小鼠脑出血后一个月内染色质可及性的动态变化
下划线 = 中文容易“变味”或丢信息的词句：红线带编号 = 12 条注释卡（悬停看、点击跳转）。这一篇是“正在发生什么”到“细胞为什么这么做”的桥：它不看 RNA，而是看染色质哪一段被打开了——也就是细胞把自己的哪一部分说明书翻到了台面上。英文原文附于文末
摘要
脑出血（ICH）是一种严重的卒中亚型，致死率与致残率均高。理解 ICH 背后的分子与细胞机制，对于开发有效的治疗干预至关重要。我们采用单核转座酶可及性染色质测序（snATAC-seq），刻画了小鼠模型 ICH 后第一个月内染色质可及性的变化。我们共检测了 262,506 个细胞核，在第 1、3、7、14、28 天鉴定出 77,978 个上调、35,023 个下调的候选顺式调控元件（cCREs）。结果揭示了小胶质细胞与巨噬细胞中多个超级增强子的破坏，以及反应性星形胶质细胞从保护性状态向神经毒性状态的转变。我们还观察到，在 ICH 后的髓鞘再生通路中，参与缺氧应答与脂质代谢的基因出现显著上调。本研究为 ICH 后发生的表观遗传改变提供了关键认识，为未来研究与潜在治疗干预奠定了坚实基础。
引言
脑出血（ICH）是脑实质内血管破裂所致，占所有卒中的约 10%–15%，但与缺血性卒中相比，其死亡率与致残率更高，而有效治疗手段有限。血肿形成后，红细胞裂解、凝血酶激活与炎症细胞浸润共同造成继发性脑损伤；这一过程涉及小胶质细胞激活、外周单核细胞浸润、星形胶质细胞反应性增生以及血脑屏障破坏。然而，既往研究多在 RNA 或蛋白层面描述这些变化，很少触及“细胞为什么会在某个时刻改变自己的表达程序”这一层。
染色质可及性（chromatin accessibility）正是回答这一层问题的入口：DNA 被核小体包裹，只有处于开放状态的调控区域才能被转录因子结合。因此，测量全基因组的开放区域，可以在转录发生之前看到细胞“准备做什么”。本文用 snATAC-seq 在 ICH 后 1、3、7、14、28 天五个时间点，对皮层与纹状体组织做了系统描绘，目的是把 ICH 后的脑内反应从“细胞与分子的清单”推进到“基因调控程序的时序图”。
结果 1 · ICH 的 snATAC-seq 图谱
为考察细胞对 ICH 的响应，我们使用胶原酶诱导模型：向 C57BL/6 小鼠纹状体立体定位注射胶原酶。该方法模拟急性脑血管损伤，出血部位与体积较为一致。我们在 ICH 后第 1、3、7、14、28 天以及正常对照组，收集皮层与纹状体组织，使用 DNBelab C4 平台进行 snATAC-seq。经质量控制后保留 262,506 个细胞核，每个细胞核的片段数中位数为 27,443。
使用 Harmony 进行批次校正后，我们鉴定出 18 种不同的细胞类型，包括星形胶质细胞（AC，23,231 个核）、脉络丛上皮细胞（CHOR，224 个）、间脑与中脑兴奋性神经元（MEGLU，5,517 个）、齿状回颗粒细胞（DGGRC，3,279 个）、内皮细胞（EN，2,131 个）、室管膜细胞（EPEN，546 个）、成纤维细胞（FB，1,058 个）、谷氨酸能神经母细胞（GNBL，670 个）、巨噬细胞（MAC，4,586 个）、小胶质细胞（MGL，10,689 个）、中型棘状神经元（MSN，35,112 个）、非谷氨酸能神经母细胞（NGNBL，1,552 个）、少突胶质细胞（OLG，2,919 个）、少突胶质前体细胞（OPC，8,155 个）、肽能神经元（PEP，26,184 个）、T 细胞（470 个）、端脑投射兴奋性神经元（TEGLU，95,096 个）与端脑抑制性中间神经元（TEINH，22,087 个）。细胞类型注释依据每种细胞最多两到三个标记基因在启动子区与基因体区的染色质可及性谱。进一步做亚聚类，可在这些细胞类型内区分出不同的亚型或状态；例如端脑投射兴奋性神经元被分为 12 个亚群，对应不同皮层区域的分类。
接下来，我们利用高分辨率 snATAC-seq 数据，在 18 种细胞类型中鉴定出与候选 cCREs 对应的开放染色质区域：共 463,505 个可及染色质区域，其中 379,646 个在细胞类型之间表现出差异可及性。这些区域富集了谱系决定转录因子的基序，例如端脑投射兴奋性神经元中的 Tcf21 与 Bhlhe22、星形胶质细胞中的 Nfic、巨噬细胞中的 Sfpi1。
为考察 ICH 后的时间变化，我们量化了各时间点各细胞类型的比例。值得注意的是，出现了新的免疫细胞类别——尤其是 T 细胞——同时巨噬细胞与星形胶质细胞的比例显著上升。这些变化提示，脑损伤后细胞组成发生了动态改变。作者也明确写道：这些观察仍需更多生物学重复来证实。
结果 2 · 各细胞类型 cCRE 的变化
我们比较了 ICH 组与正常组在各类细胞中 cCREs 的染色质可及性。在 ICH 后的各时间点（第 1、3、7、14、28 天）共鉴定出 77,978 个上调、35,023 个下调的独特 cCREs。差异性峰的数量与各细胞类型的总峰数相关，因此我们计算了每种细胞中“差异峰／总峰”的比例：星形胶质细胞与巨噬细胞／小胶质细胞的变化比例较高（分别为 28.57% 与 26.75%），而 OPC 与 TEINH 的比例较低（分别为 9.25% 与 12.37%）。
进一步分析发现，多数差异峰为各细胞类型所特有，但亲缘较近的细胞类型之间共享 ICH 依赖性的变化。例如，间脑／中脑兴奋性神经元（MEGLU）与肽能神经元（PEP）之间有 681 个共享峰，星形胶质细胞（AC）与室管膜细胞（EPEN）之间有 565 个共享峰。与出现 20 次以上差异峰相关的基因包括：炎症相关基因（如 Zbtb16、Cxcr4、Fth1、Stat1、C8b、Klf4）、应激反应调控基因（如 Fos、Fkbp5）、突触形态与可塑性相关基因（如 Cpeb2、Epha6、Tns3）以及溶质载体家族基因（如 Slc2a1）。这些结果说明，上述基因相关的 cCREs 在 ICH 后于多种细胞类型中发生了失调。
对这些上调的动态 cCREs 做基序与基因集富集分析，可提示参与脑修复与损伤的转录因子与生物学通路。例如在非谷氨酸能神经母细胞（NGNBL）中，SP1 与 SP4 的基序在 ICH 后第 3、7 天富集，Smad1 与 Smad5 的基序在第 28 天富集；这些 cCREs 附近的基因在第 1 天富集于“对肽的细胞应答”，第 3 天富集于“神经发生调控”。在成纤维细胞（FB）中，Sp1／Sp3／Sp4／Sp6 家族基序富集，GO 分析提示其参与“出芽式血管新生”。昼夜节律转录因子 Dbp 的 DNA 识别基序在多种神经元谱系的上调 cCREs 中富集，尤以 MSN 为著。
结果 3 · 小胶质／巨噬细胞超级增强子的得失
ICH 后，驻留的小胶质细胞迅速激活并调控神经炎症过程；而源自外周单核细胞的巨噬细胞则通过受损的血脑屏障浸润损伤脑组织。这两类细胞在 ICH 后的损伤与修复中都扮演关键角色。超级增强子是大型的转录增强子簇，驱动控制细胞身份与疾病的基因表达。我们考察了这些动态变化的 cCREs 中超级增强子的数量与功能。
我们从 SEdb 2.0 数据库收集了小胶质细胞与巨噬细胞的超级增强子元件，再与巨噬／小胶质细胞的差异峰取交集。经严格筛选后，发现 45 个下调、146 个上调的超级增强子。在下调的超级增强子中，有与 Vsir 等维持小胶质细胞稳态至关重要的基因相关的元件。例如，Tgfbr2 的失调与小胶质细胞激活相关；Inpp5d 活性降低则与自噬及炎症小体激活相关。在上调的超级增强子中，有与多能性基因 Nanog 和 Dppa3 相关的元件（尽管 Nanog 的启动子区并未显示开放）。在 ICH 后第一天，我们鉴定出与 Fos 相关的高度上调超级增强子。第 7 天之后，观察到与脂质代谢基因相关的超级增强子显著上调，如 Ch25h、Gnptab、Apoe、Apoc1 与 Nceh1。Apoe 的空间分布显示，其在 ICH 后第 7–28 天于创区显著表达——这些结果表明该时段创区存在旺盛的脂质代谢。
Bcl3 是一种转录共激活因子，ATAC-seq 分析显示其超级增强子从 ICH 后第 1 天至第 28 天持续上调；空间转录组数据同样显示 Bcl3 在损伤区域显著上调。为考察 BCL3 蛋白是否在小胶质细胞／巨噬细胞中表达，我们在正常样本与 ICH 后第 7 天，对 IBA1（小胶质细胞／巨噬细胞标志物）与 BCL3 做了共免疫染色。结果显示：ICH 后 BCL3 特异性地表达于损伤区域，并与 Iba1 共定位于同一细胞；而正常组中未检测到 BCL3 表达。
我们进一步用 Pando 对受这些超级增强子调控的基因做基因表达调控网络分析，以识别关键转录因子。这些转录因子包括抗氧化转录因子 Nfe2l2、缺氧诱导因子 Hif3a、干扰素调节因子 Irf2 等。Fos 调控 Rara、Prkch、Mcl1 与 Tmed10，而 Bcl3 主要由 Relb 与 Klf1 调控；脂质代谢基因 Apoe 主要由 Stat1 与 Ctcfl 控制，Nceh1 主要由 Elf2 调控。
结果 4 · 小胶质细胞与巨噬细胞的差别
ICH 后活化的小胶质细胞与巨噬细胞之间的功能差异，长期是研究焦点。既往研究表明，巨噬细胞主要参与吞噬活动，且相较小胶质细胞具有更强的噬红细胞作用；但二者功能差异背后的调控机制在很大程度上仍未被探索。
小胶质细胞与巨噬细胞可通过 Itgam 与 Ptprc 的差异表达水平来区分：小胶质细胞 Itgam 表达更高，巨噬细胞 Ptprc 表达更高。为进一步探究二者的差异，我们对各自的差异峰做了 GO 富集分析，结果显示：小胶质细胞主要富集于“突触后组织结构”与“轴突发生”相关条目，而巨噬细胞富集于“免疫应答”相关条目。不过，这些差异可能很大程度上反映的是稳态条件下正常小胶质细胞与外周浸润巨噬细胞的差别。然而，ICH 后小胶质细胞被激活并特异性地聚集在损伤部位周围，这就引出一个更有意思的问题：活化的小胶质细胞与巨噬细胞之间的调控差异是什么？
我们用 AUCell 算法评估小胶质细胞的活化状态。AUCell 用于在单细胞测序数据中识别“基因集”（即基因特征）活跃的细胞；本研究所用的基因集由已知的小胶质细胞活化标志物构成。活化的小胶质细胞集中在 UMAP 投影的一端，据此我们鉴定出 1,205 个活化小胶质细胞。聚焦 ICH 后第 7 天，超级增强子分析揭示了两类细胞不同的功能特征：活化的小胶质细胞保留了维持其核心功能与细胞身份所必需的超级增强子，如 Sall1 与 Siglech；此外还上调了与 Cmtm7、Cdap74、Rapgef1、Plcl2、Vps37b、Ccr6 等基因相关的超级增强子。相比之下，巨噬细胞上调了与 Mmp27、Gpnmb、F7、F10、Lgals3、Clec4d、Mylk2、Sash1、Mertk 等基因相关的超级增强子。
我们对差异富集的峰做了 GO 富集分析：活化的小胶质细胞主要富集于“神经元死亡”“突触后细胞骨架组织”与“树突棘”相关通路；而巨噬细胞显著富集于“血管发生”与“吞噬作用”相关通路。两类细胞的峰中都富集了免疫细胞相关通路，但具体功能不同：小胶质细胞富集于“白细胞迁移”与“白细胞间黏附”，巨噬细胞富集于“髓系细胞分化”与“白细胞活化”。基序分析显示，小胶质细胞富集 Stat2、Irf2、Bcl11b、Bde11a 等基序，而巨噬细胞富集 Nfe2、Cux1、Hif 等基序。
总体而言，活化的小胶质细胞更多参与神经元相关的修复功能，而巨噬细胞在 ICH 后参与组织重塑与血管发生；两类细胞都参与免疫应答相关功能。
结果 5 · 反应性星形胶质细胞 cCRE 的动态
我们发现星形胶质细胞的 AC_2 亚群几乎只出现于 ICH 之后，并表现出反应性星形胶质细胞的特征，如表达 Gfap、Serpina3n、Vim 与 Cxcl10。
我们聚焦这一特定细胞亚群，考察其在 ICH 不同阶段的动态变化。在我们既往发现的基础上，我们观察到反应性星形胶质细胞发生了状态转变：从保护性状态（ICH 后 1–3 天）转向神经毒性状态（ICH 后 7–28 天），本研究的结果印证了这一转变。Tgm1 是保护性星形胶质细胞的标志物，其相关 cCREs 在 AC_2 的早期（ICH 后 1–3 天）上调；相反，神经毒性标志物 Serping1 的 cCREs 在 ICH 后 7–14 天上调；而被视为经典神经毒性标志物的 C3，其 cCREs 在 ICH 后第 14 天上调。这些基因的 cCRE 活性变化模式，与我们此前空间转录组研究中发现的趋势一致，为 ICH 后反应性星形胶质细胞从保护性向神经毒性表型动态转变提供了进一步证据。
为进一步阐明 AC_2 亚群在不同时间点的差异，我们计算了差异峰并对相关基因做 GO 富集分析。第 1 天富集到“钙调蛋白依赖性蛋白激酶活性”；第 3 天富集“黏着连接”“胶质发生”与“胶质细胞发育”，与 ICH 后观察到的胶质瘢痕形成相符；第 7 天富集“GTP 酶调节活性”“生长锥”与“轴突发生”；第 14 天“Wnt 介导的细胞间信号”与“轴突再生”突出；第 28 天则富集“轴突发生”“生长锥”与“极化生长位点”。GO 富集结果提示，星形胶质细胞可能参与 ICH 后的轴突形成与修复。
借助 chromVAR 预测转录因子活性的富集情况，我们识别出具有时间特异性的基序：与抗氧化应答相关的 Nfe2l2 基序在早期富集，同时还有与即刻早期基因（如 Fos／Jun）相关的基序；相比之下，后期以 Stat2、Irf1、Irf7 对应基序的富集为特征，提示随着 ICH 后星形胶质细胞反应的演进，其调控格局发生了转变。
结果 6 · 髓鞘再生的 cCRE
ICH 后，神经元先经历脱髓鞘，随后发生髓鞘再生，后者主要由源自 OPC 的新生成少突胶质细胞完成。我们鉴定出五种不同的细胞亚型：OPC（Cspg4⁺）、定向少突胶质祖细胞（COP，Bmp4⁺）、新形成少突胶质细胞（NFOL，Prom1⁺）与成熟少突胶质细胞（MOLG1／MOLG2，Mag⁺）。这些亚型指示了连续的分化阶段，且处于中间状态（COP 与 NFOL）的细胞数少于起始与终末状态（OPC 与 MOLG）的细胞数。
我们进一步在正常与 ICH 条件下，构建了从 OPC 到 MOLG 的半监督拟时序（pseudotime）轨迹，展示了 OPC → COP → NFOL → OLG 的分化路径。分析揭示了 ICH 后不同时间点细胞比例的变化：值得注意的是，COP 的比例在 ICH 后第 3 天达峰，NFOL 在第 7 天达峰——这提示第 7 天可能是髓鞘修复的高峰期。将沿 ICH 轨迹变异最大的 10% 峰可视化，可见可及染色质呈连续、渐进的调制；同样，沿轨迹变异最大的 10% 基序可视化显示出关键转录因子的变化。例如 Ascl1 参与 OPC 增殖，Sox9 主要在少突胶质细胞定向中发挥作用，而 Sox10 是诱导终末分化与髓鞘化所独有的必需因子。此外，Klf9 可提升少突胶质细胞中的髓鞘基因表达，与 Sox10、Myrf 协同作用。
我们用基因评分矩阵比较 ICH 组与正常组中 OPC 与 OLG 的差异表达。ICH 后上调的包括若干免疫应答相关基因（如 Tat、Cd101、Cebpd）以及少突胶质细胞损伤相关基因（如 C4b）。为进一步说明两条轨迹的差异，我们将 ICH 轨迹中沿拟时序的上调差异峰与差异基因做了可视化，观察到若干与髓鞘再生及细胞损伤应答各方面相关的基因峰值变化。在轨迹早期细胞中，Mertk 上调，提示 OPC 可能最初先吞噬受损的髓鞘；随后，负责转运甲状腺激素的蛋白 Ttr 上调；缺氧相关因子（如 Egln3 与 Hif3a）也沿分化轨迹上调——这些发现提示，ICH 后髓鞘再生的环境可能是缺氧的。此外，在 ICH 后髓鞘再生的背景下，观察到脂质代谢相关基因（如 Pla2g3、Abca2、Plin4、Apod、Napepld）上调。
讨论
我们对小鼠 ICH 后染色质可及性的动态做了全面分析。使用 snATAC-seq，我们在多个时间点描绘了皮层与纹状体组织的表观遗传景观，揭示了各细胞类型特异的 cCREs 及其动态变化。
我们的发现强调了 Zbtb16 与 Fkbp5 在多种细胞类型中的上调。Zbtb16 是已知调控炎症小体活性的基因，Fkbp5 是与应激反应相关的蛋白，二者共同提示 ICH 引起了广泛的细胞应激与炎症。
对小胶质细胞与巨噬细胞超级增强子的分析揭示了 ICH 后的显著变化：与维持小胶质细胞稳态相关的基因（如 Tgfbr2 与 Inpp5d）的超级增强子下调。ICH 后，即刻早期基因 Fos 最初在小胶质细胞／巨噬细胞中上调，随后调控包括 Prkch（蛋白激酶 C）与 Mcl1（抗凋亡的 Bcl-2 家族成员）在内的多个基因；既往转录组分析显示，Fos 在缺血性卒中中发挥抗凋亡与抗氧化作用——这提示 Fos 可能是减轻卒中后凋亡与氧化应激的潜在靶点。在 ICH 后第 7 天，更多与脂质代谢基因（如 Apoe、Nceh1、Ch25h）相关的超级增强子上调。我们此前的研究显示，出血部位的巨噬细胞中会积聚脂滴，转化为泡沫细胞；泡沫细胞的形成很可能源于髓鞘脂质清除机制被压垮，并且是卒中后慢性炎症的关键驱动因素。靶向泡沫细胞中的脂质蓄积，可能是改善恢复的一个有前景的策略。此外，控制多能性基因（如 Nanog 与 Dppa3）的超级增强子上调，尽管 Nanog 的启动子区仍处于关闭状态；这可能与小胶质细胞和巨噬细胞的重编程有关，使其能在 ICH 后动态变化的环境中执行多种功能。
我们分析了小胶质细胞与巨噬细胞的差异。ICH 后，活化的小胶质细胞与神经元死亡及显著的组装（assembly）相关；相比之下，巨噬细胞主要参与吞噬作用、髓系细胞分化与血管发生。这些功能与我们此前关于小胶质细胞与巨噬细胞空间分布的发现一致：活化的小胶质细胞位于病灶外周，而巨噬细胞集中于病灶中心。
我们进一步分析了 ICH 后髓鞘再生的分化轨迹。最初，OPC 迁移至损伤部位，随后分化为少突胶质细胞以修复髓鞘。与典型的分化轨迹相比，我们观察到缺氧诱导因子 Hif3a 以及参与脂质代谢的基因上调。由 OPC 编码的缺氧诱导因子介导氧张力，是出生后髓鞘化的关键调控因素；在 ICH 后，缺氧诱导因子可能成为髓鞘化的关键调节者。
总之，本研究提供了 ICH 后脑内反应的详细表观遗传图谱，突出了参与恢复的关键调控元件与通路。这些发现为驱动脑修复的分子机制提供了有价值的见解，并可能为 ICH 的靶向治疗开发提供线索。
研究局限（作者自述）
作者在本节列出了本研究的主要局限，其中最值得记住的两条是：（1）样本量有限，时间点的细胞比例变化仍需更多生物学重复确认；（2）snATAC-seq 只测量“染色质是否开放”，并不直接等于“基因是否表达”——因此所有关于功能的推论，都需要转录组与蛋白层面的独立验证。这两条恰好是读任何组学文章时应当先问的两个问题。
▲ 本部分翻译难点（对照学习用）
附录 · 英文原文
Dynamic changes in chromatin accessibility within one month of mouse intracerebral hemorrhage
Abstract
Intracerebral hemorrhage (ICH) is a severe stroke subtype associated with high mortality and disability. Understanding the molecular and cellular mechanisms behind ICH is vital for developing effective therapeutic interventions. We utilized a single-nucleus assay for transposase-accessible chromatin sequencing (snATAC-seq) to profile changes in chromatin accessibility over the first month following ICH in murine models. We examined 262,506 cells, identifying 77,978 upregulated and 35,023 downregulated candidate cis-regulatory elements (cCREs) across days 1, 3, 7, 14, and 28. Our results revealed disruptions in multiple super-enhancers in microglia and macrophages, along with a transition of reactive astrocytes from a protective to a neurotoxic state. We also observed a significant upregulation of genes involved in hypoxia response and lipid metabolism within the remyelination pathways post-ICH. This study provides crucial insights into the epigenetic changes that occur following ICH, laying a solid foundation for future research and potential therapeutic interventions.
Subject areas: genomics
Introduction
Intracerebral hemorrhage (ICH) represents a devastating form of stroke, inflicting significant disability and mortality. Treatment options for ICH include medical and surgical approaches to reduce hematoma expansion, control intracranial pressure, and minimize brain damage. Medical treatments typically involve blood pressure management, correction of coagulopathy, and neuroprotective measures., Surgical options range from traditional craniotomy to minimally invasive techniques such as endoscopic evacuation and catheter-based thrombolytic therapy., Despite advances in identifying ICH causes and predicting outcomes and long-term vascular risks,,, no specific treatment for ICH currently exists. Limited therapeutic options for ICH underscore the urgency for a deeper understanding of its pathophysiology.
Rapid onset of bleeding causes severe brain damage, and early intervention is crucial to prevent further neuronal loss. Timely and aggressive treatment, such as rapid blood pressure control, correction of coagulopathies, and early surgical evaluation, can mitigate hematoma expansion and reduce mortality and disability., However, secondary injury begins and progresses within hours to days after ICH, leading to tissue damage and remodeling. Secondary injury triggers a complex cascade of cellular and molecular events that restructure the brain.,,,, Central to this process are macrophages and glial cells, including astrocytes, microglia, oligodendrocyte precursor cells (OPCs), and oligodendrocytes. These cells play a dual role, offering both protection and contributing to secondary damage, positioning them as pivotal in the aftermath of ICH.,,,,,,,
Single-cell technologies have begun to unravel the complex heterogeneity of cell states in response to stroke.,, Our previous spatiotemporal transcriptomic atlas elucidated the activation of astrocytes, oligodendrocytes, and microglia within and around the hemorrhagic zone post-ICH. Astrocytes form a dense scar after seven days, segregating the inflamed environment from the brain parenchyma, with a shift from protective to neurotoxic states. Post-ICH remyelination occurs around the lesion site and is linked with the functions of OPCs and oligodendrocytes. Microglia and infiltrating peripheral immune cells play crucial roles in hematoma phagocytosis, tissue remodeling, and modulation of the immune-inflammatory response. However, we still lack a clear understanding of how these diverse states, particularly in astrocytes, microglia/macrophages, and oligodendrocyte lineage cells, affect functional outcomes during the acute and subacute phases of ICH. Little is known about the chromatin accessibility changes that dynamically reshape gene expression in these cells after ICH.
This study utilized snATAC-seq to explore dynamic chromatin accessibility changes within the first month post-ICH in the mouse cortex and striatum. We analyzed a total of 262,506 cells from these tissues in naive mice and on post-ICH days 1, 3, 7, 14, and 28. Our analysis focused on alterations in chromatin accessibility among various glial cell types.
Results
snATAC-seq atlas of ICH
To examine the cellular response to ICH, we used a collagenase-induced model in C57BL/6 mice, administering the enzyme into the striatum via stereotactic injection., This method simulates acute cerebrovascular injury, resulting in consistent hemorrhage in location and size., We collected cortical and striatal tissues at 1, 3, 7, 14, and 28 days post-ICH and from a naive control group, for snATAC-seq using DNBelab C4 platform (A). After quality control, 262,506 nuclei were retained, with a median of 27,443 fragments per nucleus (A–S1D;).
[Figure 1] Single-nucleus ATAC-seq atlas of ICH(A) Study schematic outlining the experimental design, from sample collection and nuclei isolation to snATAC-seq.(B) UMAP representation of all snATAC-seq cells passing quality control, colored by annotated subclusters. The abbreviation of cell types: DGGRC, dentate gyrus granule neurons; MEGLU, diencephalon and mesencephalon excitatory neurons; GNBL, glutamatergic neuroblasts; NGNBL, non-glutamatergic neuroblasts; PEP, peptidergic neurons; TEINH, telencephalon inhibitory interneurons; TEGLU, telencephalon projecting excitatory neurons; MSN, telencephalon projecting inhibitory neurons; MAC, macrophages; NEUT, neutrophils; DC, dendritic cells; AC, astrocytes; MGL, microglia; OPC, oligodendrocyte precursor cells; OLG, oligodendrocytes; EPEN, ependymal cells; CHOR, choroid plexus epithelial cells; EN, endothelial cells; PER, pericytes; FB, fibroblasts; VLM, vascular and leptomeningeal cells.(C) Gene activity scores for marker genes across different cell types.(D) Differential cCREs for each cell type (FDR < 0.1, log2(fold change) >0.5).(E) The composition percentage of cell types at different time points post-ICH.
After batch correction using Harmony, we identified 18 distinct cell types, including astrocytes (AC, 23,231 nuclei), choroid plexus epithelial cells (CHOR, 224 nuclei), diencephalon and mesencephalon excitatory neurons (MEGLU, 5,517 nuclei), dentate gyrus granule cells (DGGRC, 3,279 nuclei), endothelial cells (EN, 2,131 nuclei), ependymal cells (EPEN, 546 nuclei), fibroblasts (FB, 1,058 nuclei), glutamatergic neuroblasts (GNBL, 670 nuclei), macrophages (MAC, 4,586 nuclei), microglia (MGL, 10,689 nuclei), medium spiny neurons (MSN, 35,112 nuclei), non-glutamatergic neuroblasts (NGNBL, 1,552 nuclei), oligodendrocytes (OLG, 2,919 nuclei), oligodendrocyte precursor cells (OPC, 8,155 nuclei), peptidergic neurons (PEP, 26,184 nuclei), T cells (470 nuclei), telencephalon projecting excitatory neurons (TEGLU, 95,096 nuclei), and telencephalon inhibitory interneurons (TEINH, 22,087 nuclei) (E and S1F; B). Cell-type annotation was based on chromatin accessibility profiles at the promoter regions and gene bodies of up to two or three marker genes per identified cell type (C; A;). Further subclustering revealed distinct subtypes or states within these cell types. For example, telencephalon projecting excitatory neurons was divided into 12 subclusters, representing the classification of telencephalon projecting excitatory neurons in different cortical regions (B; B and S2C).
Next, we identified open chromatin regions corresponding to candidate cCREs across the 18 cell types using high-resolution snATAC-seq data. We identified 463,505 accessible chromatin regions, with 379,646 showing differential accessibility across the cell types (D; Wilcoxon rank-sum test, false discovery rate ≤0.1, log2(fold change) ≥ 0.5). These regions were enriched for motifs of lineage-determining transcription factors such as Tcf21 and Bhlhe22 in telencephalon projecting excitatory neurons, Nfic in astrocytes, and Sfpi1 in macrophages (D).
To explore temporal changes post-ICH, we quantified the proportional representation of each cell type at the designated time points. Notably, new immune cell categories emerged, particularly T cells, while the proportion of macrophages and astrocytes increased significantly. These changes suggest a dynamic shift in cellular composition in response to brain injury (E). Further studies with more biological replicates are necessary to confirm these observations.
ICH-associated changes in cCREs across different cell types
We investigated changes in cCREs across various cell types following ICH. We compared the chromatin accessibility profiles of cCREs between the ICH-affected and naive groups (A). We identified 77,978 unique cCREs as upregulated and 35,023 as downregulated across post-ICH time points (days 1, 3, 7, 14, and 28) (B;). The prevalence of differential peaks correlates with the total number of peaks per cell type, prompting us to calculate the ratio of differential to total peaks for each cell type (A). Astrocytes and macrophages/microglia exhibited higher change ratios (28.57% and 26.75%, respectively), whereas OPCs and TEINH displayed lower ratios (9.25% and 12.37%, respectively).
[Figure 2] Cell-type specific changes in chromatin accessibility in different cell types after ICH(A) Schematic diagram of the data analysis strategy used to identify ICH-dependent changes in cCREs, enriched transcription factor motifs, and biological pathways for each cell type.(B) A stacked bar chart shows the number of ICH-related cCREs for each cell type, with macrophages and microglia merged into one cluster. The left side of the bar chart represents the number of downregulated cCREs, while the right side shows the number of upregulated cCREs, colored by different days post-ICH (ICH-related cCREs: (∗p ≤ 0.05 and abs(log2(fold change)) ≥ 1)).(C) Frequency of upregulated cCREs summarized by gene.(D) Genomic tracks for chromatin accessibility around the Fkbp5 and Klf4 loci in astrocytes (AC), endothelial cells (EN), and telencephalon projecting excitatory neurons (TEGLU) at different time points.(E) Motif enrichment (left) and enriched GO biological pathways (right) for upregulated cCREs in non-glutamatergic neuroblasts (NGNBL), fibroblasts (FB), and telencephalon projecting inhibitory neurons (MSN).
Further analysis identified the overlap of differential peaks among cell types (B and S3C). Most differential peaks were specific to individual cell types, but closely related cell types shared ICH-dependent changes. For instance, there was considerable overlap between diencephalon and mesencephalon excitatory neurons (MEGLU) and peptidergic neurons (PEP), with 681 shared peaks, and between astrocytes (AC) and ependymal cells (EPEN), with 565 shared peaks. C highlights genes associated with differential peaks occurring more than 20 times, including genes implicated in inflammation (e.g., Zbtb16, Cxcr4, Fth1, Stat1, C8b, and Klf4), stress response regulation (e.g., Fos, and Fkbp5), synaptic morphology and plasticity (e.g., Cpeb2, Epha6, and Tns3), and the solute carrier family (e.g., Slc2a1). D shows changes in chromatin accessibility for Fkbp5 and Klf4 in three representative cell types. These results indicate that cCREs associated with these genes are dysregulated across multiple cell types following ICH.
Motif and gene set enrichment analysis of these upregulated dynamic cCREs revealed putative transcription factors and biological pathways involved in brain repair and damage (E; and). For instance, in non-glutamatergic neuroblasts (NGNBL), motifs for SP1 and SP4 were enriched after ICH on days 3 and 7, and motifs for Smad1 and Smad5 were enriched on day 28. Genes near these cCREs were enriched for functions such as cellular response to peptide (day 1) and regulation of neurogenesis (day 3). In fibroblasts (FB), motifs for the specific protein family (Sp1, Sp3, Sp4, and Sp6) were enriched and involved in sprouting angiogenesis based on gene ontology (GO) analysis. DNA recognition motifs of the circadian transcription factor–Dbp, were enriched in increased cCREs in many neuronal cell lineages, particularly in MSN.
Gain and loss of super-enhancer in macrophages/microglia after ICH
After ICH, resident microglia rapidly activate and regulate neuroinflammatory processes. Macrophages derived from peripheral monocytes infiltrate the injured brain through the damaged blood-brain barrier. These two cell types play crucial roles in damage and repair following ICH. Super-enhancers are large clusters of transcriptional enhancers that drive the expression of genes controlling cell identity and disease. We investigated the number and functions of super-enhancers within these dynamically changing cCREs.
We collected super-enhancer elements for microglia and macrophages from the SEdb 2.0 database. We then intersected these super-enhancers with the differential peaks of macrophages/microglia. After strict filtering (see), we found 45 downregulated and 146 upregulated super-enhancers (A; A;). Among the downregulated super-enhancers, we observed those related to Vsir and other genes crucial for maintaining microglial homeostasis (B). For example, Tgfbr2 dysregulation is associated with microglial activation,, and decreased Inpp5d activity is linked to autophagy and inflammasome activation. Among the upregulated super-enhancers, we found those associated with pluripotency genes Nanog and Dppa3 (C), though the promoter region of Nanog did not show openness. On the first day post-ICH, we identified highly upregulated super-enhancers associated with Fos (D). After day 7, we observed significant upregulation of super-enhancers related to lipid metabolism genes, such as Ch25h, Gnptab, Apoe, Apoc1, and Nceh1 (A and 3B). C shows the spatial distribution of Apoe, indicating significant expression at the wound site from days 7–28 post-ICH. These results suggest rich lipid metabolism in the wound area during this period.
[Figure 3] Changed super-enhancers and their regulatory networks in microglia/macrophages post-ICH(A) Heatmap of upregulated super-enhancers in microglia/macrophages at different time points post-ICH compared to the naive state. The heatmap values represent log2 fold changes (ICH vs. Naive). Each row represents a differential cCRE, with the nearest gene labeled on the right.(B) Genomic tracks for chromatin accessibility around the Apoe and Nech1 loci.(C) Spatial distribution of Apoe gene, with the black dashed line roughly indicating the ICH-related lesion area. Data sourced from https://db.cngb.org/stomics/stmich/.(D) Genomic tracks for chromatin accessibility around the Bcl3 loci.(E) Immunofluorescence co-staining of BCL3 and IBA1. Scale bars: 1.25 mm for panoramic images; 100 μm for inset (zoomed-in) images.(F) Regulatory network of genes associated with upregulated super-enhancers. The network identifies the transcription factors involved in regulating these genes.(G) Specific transcription factors regulating Apoe, Bcl3, Fos, and Nceh1.
Bcl3, a transcriptional coactivator, exhibited consistent upregulation of its super-enhancer from day 1 to day 28 following ICH, as revealed by ATAC-seq analysis (D). Consistent with this finding, spatial transcriptomic data demonstrated a significant upregulation of Bcl3 in the injury region (E and S4F). To investigate whether BCL3 protein is expressed in microglia/macrophages, we performed co-immunostaining for IBA1, a marker for microglia/macrophages, and BCL3 in naive samples and at day 7 post-ICH. Our results revealed that BCL3 was specifically expressed in the injury regions following ICH and colocalized with Iba1 in the same cells. Notably, no BCL3 expression was detected in the naive group.
We further used Pando for gene expression regulatory network analysis on the genes regulated by these super-enhancers, aiming to identify key transcription factors (F). These transcription factors include the antioxidant transcription factor Nfe2l2, hypoxia-inducible factor Hif3a, interferon regulatory factor Irf2, and others (F). Fos regulated Rara, Prkch, Mcl1, and Tmed10, while Bcl3 was primarily regulated by Relb and Klf1 (G). The lipid metabolism gene Apoe is primarily controlled by transcription factors Stat1 and Ctcfl, while Nceh1 is mainly regulated by Elf2 (G).
Differences between microglia and macrophages
The functional differences between activated microglia and macrophages following ICH have long been a focus of research. Previous studies indicate that macrophages are primarily involved in phagocytic activity and exhibit higher erythrophagocytosis compared to microglia. However, the regulatory mechanisms underlying their distinct functional roles remain largely unexplored.
Microglia and macrophages can be distinguished by the differential expression levels of Itgam and Ptprc, with microglia showing higher Itgam expression and macrophages exhibiting higher Ptprc expression (A). To further explore their differences, we performed GO enrichment analysis on their respective differential peaks. The results revealed that microglia were primarily enriched in GO terms related to postsynapse organization and axonogenesis, whereas macrophages were enriched in GO terms associated with immune responses. Although GO enrichment analysis highlights distinct functional roles between microglia and macrophages, these differences may largely reflect those of normal microglia and infiltrating macrophages under homeostatic conditions (B). However, following ICH, microglia become activated and specifically accumulate around the injury site. This raises intriguing questions about the regulatory differences between activated microglia and macrophages.
[Figure 4] Differential analysis of microglia and macrophages(A) UMAP plots of microglia and macrophages, with displayed markers distinguishing the two cell types.(B) GO enrichment terms associated with differentially accessible cCREs between microglia and macrophages.(C) UMAP plot showing the activation score of microglia.(D) AUC threshold identification for activated microglia.(E) Differential super-enhancers between activated microglia and macrophages.(F) GO enrichment terms associated with differentially accessible cCREs between activated microglia and macrophages.(G) Motif enrichment analysis associated with differentially accessible cCREs between activated microglia and macrophages.
We applied the AUCell algorithm to evaluate the activation status of microglia. AUCell identifies cells with active “gene sets” (i.e., gene signatures) in single-cell sequencing data. In this study, the gene set used comprised known activation markers of microglia. Activated microglia were concentrated at one end of the UMAP projection, allowing us to identify 1,205 activated microglia (C and 4D, see). Focusing on day 7 post-ICH, the super-enhancer analysis revealed distinct functional characteristics between the two cell types. Activated microglia retain super-enhancers critical for maintaining their core functions and cellular identity, such as Sall1 and Siglech. Additionally, they exhibit upregulation of super-enhancers associated with genes, such as Cmtm7, Cdap74, Rapgef1, Plcl2, Vps37b, and Ccr6. In contrast, macrophages exhibited upregulation of super-enhancers linked to genes, such as Mmp27, Gpnmb, F7, F10, Lgals3, Clec4d, Mylk2, and Sash1, Mertk (E). We performed GO enrichment analysis on the differentially enriched peaks. Activated microglia were primarily enriched for pathways related to neuron death, postsynaptic cytoskeleton organization, and dendritic spines. In contrast, macrophages showed significant enrichment in pathways related to vasculogenesis and phagocytosis. Both microglia and macrophages had immune cell-related pathways enriched within their respective peaks, but their specific functions differed: microglia were enriched for leukocyte migration and leukocyte cell-cell adhesion, while macrophages were enriched for myeloid cell differentiation and leukocyte activation (F). Motif analysis revealed that microglia were enriched for motifs, including Stat2, Irf2, Bcl11b, and Bde11a, whereas macrophages showed enrichment for motifs, such as Nfe2, Cux1, and Hif (G).
Overall, activated microglia are more involved in neuron-related repair functions, while macrophages contribute to tissue remodeling and vasculogenesis following ICH. Both cell types play roles in immune response-related functions.
Dynamic changes of cCREs in reactive astrocyte
We found that the AC_2 subcluster of astrocytes was nearly exclusive to the post-ICH period, which exhibited characteristics of reactive astrocytes, such as the expression of Gfap, Serpina3n, Vim, and Cxcl10 (A–5C).
[Figure 5] Dynamic changes of cCREs in reactive astrocyte(A) Composition percentage of astrocyte subclusters at different time points post-ICH.(B) UMAP representation of astrocytes, colored by astrocyte subclusters.(C) UMAP visualization of reactive astrocyte markers.(D) Genomic tracks for chromatin accessibility around the Tgm1, Serping1, and C3 loci in reactive astrocytes.(E) Differential cCREs in reactive astrocytes at various time points, along with enriched GO biological processes.(F) Motif Z scores for Fos, Jun, Nfe2l2, Stat2, Irf2, and Irf7 in reactive astrocytes at different time points. The Z scores indicate the bias-corrected deviation across all cells.
We focused on these specific cell subsets to investigate their dynamic changes across different stages of ICH. Building on our previous findings, we observed a transition in reactive astrocytes, shifting from a protective state (1–3 days post-ICH) to a neurotoxic state (7–28 days post-ICH). Our current study corroborates this transition. Tgm1, a marker of protective astrocytes, had its associated cCREs upregulated in AC_2 during the early stages of 1–3 days post-ICH (D). Conversely, Serping1, a marker for neurotoxicity, showed cCREs upregulation at 7–14 days post-ICH. Furthermore, C3, recognized as a classical neurotoxic marker, had its cCREs upregulation on day 14 post-ICH. The cCREs activity patterns of these genes align with trends previously identified in our spatial transcriptomics research, providing further evidence of a dynamic shift in reactive astrocytes from a protective to a neurotoxic phenotype following ICH.
To further elucidate the differences in the AC_2 subcluster at various time points, we calculated differential peaks and conducted GO enrichment analysis for the associated genes (E). On the first day, the enriched term was “calmodulin-dependent protein kinase activity”. By the third day, the enriched terms included “adherens junction,” “gliogenesis,” and “glial cell development,” correlating with the observed glial scar formation post-ICH. On the seventh day, terms such as “GTPase regulator activity,” “growth cone,” and “axonogenesis” were enriched. By the fourteenth day, “cell-cell signaling by Wnt” and “axon regeneration” were prominent. Finally, on the twenty-eighth day, “axonogenesis,” “growth cone,” and “site of polarized growth” were enriched. The GO enrichment results suggest that astrocytes may be involved in axon formation and repair following ICH.
Leveraging chromVAR to predict the enrichment of transcription factor activity, we identified motifs with temporal specificity (F). The Nfe2l2 motif, associated with antioxidative responses, was enriched in the early stages, alongside motifs related to immediate-early genes such as Fos/June. In contrast, the later stages were characterized by the enrichment of motifs corresponding to Stat2, Irf1, and Irf7, indicating a shift in the regulatory landscape as the astrocytic response evolves over time post-ICH.
cCREs of remyelination after ICH
After ICH, neurons undergo demyelination followed by remyelination, primarily executed by newly generated oligodendrocytes derived from OPCs. We identified five distinct cell subtypes: OPCs (Cspg4+), committed oligodendrocyte progenitor cells (COPs, Bmp4+), newly formed oligodendrocytes (NFOLs, Prom1+), and mature oligodendrocytes (MOLGs1/2, Mag+) (A and 6B)., These subtypes indicate sequential differentiation stages, with fewer cells in intermediate states (COPs and NFOLs) than in initial and final states (OPCs and MOLGs).
We further constructed semi-supervised pseudotime trajectories from OPCs to MOLGs under both naive and ICH conditions (C). These trajectories illustrated the differentiation path from OPCs to COPs, NFOLs, and OLGs. The analysis revealed changes in cell proportions at different time points post-ICH. Notably, the proportion of COPs peaked on day 3 post-ICH, while NFOLs peaked on day 7 post-ICH (A). This suggests that day 7 may be a peak period for myelin repair. Visualization the most variable 10% of peaks along the ICH trajectory revealed a continuous, gradual modulation of accessible chromatin. Similarly, visualization of the most variable 10% of motifs along this trajectory showed alterations in crucial transcription factors (B). For instance, Ascl1 was involved in OPC proliferation, Sox9 primarily functioned during oligodendrocyte specification, and Sox10 was uniquely required to induce terminal differentiation and myelination. Additionally, Klf9 boosts myelin gene expression in oligodendrocytes, acting as partners of Sox10 and Myrf.
[Figure 6] Differential regulatory elements in the myelination regeneration trajectory(A) UMAP representation of the oligodendrocyte lineage.(B) UMAP visualization of markers for oligodendrocyte lineage differentiation.(C) Display of cell types and pseudotime in the oligodendrocyte lineage differentiation, with the upper half showing the naive state and the lower half showing the ICH state.(D) Volcano plot showing differential genes in OPCs (upper part) and OLGs (lower part) comparing ICH to the naive state (∗p ≤ 0.01 and log2(fold change) ≥ 0.8).(E) Pseudotime heatmap of differential cCREs (left) and differential gene scores (right).(F) Genomic tracks for chromatin accessibility around the Mertk, Hif3a, and Plin4 loci.
We used the gene score matrix to compare the differential expression of OPCs and OLGs between the ICH groups and naive groups. Several genes associated with immune response, such as Tat, Cd101, and Cebpd, as well as genes related to oligodendrocyte injury, including C4b, were upregulated following ICH (D). To further illustrate the differences between the two trajectories, we visualized the upregulated differential peaks and differential genes across pseudotime in the ICH trajectory (E). We observed changes in the peak of several genes relevant to various aspects of remyelination and cellular responses to injury. In early trajectory cells, Mertk was upregulated (E), suggesting that OPCs may initially phagocytose damaged myelin. Subsequently, Ttr, a protein responsible for transporting thyroid hormones, was upregulated. Hypoxia-related factors, such as Egln3 and Hif3a (E), were also upregulated along the differentiation trajectory. These findings suggest that the environment for remyelination after ICH may be hypoxic. Additionally, we observed upregulation of genes related to lipid metabolism, such as Pla2g3, Abca2, Plin4 (E), Apod, and Napepld, in the context of remyelination after ICH. F shows changes in chromatin accessibility of Mertk, Hif3a, and Plin4 along the differentiation trajectory under naive and ICH conditions.
Discussion
We conducted a comprehensive analysis of chromatin accessibility dynamics following ICH in mice. Using snATAC-seq, we profiled the epigenetic landscape of cortical and striatal tissues at multiple time points. This analysis revealed distinct cell type-specific cCREs and their dynamic changes.
Our findings highlighted the upregulation of Zbtb16 and Fkbp5 in multiple cell types. Zbtb16, a gene known to regulate inflammasome activity, and Fkbp5, a protein associated with stress response, indicate that ICH leads to widespread cellular stress and inflammation.
Our analysis of super-enhancers in microglia and macrophages revealed significant changes post-ICH. Super-enhancers associated with genes that maintain microglial homeostasis, such as Tgfbr2 and Inpp5d, were downregulated.,, Following ICH, the Fos gene, an immediate-early gene, was initially upregulated in microglia/macrophages. This upregulation subsequently regulated multiple genes, including Prkch (protein kinase C) and Mcl1 (an anti-apoptotic Bcl-2 family member). Previous transcriptomic analyses demonstrated that Fos plays anti-apoptotic and antioxidant roles in ischemic stroke. These findings suggest that Fos could be a potential target for mitigating apoptosis and oxidative stress following stroke. On day 7 post-ICH, more super-enhancers associated with lipid metabolism genes, such as Apoe, Nceh1, and Ch25h, were upregulated. Our previous study has shown that lipid droplets accumulate in macrophages at the hemorrhage site, transforming them into foam cells. Foam cell formation likely results from an overwhelmed myelin lipid clearance mechanism and is a key driver of chronic inflammation post-stroke. Targeting lipid accumulation in foam cells may be a promising strategy for improving recovery. Additionally, super-enhancers controlling pluripotency genes such as Nanog and Dppa3 were upregulated, although the promoter region of Nanog remained closed. This may relate to microglia and macrophages’ reprogramming, enabling these cells to perform diverse functions in response to the dynamically changing environment following ICH.
We analyzed the differences between microglia and macrophages. Activated microglia were found to be associated with neuronal death and prominent assembly following ICH. In contrast, macrophages were primarily involved in phagocytosis, myeloid cell differentiation, and vasculogenesis. These functions are consistent with our previous findings regarding the spatial distribution of microglia and macrophages, where activated microglia were located at the periphery of the lesion, while macrophages were concentrated at the center of the lesion.
We further analyzed the differentiation trajectory of myelin regeneration following ICH. Initially, OPCs migrated to the injury site and then differentiated into oligodendrocytes for myelin repair. Compared to the typical differentiation trajectory, we observed upregulation of the hypoxia-inducible factor Hif3a and genes involved in lipid metabolism. Hypoxia-inducible factor encoded by OPCs, which mediate oxygen tension, is a crucial regulator of postnatal myelination. Post-ICH, the hypoxia-inducible factor may serve as an essential regulator of myelination.
In conclusion, our study provides a detailed epigenetic atlas of the brain’s response to ICH, highlighting key regulatory elements and pathways involved in recovery. These findings offer valuable insights into the molecular mechanisms driving brain repair and may inform the development of targeted therapies for ICH.
Limitations of the study
A key limitation of this study is the specificity of the BCL3 antibody. Despite validation, non-specific binding cannot be completely excluded. Therefore, any claims regarding BCL3’s role should be interpreted with caution. Future work should confirm these findings with alternative methods or more specific antibodies. Second, with only one mouse per group, the statistical robustness of our findings could be enhanced by larger sample sizes in future work. Despite these constraints, our study establishes a robust starting point for exploring ICH-related epigenetic dynamics, and we anticipate it will inspire further research to build upon these observations.
Resource availability
Lead contact
Requests for further information or reagents should be directed to the corresponding author, Mingyue Wang (wangmingyue@genomics.cn).
Materials availability
All materials used for DNBelab C4 are commercially available.
Data and code availability
Acknowledgments
M.W. is supported by grants from the National Key R&D Program of China (2022YFC3400300). L.L. is supported by the Shenzhen Basic Research Project for Excellent Young Scholars (RCYX20200714114644191) and the National Key R&D Program of China (2022YFC3400405). J.W. is supported by the National Natural Science Foundation of China (82371339). This work was also supported by China National GeneBank (CNGB). We thank the National Supercomputing Center in Zhengzhou for its computational resources.
Author contributions
Jian Wang, M.W., and Junmin Wang conceived the study. X.Z., Z.L., L.W., R.D., and H.Z performed experiments. R.X., K.O., J.T., and Z.C. analyzed and interpreted the data. L.L., J.W., C.L., J.X., Y.H., and M.W. provided feedback. R.X. drafted the manuscript. All authors reviewed, edited, and approved the manuscript.
Declaration of interests
The authors declare no competing interests.
Declaration of generative AI and AI-assisted technologies in the writing process
During the preparation of this manuscript, the authors utilized ChatGPT 4.0 to enhance the grammar. After using this tool, the authors carefully reviewed and edited the text, taking full responsibility for the content of the publication.
STAR★Methods
Key resources table
[Table]
| REAGENT or RESOURCE | SOURCE | IDENTIFIER |
|---|---|---|
| Antibodies | ||
| Anti – IBA1 | Affinity Biosciences | Cat#DF6442; RRID:AB_2838405 |
| Anti – BCL3 | proteintech | Cat#23959-1-AP; RRID:AB_2879375 |
| Chemicals, peptides, and recombinant proteins | ||
| Isoflurane | RWD | R510-22-10 |
| Collagenase VII-S | Sigma-Aldrich | C2399 |
| 4% paraformaldehyde | HAT | HT115 |
| First antibody dilution buffer | Beyotime | P0262 |
| Second antibody dilution buffer | Applygen | C1247 |
| Anti-fade agent containing DAPI | SourhernBiotech | 0100–20 |
| bovine serum albumin | Solarbio | A8020 |
| Triton X-100 | BioFrox | 1139ML500 |
| AMPURE XP beads | Beckman Coulter | Cat# A63881 |
| Tissue-Teck OCT | Sakura | Cat# 4583 |
| Pepsin | Sigma | Cat# P7000 |
| Critical commercial assays | ||
| DNBelab C Series | MGI | DNBelab C4 v2.0 |
| Deposited data | ||
| Raw sequencing reads | This paper | CNP0005762 |
| Software and algorithms | ||
| ArchR | GreenleafLab | https://github.com/GreenleafLab/ArchR |
| ClusterProfiler | YuLab | https://github.com/YuLab-SMU/clusterProfiler |
| Pando | QuaDBio Lab | https://github.com/quadbio/Pando |
| Seurat | satijalab | https://github.com/satijalab/seurat |
| Signac | satijalab | https://github.com/stuart-lab/signac |
| Aucell | aertslab | https://github.com/aertslab/AUCell |
Experimental model and study participant details
Ethical statement
All experimental processes in this study comply with the Animal Ethics Committee of Zhengzhou University and BGI-research (ZZUIRB 2022-148, BGI-IRB E23025). All experiments adhered to national and ARRIVE guidelines (http://www.nc3rs.org.uk/arrive-guidelines).
Experimental mouse models
Only mouse participants were used, and no new human participants were recruited for this study. The animal model and sample acquisition methods used in this study are consistent with those described previously. 3-month-old male C57BL/6 mice (22-25g) were obtained from SPF (Beijing) Biotechnology Co. Ltd. (SCXK(JING)2019-0010). The mice were maintained in a specific pathogen-free (SPF) environment at 23°C–25°C with 45–55% relative humidity. The mice were exposed to a 12-h light/dark cycle and had ad libitum access to food and water. The animal protocol was approved by the Animal Ethics Committee of Zhengzhou University and BGI-research (ZZUIRB 2022-148, BGI-IRB E23025). All experiments adhered to national and ARRIVE guidelines (http://www.nc3rs.org.uk/arrive-guidelines). Mice were randomly assigned to groups using a web-based tool (www.randomization.com).
Method details
ICH mouse model
Male C57BL/6 mice were anesthetized with isoflurane (4% induction, 30% O2 and 70% N2O, 2% maintenance). The mice were secured in a stereotaxic instrument, and a 1–1.5 cm incision was made along the sagittal suture. A glass micropipette, filled with paraffin wax, was connected to a micro-infusion pump. Collagenase VII-S (0.075 U in 0.5 μL sterile saline, Sigma, St. Louis, MO) was injected into the left striatum using a dual-arm brain stereo locator (coordinates: 0.8 mm anterior to bregma, 2.0 mm lateral to midline, 3.5 mm depth) (Stoelting, 51503D, Stoelting Co., Wood Dale, IL, USA). The collagenase was administered at a steady rate of 0.1 μL/min. Throughout the procedure, the mice’s body temperature was maintained at 37.0 ± 0.5°C with a temperature control device (RWD, Shenzhen, China). The control group consisted of untreated male C57BL/6 mice. Mice that did not survive within 24 h post-surgery were excluded from the analysis.
Neurologic deficit score
Mouse neurologic deficit score (NDS) was evaluated on naive, days 0, 1, 3, 7, 14, and 21 after ICH. The NDS system comprised six parameters: body symmetry, gait, climbing, circling behavior, front limb symmetry, and compulsory circling. Each parameter was rated on a scale from 0 to 4, with a maximum possible score of 24. Mice with a score of less than 4 on day 1 post-ICH were excluded from the study.
[Table] Neurological Function Score
| Item | 0 points | 1 point | 2 points | 3 points | 4 points |
|---|---|---|---|---|---|
| Gait | Normal | Stiff | Limping | Tremors, unsteady gait | No movement |
| Body symmetry | Normal | Mild asymmetry | Moderate asymmetry | Severe asymmetry | Extreme asymmetry |
| Climbing | Normal | Climbs normally, forelimb weakness | Grasps inclined board, no movement | Slides down incline, unable to hold body | Immediate fall |
| Forelimb symmetry | Normal | Mild asymmetry | Marked asymmetry | Severe asymmetry | No forelimb/body movement |
| Circling | Normal | Tendency to circle to one side | Intermittent circling to one side | Continuous circling to one side | Axial rotation, swaying, or no movement |
| Forced circling | Normal | Tendency to circle to one side | Circling to one side | Slow axial rotation to one side | No movement |
Tissue collection
Mice were selected based on their NDS. We collected naive mouse and mice at five time points post-ICH, including days 1, 3, 7, 14, and 28, resulting in a total of six groups. One mouse per group that met the criteria was chosen for anesthesia. Following anesthesia, striatal and cortical tissues were extracted and stored at −80°C for subsequent analyses.
Nuclei isolation for snATAC-seq
Nuclei were isolated from brain tissues for snATAC-seq using a slightly modified established protocol. Briefly, frozen tissue was transferred to a 2 mL KIMBLE Dounce tissue grinder (Sigma #D8938-1SET) with 2 mL of ice-cold homogenization buffer [20 mM Tris pH 8.0 (Thermo Fisher Scientific)], 500 mM sucrose (Sigma), 50 mM KCl (Thermo Fisher Scientific), 10 mM MgCl2 (Thermo Fisher Scientific), 0.1% NP-40 (Roche), 1× protease inhibitor cocktail (Roche), and 1% nuclease-free BSA, and 0.1 mM DTT). The tissue was homogenized with 15 strokes of the loose Dounce pestle, and the homogenate was filtered through a 70 μM cell strainer (Falcon). The filtered homogenate was then further homogenized with five strokes of the tight pestle to release nuclei, followed by filtration through a 30 μM cell strainer (Sysmex) into a 15 mL centrifuge tube. Nuclei pellets were obtained by centrifugation at 500g for 5 min at 4°C. The nuclei were then washed twice with 1 mL of ice-cold blocking buffer (1× PBS supplemented with 1% BSA) followed by another centrifugation step at 500 g for 5 min at 4°C. Finally, the nuclei were resuspended in 50 μL of 1× PBS containing 1% BSA and counted using DAPI staining.
snATAC-seq library construction and sequencing
snATAC-seq libraries were prepared using the DNBelab C Series Single-Cell ATAC Library Prep Set (MGI, 1000021878). The preparation process included steps for transposition, droplet encapsulation, pre-amplification, emulsion breakage, capture bead collection, DNA amplification, and purification. Library quality was assessed using an Agilent Bioanalyzer 2100 (Agilent Technologies, G2939A) and Qubit ssDNA Assay Kits (Thermo Fisher Scientific, Q10212). Paired-end sequencing was performed on an MGI DNBSEQ-T7 sequencer, with read lengths of 109 bp for read 1, 50 bp for read 2, and 10 bp for the sample index.
Immunofluorescence
Brain tissue was embedded using a brain tissue embedding medium and sectioned at a thickness of 25 μm using a cryostat. After rinsing with PBS, the sections were mounted onto slides, dried, and stored at −20°C. The brain sections were blocked at room temperature for 2 h with PBS buffer containing 10% fetal bovine serum and 0.1% Triton X-100. Primary antibodies were added and incubated overnight at 4°C. The sections were washed three times with PBS, each for 10 min. Subsequently, the appropriate secondary antibodies were added and incubated at room temperature in the dark for 2 h. After incubation, the sections were washed three times with PBS, stained with DAPI for 1 min, and mounted. The primary antibodies used were IBA1 (Cat# DF6442, Affinity Biosciences) and BCL3 (Cat# 23959-1-AP, Proteintech). The immunostained sections were imaged using a Leica Microsystems instrument (address: Ernst-Leitz-Str. 17–37, Wetzlar, Germany).
snATAC-seq data processing
The raw reads obtained from snATAC-seq were processed using PISA, and aligned to the mouse genome (mm10) using BWA to generate BAM files. Subsequently, bap2 was employed to create fragment files for each snATAC-seq library, ensuring accurate representation of the sequencing data for further analysis.
Quality control, dimensionality reduction and clustering of snATAC-seq data
Most of our data analysis was conducted using the R package ArchR (version 1.0.2). Unless otherwise specified, all functions used are from the ArchR package with default parameters. We initiated the creation of ArchR projects using the ArchRProject function. A Tn5 insertion correction was applied using a positive strand offset of +4 bp and a negative strand offset of −5 bp. To ensure high data quality, low-quality nuclei were filtered based on unique nuclear fragments (nFrags) and transcriptional start site (TSS) enrichment scores, with the following criteria: nFrags >10,000 and TSSEnrichment ≥4. Doublet scores were calculated using the addDoubletScores function with default parameters, and doublets were removed using the filterDoublets function (filterRatio = 1.5, cutScore = 5) to retain high-quality nuclei for subsequent analysis.
Dimensionality reduction was conducted using iterative latent semantic indexing (LSI) to capture key data features. Batch effects between samples were corrected using the addHarmony function. For unsupervised clustering, the addClusters function was applied with a resolution of 3 based on the Seurat Leiden clustering algorithm. Visualization was performed using UMAP with default parameters. Gene activity scores were calculated using the addGeneScoreMatrix function with default parameters, and marker genes for each cluster were identified using the getMarkerFeatures function, with criteria of FDR ≤0.01 and Log2 FC ≥ 1.25. Cell type annotation was performed using known cell-type-specific markers for each cluster. For each identified major cell type, subclasses were identified by performing reclustering with a resolution parameter set to 1.
Peak calling
We created pseudo-bulk replicates for each cell type using the addGroupCoverages function and performed peak calling using the addReproduciblePeakSet function with default parameters, except for setting the cutOff parameter to 0.00001.
Differential cCREs calculation
Cell-type differential cCREs analysis was conducted using the getMarkerFeatures function with the criteria of FDR <0.1 and Log2 FC > 0.5.
To identify ICH-differential cCREs, we compared cell types at various time points post-ICH and their corresponding naive states. This analysis included days 1, 3, 7, 14, and 28 post-ICH, with macrophages and microglia combined for this purpose. The procedure involved the following steps: First, we extracted the called peaks for each cell type, ensuring that only peaks with a score greater than 30 were selected. We then applied the getMarkerFeatures function using the Wilcoxon test method (testMethod = “wilcoxon”), without binarizing the data (binarize = FALSE), and incorporating a bias correction based on the log-transformed number of fragments (bias = c("log10(nFrags)")). The groups compared were days 1, 3, 7, 14, and 28 post-ICH (useGroups = c(‘Day 1’,‘Day 3’,‘Day 7’,‘Day 14’,‘Day 28’)), with the naive state serving as the background group (bgdGroups = 'Naive'). Differential peaks were identified based on following criteriaes: p-value<= 0.05 and an absolute Log2 FC ≥1.
Motif and gene ontology enrichment analysis
Motif enrichment analysis was performed using the peakAnnoEnrichment function, enabling the identification of enriched motifs within the differential peaks. Enriched gene ontology biological pathways were performed by clusterProfiler (version 4.2.2) for ICH-differential cCREs for each cell type. Pathways with an adjusted p-value (p.adjust) < 0.05 were considered significantly enriched.
Identification of ICH-related super enhancers in macrophages/microglia
To identify ICH-related super-enhancers in macrophages and microglia, we downloaded super-enhancer data from the SEDB2.0 database (https://bio.liclab.net/sedb) by searching for “macrophages” and “microglia”. We then intersected these super-enhancers with the differential peaks identified in macrophages/microglia using the findOverlaps function to obtain the differential super-enhancers. To ensure robustness, we retained only those super-enhancers that contained at least four differential peaks.
Construction of gene expression regulatory network
To construct the gene expression regulatory network for macrophages and microglia, we first converted the ArchR object to a Signac object using the ArchRtoSignac package (https://github.com/swaruplabUCI/ArchRtoSignac). For this conversion, we selected cell types including macrophages and microglia, sampling 3000 cells from each type.
Following the conversion, we followed the tutorial provided by Pando (version 1.0.5) (https://github.com/quadbio/Pando) to construct the gene expression regulatory network. We used the genes associated with the differential super-enhancers identified in the previous steps as candidate genes for constructing the macrophages/microglia gene expression regulatory network. For peak selection, we used peaks called from macrophages and microglia with a peak score greater than 30, keeping other parameters at their default settings.
Identification of activated microglia
To identify activated microglia, we utilized the AUCell algorithm, which evaluates the activity of predefined gene sets in single-cell sequencing data. Specifically, we curated a gene set comprising known activation markers of microglia. AUCell calculates an activation score for each cell based on the expression levels of genes in this set, enabling us to quantify the activation state of individual microglial cells. We determined the dataset of activated microglia by leveraging the bimodal distribution of AUCell scores. Most cells in the dataset exhibited lower "AUC" values, whereas a distinct population showed significantly higher sores. In this study, microglia were classified as activated if their AUCell scores for the activation gene set were greater than 0.31.
Trajectory analysis
To construct the trajectory for oligodendrocyte lineage cells in both naive and ICH states, we used the addTrajectory function from ArchR. The trajectory was built using the following lineage order: OPC, COP, NFOL, MOLG1, and MOLG2. We set the parameters preFilterQuantile = 1 and postFilterQuantile = 1. For visualizing features upregulated in ICH across pseudo-time using heatmaps, we first identified differential genes using the getMarkerFeatures function. Next, we retrieved the trajectory from the ArchRProject using the getTrajectory function with useMatrix = "GeneScoreMatrix". We then intersected the identified genes with differential genes meeting the criteria of p-value ≤ 0.01 and Log2FC > 0.8 to obtain the differential trajectory features. This allowed us to focus on the significant changes in gene expression along the trajectory in response to ICH.
Quantification and statistical analysis
The statistical analysis of this study was performed using the ArchR package analysis of R software. The specific parameters are described in detail in the steps above.
Published: June 2, 2025
Footnotes
Contributor Information
Jian Wang, Email: jianwang2020@outlook.com.
Longqi Liu, Email: liulongqi@genomics.cn.
Mingyue Wang, Email: wangmingyue@genomics.cn.