1K). Together, we show that the absolute number as well as the relative abundance of myeloid over lymphoid cells increases significantly upon aging, but that this is not driven by the presence of clonal hematopoiesis (CH).

造血干细胞 （HSC） 在保持其自我更新、增殖和分化潜力的同时不断产生血细胞。正常的血细胞产生在髓系和淋巴后代之间是平衡的。随着年龄的增长，HSC 的数量增加，但其分化潜力下降。¹ 老年 HSC 的标志之一是髓系分化偏倚，向淋巴系分化的能力较弱，导致与年龄相关的髓系偏斜。造血系统老化的另一个共同特征是 HSC 隔室内体细胞驱动突变的患病率增加。共享血液系统恶性肿瘤相关驱动基因突变的细胞亚群的克隆生长称为克隆性造血 （CH）。² 由于这两种情况的患病率都随着年龄的增长而增加，我们质疑髓系偏斜和 CH 之间是否存在关联。

为了深入了解衰老后骨髓和淋巴后代的变化，我们分析了来自荷兰人群的 Lifelines 队列 ≥18 岁具有可用骨髓和淋巴外周血计数的所有个体 （n = 144 676）。在男性中，髓系细胞占总白细胞的百分比随着年龄的增长而显着增加 （p < .001;图 S1 和 S2），而在女性中，髓细胞呈周期性模式，最初增加，然后在更年期减少，最后从 70 岁开始再次增加（图 S1 和 S2）。在骨髓细胞计数的变化方面，观察到男性和女性之间存在明显差异。这可能是由性激素水平的变化来解释的，因为女性更年期中性粒细胞的数量显着减少。然而，我们观察到衰老后骨髓细胞的平均百分比发生了明显变化（图 S1）。

为了调查髓细胞百分比与 CH 之间是否存在关联，我们评估了 Lifelines 队列中 ≥ 60 岁的所有个体 （n = 21 727），具有可用的髓细胞和淋巴细胞计数，我们之前从中生成了 CH 数据 （n = 4607;图 S3 和 S4，补充方法;数据 S1，表 S1）。髓系细胞的百分比在该队列中呈正态分布，髓系细胞的平均分布为 67.8%（图 S4）。在这些个体中，n = 1899 （41.2%） 携带至少一个变异等位基因频率 （VAF） ≥1% 的驱动基因突变。观察到骨髓细胞的百分比与 JAK2 突变之间存在显著关联 （OR 1.06,95% CI 1.03-1.09;p < .001）、SF3B1 （OR 1.03,95% CI 1.00–1.07;p = .034）和 TET2 （OR 1.01,95% CI 1.00–1.02;p = .020;图 S4）。总体而言，在队列中，骨髓细胞的百分比与克隆大小之间没有观察到显著的相关性，可用骨髓细胞计数和 CH （n = 1899;p = .891;图 1A 和 S5）。然而，我们观察到骨髓细胞的百分比与 JAK2 的克隆大小呈正相关 （Spearman 秩相关系数 0.319;p = .012;图 1B）和 ASXL1 （Spearman 秩相关系数 0.279;p = .002;图 1C）。与此一致，已经表明，与杂合突变相比，纯合 JAK2-V617F 突变与白细胞计数增加相关。³ 我们的数据表明，这种剂量效应可能已经存在于癌前杂合状态。

随后，我们研究了 CH 与衰老时异常髓系细胞计数之间的关联。在没有普遍接受的固定阈值来描述衰老时骨髓偏斜的个体 ≥60 岁）髓细胞（n = 218,138 名男性，80 名女性，骨髓百分比≥ 83.82%）和最低（第 1 个）百分位数（n = 218,47 名男性，171 名女性，骨髓百分比≤ 46.08%），以及 1：1 年龄和性别匹配的对照（表 S2 和 S3;图 S6 和 S7）。与匹配的对照组相比，高髓病例的 CH 患病率没有显著差异 （42.8% vs. 38.9%;p = .431;图 S7），以及低髓系病例与匹配的对照相比（33.3% 对 36.4%;p = .543）。尽管它们的频率较低，但我们观察到剪接体相关基因 （SF3B1 和 SRSF2） 的突变与匹配的对照 （4.2% 对 0.5%;p = .020;图 1D、G、S8 和 S9）。在并发贫血的病例中检测到所有剪接体突变 （n = 9）。我们随后研究了骨髓细胞百分比低或高的情况下的克隆大小，并观察到与匹配的对照相比，TET2 突变低髓病例的克隆大小显着更大（中位 VAF 13.0% 对 2.0%;p = .005;图 1H），而高髓病例的克隆大小与对照没有显着差异（图 1E）。我们假设携带 TET2 突变的低髓病例代表具有潜在淋巴恶性肿瘤的病例，因为 TET2 突变通常在淋巴瘤或弥漫性大 B 细胞淋巴瘤中被发现。⁴

为了确定 CH 相关突变是否仅限于高髓或低髓病例中来自髓系或淋巴谱系的细胞，我们从选定病例中对成熟髓系和淋巴细胞进行了分类，并确定了突变的存在和克隆大小（补充方法;数据 S1）。我们选择了携带 DNMT3A 或 TET2 突变的样本，因为这些是最常见的突变基因。总体而言，与来自淋巴系的细胞（T 淋巴细胞和 B 淋巴细胞;图 S10）。突变的 NK 细胞克隆的 VAF 与髓系 （单核细胞、粒细胞） 相当。⁵ 尽管我们观察到与匹配的对照相比，低髓类病例的 TET2 突变克隆大小显着增加，但与髓系组分相比，没有发现淋巴组分中 TET2 突变的特异性富集（表 S4 和 S5）。

随访数据来自骨髓百分比高或低的病例子集，但显示随访时骨髓百分比持续高或低的病例或校正骨髓细胞百分比的病例之间的 CH 比例没有显着差异（图 S11 和 S12）。此外，髓系细胞百分比随时间的变化不取决于基线时 CH 的存在（图 S11）也不取决于克隆大小（图 S13）。骨髓百分比高的病例血小板数量显著增加 （p = .023），与匹配的对照相比，骨髓细胞百分比高的病例中炎症标志物高敏 C 反应蛋白 （hsCRP） 显着增加（p < .001，表 S2），而在骨髓百分比低的情况下，血小板计数 （p = .035） 和 hsCRP 水平 （p= .015） 显著降低。升高的 hsCRP 水平可能是骨髓偏斜的一个促成因素。

为了评估髓系细胞计数异常的后果，我们确定了 CH 的存在是否会影响髓系偏斜高或低的病例的生存率。与年龄和性别匹配的对照组相比，高髓偏斜病例的生存率较差 （p < .001;HR 2.47,95% CI 1.54-3.97），这与研究表明高中性粒细胞与淋巴细胞比值 （NLR） 与全因死亡率相关的研究一致。⁶ 然而，CH 的存在并没有显着影响高髓偏斜病例的生存率 （p = .427;HR 1.25,95% CI 0.72–2.19;图 1F;表 S6A）。此外，与匹配的对照组相比，低髓偏斜病例显示出生存率较差的趋势 （p = .051;HR 2.19,95% CI 1.00-4.82），但 CH 的存在对这些病例的生存率没有影响 （p = .345;HR 1.53,95% CI 0.63-3.72;图 1I;表 S6B）。在对特定突变基因的分析进行分层时，骨髓偏斜低的病例显示携带 TET2 突变的个体的全因死亡率更高 （p = .020;HR 3.42,95% CI 1.22–9.64;图 S4;表 S7）。

通过将 Lifelines 队列与全国基于人群的国家癌症登记数据库联系起来，我们调查了血液肿瘤的发病诊断。首先，我们评估了 Lifelines 队列中所有个体 ≥ 60 岁血液恶性肿瘤的累积发病率，并校正了可用髓细胞计数 （n = 21 599），并校正了年龄和性别。在髓系细胞百分比最高的个体中观察到较高的累积发生率 （>80%，n = 739;p < .001;HR 3.13,95% CI 1.94-5.03）和最低的髓系细胞百分比 （≤50%，n = 621;p < .001;HR 3.06,95% CI 1.76–5.32;图 S14）。此外，在骨髓细胞百分比为 51-60% 的个体中观察到更高的累积发生率 （p = .011;HR 1.55,95% CI 1.11–2.17;图 S14）。此后，我们评估了与匹配对照相比，高髓或低髓病例血液恶性肿瘤的累积发生率。由于发生率低，我们无法对 CH 的存在进行分层分析。经过 10 年的随访期，6 例高位髓系病例被诊断为血液系统恶性肿瘤 （2.8% 的病例，n = 3 发展为髓系恶性肿瘤，n = 3 发展为淋巴恶性肿瘤），与匹配的对照组相比，累积发生率更高，尽管这没有达到统计学意义 （p = .052;图 1J）。血液系统恶性肿瘤的发病率在低髓系病例中相当 （n = 6 次事件诊断，2.9%，n = 2 发生骨髓恶性肿瘤，n = 4 发生淋巴恶性肿瘤）及其匹配的对照 （p = .316;图 1K）。总之，我们表明骨髓相对于淋巴细胞的绝对数量以及相对丰度随着衰老而显着增加，但这并不是由克隆造血 （CH） 的存在驱动的。