一个碱基如何把抑郁风险写进伏隔核：rs1950834 降低 LRFN5 表达
下划线 = 中文容易“变味”或丢信息的词句：红线带编号 = 15 条注释卡（悬停看、点击跳转）。如果 02 页是“哪些基因值得追”，这一页就是“追到之后怎么把它钉死”：从人群统计一路做到细胞核里哪两个蛋白结合变松、小鼠伏隔核里哪一撮神经元出了问题。英文原文附于文末
摘要（全译）
全基因组关联研究已把 14q21.1 鉴定为重度抑郁障碍（major depressive disorder，MDD）的一个稳健风险位点。但其背后的机制仍不清楚。本研究要探索的是 rs1950834 对富含亮氨酸重复与纤连蛋白 III 型结构域蛋白 5（leucine-rich repeat and fibronectin type III domain containing 5，LRFN5）表达的调控功能，以及它在 MDD 中的作用。
方法上有四块：用 CRISPR/Cas9 介导的基因组敲除与单碱基编辑，判定 rs1950834 对转录因子结合以及靶基因 LRFN5 表达的影响；对多套转录组数据集做 meta 分析，以厘清 MDD 患者中 LRFN5 下调究竟发生在哪个脑区；用腺相关病毒（AAV）介导 Lrfn5 在伏隔核（nucleus accumbens，NAc）的过表达或敲低，检验它们对雄性小鼠抑郁样行为、以及对慢性不可预见温和应激（CUMS）敏感性的影响；用突触蛋白表达检测、高尔基染色和电生理分析监测突触结构与功能。
结果有三层。第一，rs1950834 的风险等位基因（A）降低了对 RNA 聚合酶 II 亚基 A（POLR2A）和转录因子 RAD21 的结合亲和力，从而导致 LRFN5 表达下降。第二，与健康对照相比，MDD 患者的 LRFN5 表达在 NAc 中特异性地下调。第三，在 NAc 神经元中敲低 Lrfn5 会诱导抑郁样行为，并通过突触损伤进一步加重 CUMS 诱导的表型；而在小鼠 NAc 中过表达 Lrfn5 则产生对 CUMS 的抗性。
作者认为，这些发现说明 14q21.1 上的功能性风险单核苷酸多态性 rs1950834 调控 NAc 中 LRFN5 的表达与功能，为 MDD 的分子诊断和靶向干预提供了新的视角。
背景与引言（原文作 Background；精译）
重度抑郁障碍（MDD）是一种致衰性的精神疾病，遗传度为 37%。大量 GWAS 已鉴定出多个与 MDD 相关的变异。遗憾的是，MDD 的致病基因仍然不清楚，导致临床实践中既缺诊断标志物，也缺治疗靶点。MDD 风险位点中处于连锁不平衡（LD）的遗传变异往往显示出相近的 P 值，这阻碍了风险变异的鉴定；而约 90% 的风险位点位于非编码区。此外，风险变异的功能可能具有脑区与细胞类型特异性，使背后的机制更难推断。因此，从 MDD 风险位点中鉴定出风险变异、再做功能验证，是一项迫切的需求。
14q21.1 上的 rs1950834 变异位于 LRFN5 的启动子区，与抑郁风险显著相关。此外，这个变异还与认知功能或脑结构显示出提示性关联。作者此前的功能基因组学研究初步提示：在 34 个扰乱 15 个转录因子结合的 MDD 风险 SNP 中，rs1950834 可能是一个有前景的功能性风险变异。rs1950834 有两个等位基因；此前研究显示，G 等位基因在细胞系中带来的荧光素酶活性高于 A 等位基因。进一步预测认为：rs1950834-A 是加重抑郁的风险等位基因，而 rs1950834-G 起保护作用。按 1000 Genomes 计划第 3 阶段的数据，A 等位基因的频率在世界人群中从 48% 到 80% 不等。
大量研究一致表明 LRFN5 可能参与 MDD 的发病。LRFN5 又名突触黏附样分子 5（synaptic adhesion-like molecule 5，SALM5），是一种 I 型跨膜蛋白，主要在神经系统中表达，影响突触的发育与功能。但 LRFN5 在抑郁中的调控与功能仍不清楚。判定 LRFN5 是否选择性地调节情绪相关脑区中特定神经元亚型的功能，是 MDD 靶向神经调控的一条策略性路径。
结果一：先把 rs1950834 钉在“开着的染色质”里（精译）
基于此前的工作，作者先用生物信息学分析预测 rs1950834 是否是一个功能性风险 SNP。第一步定位：rs1950834 位于 LRFN5 的上游启动子区。第二步看染色质状态：在人脑组织中，这个位置明显富集与活跃转录相关的信号，包括 H3K4me3、H3K4me1 和 H3K27ac。第三步看染色质开放性：人脑组织与神经源细胞的 DNase-seq 一致显示，rs1950834 位于开放染色质区域——这支持它落在一段有促转录活性的 DNA 里。第四步看结合：人神经源细胞的 ChIP-seq 显示，rs1950834 这段序列对转录因子 RAD21 与 POLR2A 的结合最强。第五步看等位基因差异：位置权重矩阵（PWM）分析显示，与 G 等位基因相比，rs1950834 的 A 等位基因对 RAD21 和 POLR2A 的结合力下降。
这一节是纯预测，但它做的每一步都很讲究顺序：先证明“这段 DNA 是活跃的”，再证明“有蛋白结合在这里”，最后才说“换一个碱基，结合变弱”。如果第一步就失败——比如这个位点埋在异染色质里——后面所有推论都不用做了。你以后做功能变异，第一件事就是查 ENCODE 和 Roadmap，看它是不是落在开放染色质和活性组蛋白修饰区。
顺带说一句这个位点的“先天条件”：A 等位基因在世界人群里的频率是 48%–80%，也就是说它在多数人群里是常见等位基因，而不是罕见突变。这决定了它的效应量不可能很大——常见变异+中等效应，正是 MDD 这类复杂疾病的典型遗传结构。
结果二：报告基因、EMSA、ChIP——把“A 结合力低”做实（精译）
报告基因。为检验 rs1950834 不同等位基因的启动子活性，作者做了双荧光素酶报告基因实验：把含 rs1950834 的 513 bp 片段插入报告载体中萤火虫荧光素酶的启动子区，用海肾荧光素酶做内参归一化。结果：在神经母细胞瘤细胞 SH-SY5Y 和胶质母细胞瘤细胞 U251 中，MDD 风险等位基因 rs1950834-A 的启动子活性都显著低于 G 等位基因（每组 N = 16，双尾非配对 t 检验，P < 0.001）。
EMSA。为看这个位点是否真的被核蛋白结合，作者做了电泳迁移率变动分析（EMSA）：用 SH-SY5Y 的核蛋白与生物素标记的探针孵育。结果显示，G 等位基因对 6.4 μg 与 12.8 μg 核蛋白的结合都强于 A 等位基因（每组 N = 2）。竞争实验进一步显示，G 等位基因对重组 RAD21 蛋白和 Ile1231–Lys1350 片段的 POLR2A 蛋白的结合也强于 A 等位基因。
ChIP。为在细胞内验证，作者在 SK-N-AS 细胞里做了 ChIP-qPCR：以 IgG 为阴性对照、H3 为阳性对照，用 RAD21 和 POLR2A 抗体做免疫沉淀，结果显示这两个蛋白在含 rs1950834 的基因组序列上显著富集（每组 N = 4）。再把 ChIP 下来的 DNA 做 Sanger 测序，看富集到的到底是哪个等位基因：结果表明 RAD21 与 POLR2A 都偏好结合 G 等位基因。在全基因组层面，ChIP-seq 也显示 A 与 G 等位基因对 RAD21、POLR2A 的结合能力不同。
把上游调控者拿掉会怎样。作者转而去看公共数据：GSE172425 里，用 siRNA 在小鼠原代皮层神经元中敲低 Rad21（11 例对照、12 例 siRad21），Lrfn5 表达下降；作者自己用 shRNA 敲低 RAD21A/B，SH-SY5Y 与 U251 细胞的 LRFN5 mRNA 与蛋白都下降。GSE125413 里，携带 Polr2a R749H 突变的小鼠成熟神经元（3 例对照、3 例突变），Lrfn5 表达被抑制；作者用 shPOLR2A 敲低，同样降低 LRFN5。
这一节的设计值得画下来：体外（报告基因、EMSA、ChIP）证明“A 等位基因结合弱、活性低”，体内/内源（siRNA、shRNA、突变小鼠）证明“把这两个蛋白拿掉，基因表达就降”。两条腿合起来，才能说“rs1950834 通过 RAD21 与 POLR2A 调控 LRFN5”。只有报告基因那张图，说明不了内源基因；只有敲低实验，说明不了这个 SNP。
结果三：eQTL、CRISPR 与单碱基编辑——把“LRFN5 降低”做实（精译）
eQTL 旁证。作者查了 rs1950834 上下游 1 Mb 范围内的 cis-eQTL（BrainMeta v2 与 PsychENCODE 两套人脑数据）：在这个窗口里，只有 LRFN5 的表达与 rs1950834 显著相关，并且呈基因剂量效应——A 等位基因在多个脑区都与更低的 LRFN5 表达相关。换句话说，GWAS 里那个“抑郁风险”，在表达层面确实对应着“LRFN5 更低”，而且这个对应关系在人脑里是可以在线查到的。
CRISPR 敲除整段序列。作者用 CRISPR-Cas9 把含 rs1950834 的片段整段敲掉（两条 sgRNA 装入 PX-459 载体），SH-SY5Y 细胞的 LRFN5 mRNA 与蛋白都下降。这证明这段序列是 LRFN5 表达所需要的。
单碱基编辑只换一个字母。整段敲除还不算精确——被敲掉的那段里可能有别的调控元件。于是作者用单碱基编辑在 SK-N-AS 细胞里造出 rs1950834-AA 与 rs1950834-GG 两种纯合细胞，直接比较同一个基因座上只差一个碱基的两种细胞：AA 细胞的 LRFN5 mRNA 与蛋白，在未处理（naive）状态、以及在 1 μM 地塞米松（DEX）应激 72 小时之后，都低于 GG 细胞。这是全篇最硬的一步：不是“敲掉一段”，而是“改一个字母”，其他一切不变。
下游转录组。对这两种细胞做 RNA-seq（阈值 Log2|FC| > 2、FDR < 0.05）：naive 状态下 AA 相对 GG 上调 2087 个、下调 2480 个基因，共 4567 个差异表达基因；DEX 应激下上调 2800 个、下调 3191 个，共 5991 个；两者交集 3764 个。GO 富集到的生物学过程是神经元突起形态发生、细胞—细胞黏附、轴突发生；细胞组分是突触结构；分子功能是细胞黏附。KEGG 富集到黏着斑、轴突导向、细胞黏附分子。也就是说，一个碱基的差别，最后落在“突触怎么连、轴突怎么长”这一整套程序上。
结果四：23 套数据里，只有伏隔核在变（精译）
先看这个基因在哪。GTEx 显示 LRFN5 在多种组织中都有表达，但在脑内相对更高；Allen Brain Map 显示它在神经元中高、在胶质细胞中低。这符合一个突触黏附分子的画像。
再看它是不是“抑郁专属”。作者查了其它精神疾病的 GWAS（注意缺陷多动障碍、神经性厌食、焦虑障碍、孤独症谱系障碍、双相障碍、精神分裂症）：在 LRFN5 上下游各 250 kb 的范围内，这些疾病都没有达到全基因组显著（P < 5×10⁻⁸）。这提示 14q21.1–LRFN5 这条关联是 MDD 相对特异的，而不是“所有精神疾病共享的一个杂音”。
然后去找脑区。这是全篇最像“侦探工作”的一段：作者对 GEO 上 23 套人脑转录组数据做了 meta 分析，覆盖外周血、前额叶皮层、前扣带、杏仁核、海马、纹状体、伏隔核等多个区域。结论是：LRFN5 在外周血、额叶皮层、杏仁核、扣带回等区域都没有显著差异，只有伏隔核（NAc）出现了选择性降低。
这一步为什么重要？因为如果只看一套数据，你很可能得到一个“全脑都在降”的假象，或者因为样本量太小什么都看不到。23 套数据一起看，等于把“哪个脑区”这个问题从噪声里捞出来。伏隔核是奖赏环路的核心节点，也是抑郁的经典脑区之一——一个突触黏附分子在奖赏中枢选择性地降低，这个故事才立得住。
结果五：NAc 里敲低与过表达（精译）
先造抑郁模型。作者用慢性不可预见温和应激（CUMS）造抑郁样模型，行为学上看到蔗糖偏好下降（P < 0.01）、悬尾实验与强迫游泳实验的不动时间增加（P < 0.01、P < 0.05，N = 7）。在 NAc 里看 LRFN5：它只在 NeuN⁺ 神经元中下降，而在 GFAP⁺ 星形胶质细胞和 Iba1⁺ 小胶质细胞中没有差异（P > 0.5）；并且神经元内 LRFN5 的水平与蔗糖偏好、强迫游泳的不动时间呈负相关——LRFN5 越低，抑郁样表型越重。
再看是哪一撮神经元。NAc 分三个亚区（core、medial shell、lateral shell），其中约 95% 的神经元是中等棘状神经元（MSN），分为 DRD1⁺ 与 DRD2⁺ 两个亚型。作者发现 CUMS 小鼠的 DRD1⁺ 与 DRD2⁺ MSN 内 LRFN5 普遍下降，没有亚区特异性，也没有亚型特异性——不是某一个小圈子出了问题，而是整个 NAc 的神经元都在降。
然后用病毒直接改。作者用 rAAV（1.0×10¹³ GC/mL，每侧 0.5 μL，双侧 NAc，坐标 AP +1.50、ML ±0.75、DV −4.50，0.25 μL/min）在 NAc 里敲低或过表达 Lrfn5。敲低的结果是：蔗糖偏好下降、强迫游泳不动时间增加，也就是单纯敲低就能造成抑郁样行为；把它叠加在 CUMS 上，表型更重。过表达的结果相反：小鼠对 CUMS 表现出抗性（resilience）。
到这里，因果链才算闭合：SNP → 表达下降 → NAc 神经元功能受损 → 抑郁样行为；反过来，把表达补回去 → 抗应激。注意顺序不要搞反：是先有“病人脑里 NAc 降低”的观察，再有“小鼠里敲低能造出表型”的验证。前者是相关性，后者才是因果。
结果六：突触、树突棘与电生理（精译）
突触蛋白。CUMS 之后，突触相关蛋白 PTPRD、PSD95、SNAP25 都下降；CUMS＋shLrfn5 组降得更低；过表达 Lrfn5 则逆转了这一变化。
树突棘。高尔基染色显示，CUMS 后总树突棘密度下降，其中 stubby 型棘（粗短、没有明显的头颈之分）的密度与比例都下降。作者的解释是：stubby 棘缺少细颈，突触处升高的钙瞬变容易扩散进树突干，因此这类棘的减少会影响突触信号的区隔化。同样的趋势在 CUMS＋shLrfn5 组更重，在过表达组被逆转（各组 N = 22/22/20/19/23）。
电生理。作者记录了微小兴奋性突触后电流（mEPSC）与微小抑制性突触后电流（mIPSC）。CUMS 组 mEPSC 的频率下降而幅度不变，过表达 Lrfn5 可以改善（N = 10/8/8）；mIPSC 的频率与幅度都没有差异（N = 8）。此外，配对脉冲比（PPR）在 CUMS 组升高，过表达 Lrfn5 后回落。
这三个结果要连起来读。mEPSC 频率下降、幅度不变，通常被解释为“突触前释放变少或突触数量变少”，而不是“单个突触的受体变少”；PPR 升高同样指向突触前释放概率下降（第一次刺激把递质囊泡用掉了，第二次就释放得少）。而突触蛋白与树突棘密度的下降，正好给“突触数量变少”提供了形态学证据。三条独立证据指向同一个方向，这比单看一张图可信得多。
讨论与局限（精译）
作者在讨论里补了三块背景。第一，LRFN 家族就是 SALM 家族，成员为 LRFN1–5；LRFN5 的胞外部分含六个富含亮氨酸重复（LRR）——两端各有一个富半胱氨酸的帽区——加一个免疫球蛋白（Ig）域和一个纤连蛋白 III 型域，再加一个单次跨膜域。第二，LRFN5 富集于突触组分，可与 PSD95 在兴奋性突触处形成复合物。第三，在培养的小鼠海马神经元里敲低 LRFN5，会减少兴奋性与抑制性突触的数量，降低 mEPSC 与 mIPSC 的频率和幅度；突触后的 LRFN5 通过反式作用与突触前的 LAR-RPTP 相互作用，促进突触发育并调节 AMPA 受体介导的突触强度。这也解释了为什么本研究会去测 PSD95、PTPRD 这类蛋白。
作者强调的贡献是：把一个 GWAS 位点（14q21.1）里的一个具体碱基（rs1950834）、两个具体蛋白（RAD21、POLR2A）、一个具体基因（LRFN5）、一个具体脑区（NAc）、一类具体细胞（MSN）、一套具体表型（突触与行为）串成了一条线。这种“从统计关联一路做到行为”的工作，正是后 GWAS 时代最缺的东西。
局限。作者自己承认了四条。第一，也是最要命的一条：rs1950834 在人身上位于 14q21.1，而小鼠的 Lrfn5 位于 12qC1——两个位点不同源，所以没法造出位点特异的单碱基编辑小鼠，只能用病毒在 NAc 里改表达量，这损失了“内源位点调控”的精确性。第二，没有 Lrfn5 敲除或条件敲除小鼠，所以没有发育期的证据，也看不出“出生就没有这个基因”会怎样。第三，其它脑区也可能参与——已有研究报道过额叶皮层的变化，而本研究把重点放在 NAc，不能排除其它区域的作用。第四，未来方向是检测 MDD 患者外周血中神经元来源外泌体里的 LRFN5，作为诊断标志物；这条思路的前提是“脑内的变化能漏到血里”，还需要验证。
我在读的时候还注意到两点，作者没有展开，但值得你知道。一是动物只用了雄性（原文写雄性 C57BL/6J，4–6 周龄）——MDD 在女性中更常见，只用雄性是这个领域长期的老问题；二是 4–6 周龄属于刚断奶不久的青春期小鼠，而多数抑郁样行为研究用的是成年鼠。这两点不是致命伤，但它们决定了这个结论能推广到多远。
这篇在轨迹里的位置
这一篇是王俊阳 2025 年三篇里最“靠近机制源头”的一篇，也是把他的两套语言缝在一起的地方。作者贡献声明写得很具体：Z.Z.（张志军）、X.L.（罗雄剑）、Y.K.（孔岩）负责总体设计与指导；D.L.（栾迪）、Q.B.、T.Z. 完成实验；“Y. L. guided graduate students X. C., X. D., and J. W. through the in vitro experiments and bioinformatics analysis, respectively”——也就是说，他在李一帆（Yifan Li）的指导下，做的是体外实验（报告基因、EMSA、ChIP、细胞培养这类）与生物信息学分析（eQTL、meta 分析、RNA-seq 这类）两块。这句话几乎就是他这两年能力结构的一张快照：一手湿实验，一手组学分析。
把它和 02 页那篇 sQTL 并排看，关系立刻清楚：02 页是“用统计方法从 GWAS 里挑出候选基因”（27 个，最后收敛到 5 个），03 页是“把一个候选变异从统计一路验证到小鼠行为”（1 个 SNP，做到突触和电生理）。前者广度大、深度浅；后者广度小、深度大。同一年里既参与“广度”也参与“深度”，这是很少见的组合。
再往后看，2026 年 Nature Communications 上那篇 1p36.23 变异经 RERE 影响精神分裂症风险（16 人里第 2、共同第一作者），正是这两种能力的合流：既要做位点的功能验证，又要处理大样本遗传数据。从 2024 年“在脑出血论文的作者贡献里挂上名字”，到 2025 年“独立承担体外实验与生信分析”，再到 2026 年“共同第一作者”，这条线不是跳变，是一格一格挪出来的。
还有一个小小的呼应值得注意：他的人体解剖学系背景，在这篇里第一次变得“有用”——伏隔核的三亚区（core、medial shell、lateral shell）、MSN 的 DRD1/DRD2 分型、Golgi 染色的棘形态分类，这些都是解剖学与组织学的基本功。做遗传学的人常常卡在“这个基因在哪个核团、哪类细胞”，而这恰好是他不用现学的东西。
给刚进组的你
这一页最值得抄下来的是那张“功能变异七问”清单。以后你读到任何一篇“某个 SNP 导致某个病”的论文，就按这七条对：①这个位点在病人的脑组织里是不是开放染色质、有没有活性组蛋白修饰？②哪个转录因子/蛋白结合在这里？③换个碱基，结合强度变不变（EMSA、PWM）？④换个碱基，报告基因活性变不变？⑤内源基因的表达变不变（eQTL、CRISPR、单碱基编辑）？⑥这个基因在病人脑的哪个区、哪类细胞里变？⑦把基因在小鼠里敲低/过表达，表型会不会出现？七条全过的论文不多，但每多过一条，可信度就高一档。
第二件事：把三种“改基因”的手段分清。CRISPR 敲除整段序列＝把一段 DNA 删掉，可能连带删掉别的调控元件；单碱基编辑＝只换一个字母，是这里最精确的证据；shRNA/siRNA 敲低＝只降 mRNA，基因还在。这篇论文把这三种都用上了，而且用在了不同的问题上——这是它方法部分最值得学的地方。
第三件事：去查一次你自己的 rs1950834 基因型（如果你做过全基因组芯片或全外显子测序）。它是 A 还是 G？A 等位基因在世界人群里占 48%–80%，你大概率是 A/A 或 A/G。这不是让你担心，是让你亲身体会一件事：常见变异 + 小效应是复杂疾病的常态，一个风险等位基因绝不等于“会得病”。
最后三个可以问师姐的问题：①咱们组做 ChIP 用的是哪家的抗体，IgG 和 H3 这两个对照每次都跑吗？②高尔基染色是我们自己染还是送出去？棘的分类（stubby/thin/mushroom）谁来判、盲法怎么做的？③如果要在小鼠里验证一个人源 SNP，除了病毒过表达/敲低，还有别的办法吗？第三问的答案会带你走到“人源化小鼠”和“位点编辑”这些更远的地方去。
▲ 本部分翻译难点（对照学习用）
附录 · 英文原文
The regulatory variant rs1950834 confers the risk of depressive disorder by reducing LRFN5 expression
Abstract
Genome-wide association studies have identified 14q21.1 as a robust risk locus for major depressive disorder (MDD). However, the underlying mechanism remains elusive. Here, we aim to explore the regulatory function of rs1950834 on leucine-rich repeat and fibronectin type III domain containing 5 (LRFN5) expression in MDD.
Clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9)-mediated genome knockout and single-base editing were used to determine the effects of rs1950834 on the binding of transcriptional factors and the expression of the target gene LRFN5. Meta-analysis of multiple transcriptomic datasets was performed to clarify the brain region responsible for LRFN5 downregulation in MDD patients. Adeno-associated virus (AAV)-mediated Lrfn5 overexpression or knockdown in the nucleus accumbens (NAc) was used to test their effects on depression-like behaviors and sensitivity to chronic unpredictable mild stress (CUMS) in male mice. Synaptic structure and functions were monitored by synaptic protein expression assay, Golgi staining, and electrophysiological analysis.
The risk allele (A) of rs1950834 reduced the binding affinity to RNA polymerase II subunit A (POLR2A) and the transcription factor RAD21 cohesin complex component (RAD21), leading to decreased expression of LRFN5. LRFN5 expression was downregulated specifically in the NAc of MDD patients as compared to healthy controls. Knockdown of Lrfn5 in NAc neurons induced depression-like behaviors and further exacerbated CUMS-induced phenotypes via synaptic damage, but overexpression of Lrfn5 in mouse NAc induced resilience to CUMS.
These findings reveal that the functional risk single nucleotide polymorphism rs1950834 at 14q21.1 regulates LRNN5 expression and function in NAc, providing a novel perspective for molecular diagnosis and targeted interventions of MDD.
The risk allele (A) of rs1950834 reduced the binding affinity to transcription factors POLR2A and the RAD21, leading to decreased expression of LRFN5. LRFN5expression was downregulated specifically in the NAc of MDD patients and depression-like mice. Knockdown of Lrfn5 in NAc neurons induced depression-like behaviors and further exacerbated CUMS-induced phenotypes via synaptic damage, but overexpression of Lrfn5 in mouse NAc induces resilience to CUMS. These findings reveal that the functional risk single nucleotide polymorphism rs1950834 at 14q21.1 regulatesLRFN5 expression and function in NAc, providing a novel perspective for molecular diagnosis and targeted interventions of MDD.
The online version contains supplementary material available at 10.1186/s12916-025-04141-8.
Keywords: Major depressive disorder, rs1950834, RAD21, POLR2A, LRNF5, Nucleus accumbens
Background
Major depressive disorder (MDD) represents a debilitating mental illness with a heritability of 37% [[1]]. Numerous genome-wide association studies (GWASs) have identified multiple variants associated with MDD [[2]–[6]]. Unfortunately, the MDD causative genes are still elusive, leading to the lack of diagnostic markers and therapeutic targets in clinical practice. Genetic variants in linkage disequilibrium (LD) of MDD risk loci often show similar P values, impeding the identification of risk variants, and approximately 90% of the risk loci are located in non-coding regions. Furthermore, the function of risk variants may be brain region and cell-type specific, rendering it difficult to infer the underlying mechanisms [[7]]. Identification of risk variants from MDD risk loci followed by functional validation is an imperative requirement.
The rs1950834 variant of 14q21.1, located in the promoter region of leucine-rich repeat and fibronectin type III domain containing 5 (LRFN5), is significantly associated with depression risk [[8], [9]]. In addition, this variant also showed suggestive associations with cognitive function or brain structures (Additional file 1: Table S1–S3). Our previous functional genomics study preliminarily suggested that rs1950834 might be a promising functional risk variant among the 34 MDD risk SNPs disrupting the binding of 15 transcription factors (TFs) [[10]]. There are two alleles of rs1950834, and our previous study showed that the G allele conferred higher luciferase activities than the A allele in cell lines [[10]]. Further, it was predicted that the rs1950834-A was a risk allele exacerbating depression, while the rs1950834-G allele played a protective role [[11]]. In addition, based on the data from the 1000 Genomes Project Phase 3 [[12]], the frequency of the A allele varies from 48 to 80% in world populations (Additional file 1: Fig. S1). Numerous studies have consistently demonstrated that LRFN5 may be involved in MDD pathogenesis [[9], [13]–[15]]. LRFN5, also known as synaptic adhesion-like molecule 5 (SALM5), is a type I transmembrane protein mainly expressed in the nervous system and affects synaptic development and function [[16]–[20]]. However, it is still elusive for the regulation and function of LRFN5 in depression. Determining whether LRFN5 selectively modulates the function of specific neuronal subtypes in emotion-related brain regions represents a strategic approach for targeted neuromodulation in MDD.
Methods
Functional genomics analysis
To determine whether rs1950834 was located in the transcriptionally active region of the chromatin, the H3k4me3, H3k4me1, and H3k27ac chromatin immunoprecipitation-sequencing (ChIP-seq) data at the position of rs1950834 in the human brain were downloaded from Epigenome Roadmap [[21]], DNase I hypersensitive sites sequencing (DNase-seq) and ChIP-seq for transcription factors at the position of rs1950834 in the human brain, human neurogenic cells, and human neurogenic cell lines were queried via the Encyclopedia of DNA Elements (ENCODE) [[22], [23]]. The position weight matrix (PWM) data obtained from ENCODE (ENCFF606TCO and ENCFF246HIM) was analyzed using the R software package “ggseqlogo” [[24]].
Analysis of the correlation between transcription factors RAD21 and POLR2 A, and LRFN5 expression in the GEO datasets
Two datasets, GSE172425 [[25]] and GSE125413 [[26]], were downloaded. GSE172425 (11 controls, 12 siRad21 cases) was used to analyze the effect of Rad21 knockdown on Lrfn5 expression in mouse cortical neurons, while GSE125413 (3 controls, 3 Polr2a[R749H] mutant cases) was employed to investigate the impact of Polr2a mutant in mouse mature neurons on Lrfn5 expression.
Meta-analysis of LRFN5 expression from GEO datasets
Meta-analysis using GEO datasets was used to identify brain regions where LRFN5 was differentially expressed in depressed patients. A total of 23 datasets were downloaded, namely GSE101521 [[27]] (dorsolateral prefrontal cortex: 30 MDD cases, 29 healthy controls), GSE102556 [[28]] (anterior insula: 26 MDD cases, 22 healthy controls; cingulate gyrus: 25 MDD cases, 21 healthy controls; dorsolateral prefrontal cortex: 26 MDD cases, 22 healthy controls; nucleus accumbens: 26 MDD cases, 22 healthy controls; orbitofrontal: 26 MDD cases, 22 healthy controls; subiculum: 24 MDD cases, 19 healthy controls), GSE125664 [[29]] (forebrain neurons derived-iPSCs: 6 MDD cases, 3 healthy controls), GSE126512 [[30]] (serotonergic neurons derived-iPSCs: 6 MDD cases, 3 healthy controls), GSE169459 [[31]] (whole blood: 3 MDD cases, 3 healthy controls), GSE193417 [[32]] (anterior cingulate cortex: 6 MDD cases, 6 healthy controls), GSE32280 [[33]] (peripheral blood leukocyte cells: 8 MDD cases, 8 healthy controls), GSE38206 [[34]] (peripheral blood mononuclear cells: 9 MDD cases, 9 healthy controls), GSE39653 [[35]] (peripheral blood mononuclear cells: 21 MDD cases, 24 healthy controls), GSE44593 [[36]] (amygdala: 14 MDD cases, 14 healthy controls), GSE53987 [[37]] (hippocampus: 17 MDD cases, 18 healthy controls; prefrontal cortex: 17 MDD cases, 19 healthy controls; striatum: 16 MDD cases, 18 healthy controls), GSE54562 [[38]] (anterior cingulate cortex: 10 MDD cases, 10 healthy controls), GSE54563 [[39]] (anterior cingulate cortex: 25 MDD cases, 25 healthy controls), GSE54564 [[40]] (amygdala: 21 MDD cases, 21 healthy controls), GSE54565 [[41]] (anterior cingulate cortex: 16 MDD cases, 16 healthy controls), GSE54566 [[42]] (amygdala: 14 MDD cases, 14 healthy controls), GSE54567 [[43]] (dorsolateral prefrontal cortex: 14 MDD cases, 14 healthy controls), GSE54568 [[44]] (dorsolateral prefrontal cortex: 15 MDD cases, 15 healthy controls), GSE54571 [[45]] (anterior cingulate cortex: 13 MDD cases, 13 healthy controls), GSE54572 [[46]] (anterior cingulate cortex: 12 MDD cases, 12 healthy controls), GSE76826 [[47]] (peripheral blood leukocyte cells: 20 MDD cases, 12 healthy controls), GSE80655 [[48]] (anterior cingulate gyrus: 24 MDD cases, 24 healthy controls; dorsolateral prefrontal cortex: 23 MDD cases, 24 healthy controls; nucleus accumbens: 22 MDD cases, 22 healthy controls), and GSE98793 [[49]] (whole blood: 128 MDD cases, 64 healthy controls). Meta-analysis was performed using Review Manager 5.4. The expression of LRFN5 was treated as continuous data, and the standard mean difference was used to assess the combined effect sizes. Heterogeneity was quantified using the estimated I2 with a Cochrane Q test. The random effects model was applied when the I2 level was ≥ 50% or P ≤ 0.10. Otherwise, the fixed effects model was considered.
Dual luciferase reporter gene assay
To test the impact of rs1950834 on promoter activity, the DNA fragment (513 bp) containing rs1950834 was amplified with specific primers (Additional file 1: Table S4) and inserted into the pGL4-basic vector (Promega, E6661). Vectors containing the alternative allele were generated by PCR-mediated point mutation using Golden Star T6 Super PCR mix (TSINGKE, TSE101) and Dpn I (NEB, R0176S). All constructed vectors were verified by Sanger sequencing. SH-SY5Y (4 × 104/well) and U251 (2 × 104/well) cells were plated into 96-well plates and cultured for 12 h. The reporter vectors containing the cloned fragments and control pRL-TK vectors (Renilla Luciferase) (Promega, E2241) were co-transfected into cultured cells using Lipofectamine 3000 reagent (ThermoFisher, L3000015). Forty-eight hours post-transfection, cells were harvested, and the luciferase activity was determined using Dual-Luciferase Reporter Assay System (Promega, Cat. No. E1960) and Luminoskan Ascent instrument (Thermo Scientific).
Electrophoretic mobility shift assay
Electrophoretic mobility shift assay (EMSA) was used to study the interaction between DNA sequences and proteins (including transcription factors). The nuclear proteins of SH-SY5Y cells were extracted with the nuclear protein extraction kit (Beyotime, P0028). The biotin-labeled probes (Additional file 1: Table S5) and nucleus extracts or recombinant RAD21 and Ile1231-Lys1350 POLR2A protein from the goldORF clone were then incubated to form protein-DNA complexes. EMSA was performed using the chemiluminescence EMSA kit (Beyotime, GS009) according to the manufacturer’s instructions.
ChIP-quantitative polymerase chain reaction (qPCR)
ChIP assay was performed to detect whether RAD21 and POLR2A bind to the genomic sequence containing rs1950834 in SK-N-AS cells using the ChIP kit (CST, 9002). Briefly, cells were cross-linked with 1% formaldehyde solution at room temperature for 10 min. The reaction was quenched with 2 ml 10 × glycine solution for 5 min. The cells were washed twice with cold PBS and collected with 2 ml 1 × PBS (contains 200 × protease inhibitor compound, PIC). Micrococcal nuclease (0.5 µl per 4 × 106 cells) was added and incubated at 37 °C for 20 min. Agarose gel electrophoresis was used to determine the size of DNA fragments. The IgG, H3, RAD21, and POLR2A were set as four IP groups. Crosslinked chromatin fragments (10 µg) and corresponding antibodies were added to each IP reaction, incubating at 4 °C overnight on a rotor. ChIP-level protein G magnetic beads (30 µl) were added to each immunoprecipitation reaction and incubated at 4 °C for 2 h. The enriched products were obtained by separating and rinsing the magnetic beads. The binding ability of RAD21 and POLR2A was determined by real-time qPCR (Additional file 1: Table S6).
CRISPR-Cas9-mediated genome editing
To knock out the genomic sequence containing rs1950834, we designed two sgRNAs (located upstream and downstream of rs1950834, respectively) (Additional file 1: Table S7) and cloned the sgRNAs into the PX-459 vector. RT-qPCR and Western blotting were used to verify the effect of the rs1950834 knockout on LRFN5 expression. Primer sequences were listed in Additional file 1: Table S8.
Single-base editing was performed in SK-N-AS cells using the principle of clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (Cas9) gene editing technology. The gRNA and ssDNA sequences are given in Additional file 1:TableS9. The successfully constructed single-base replaced cells were validated by Sanger DNA sequencing. Control (rs1950834-GG) and mutated (rs1950834-AA) SK-N-AS cells were stressed with 1 µM dexamethasone (MedChemExpress, 50–02-2,) for 72 h [[50]]. RT-qPCR and Western blotting were used to verify the expression of LRFN5.
Depression-like mouse model
Four- to six-week-old male C57BL/6J mice were purchased from GemPharmatech (Nanjing, China). All animals were subjected to a 12-h light–dark cycle (light on from 7:00 to 19:00) with food and water ad libitum at 23–25 °C and 50% humidity. The control groups housed 3–4 mice per cage, while the model groups were housed individually. Chronic unpredictable mild stress (CUMS) was used to induce depressive-like behavior in mice. The mice became acclimated to the environment and were randomly grouped after excluding outliers (baseline sucrose preference score < 70% of average). According to the randomization principle, we used a random number table for the allocation of mice to CUMS and control groups. Mice in the CUMS group were exposed to various stressors, such as swimming at 4 ℃/45 ℃ for 5 min, horizontal cage shaking for 5 min, tilting the cage at 45° for 10 h, restraint for 2 h, food/water deprivation for 12 h, clipping the tail for 1 min, tail suspension for 6 min, wet pad treatment for 12 h, followed by 12 h without padding, rat bedding for 12 h, day and night reversal, and stroboscopic light treatment for 12 h.
The sucrose preference test (SPT), tail suspension test (TST), and forced swimming test (FST) were used to measure depression-like behaviors, while the open-field test (OFT) was used to assess anxiety in mice according to the protocol published in our previous study [[51], [52]]. The ANY-maze animal behavior analysis system (Stoelting Co.) was used for the above detection process. For all experiments, investigators were blinded to group allocation during experimentation and data analysis. The offline data analysis was performed blindly.
Virus stereotactic injection
Mice anesthetized with sevoflurane were injected with recombinant adeno-associated virus (rAAV) (1.0 × 1013 GC/ml, 0.5 µl) in bilateral NAc (AP + 1.50 mm, ML ± 0.75 mm, and DV − 4.50 mm) to knockdown or overexpress neuron-derived Lrfn5. The injection rate was 0.25 µl/min. After administering the injection into the NAc, the skull wound was sealed with bone wax. The neuron-specific overexpression and knockdown efficiency of rAAVs were validated by Western blot and immunofluorescence assay. Animals were excluded after behavior tests if they showed misinjection at postmortem examination.
Golgi-cox staining
The Hito Golgi-Cox OptimStain™ Kit was used for Golgi-cox staining. Mouse brain specimens were soaked in Golgi staining solution at 26 ℃ for 14 days. Then, the tissue treatment solution containing brain specimens was protected from light at 4 ℃ for 3 days, and Golgi developer was added dropwise for 30 min after sectioning.
Electrophysiological analysis
Whole-cell recordings were performed as described previously [[53], [54]]. Mice were anesthetized with isoflurane and decapitated. The brain was rapidly removed and placed in an ice-cold choline-based solution containing the following (in mM): 110 choline chloride, 2.5 KCl, 0.5 CaCl2, 7 MgCl2, 1.3 NaH2PO4, 1.3 Na-ascorbate, 0.6 Na-pyruvate, 25 glucose, and 25 NaHCO3, saturated with 95% O2 and 5% CO2. Two hundred fifty to three hundred micrometers of coronal slices containing the NAc was prepared using a vibratome. Slices were transferred to a holding chamber in an incubator containing oxygenated (95% O2 and 5% CO2) artificial cerebrospinal fluid (ACSF) composed of the following (in mM): 125 NaCl, 2.5 KCl, 2 CaCl2, 1.3 MgCl2, 1.3 NaH2PO4, 1.3 Na-ascorbate, 0.6 Na-pyruvate, 25 glucose, and 25 NaHCO3 at 32 °C for at least 1 h before recording readings. After recovery, slices were transferred to the recording chamber, where they were perfused continuously with ACSF at a flow rate of 2 ml/ min, and all experiments were performed at 23–25 °C. Recordings were obtained using a Multiclamp 700B amplifier and a Digidata 1550B (Molecular Devices). Data were sampled at 10 kHz and analyzed with a pClamp10 (Molecular Devices). Patch pipettes (4–6 MΩ) were filled with a Cs-based low Cl− internal solution containing the following (in mM): 135 CsMeSO3, 10 HEPES, 1 EGTA, 3.3 QX-314, 4 Mg-ATP, 0.3 Na-GTP, 8 Na2-phosphocreatine, 290 mOsm kg−1, adjusted to pH 7.3 with CsOH. The miniature excitatory postsynaptic currents (mEPSCs) and miniature inhibitory postsynaptic currents (mIPSCs) values were recorded at − 70 mV and 0 mV, respectively. The extracellular recording solution consisted of ACSF supplemented with picrotoxin (100 µM) and tetrodotoxin (1 µM) for the mEPSC experiment; and tetrodotoxin (1 µM), AP5 (50 µM), and CNQX (20 µM) for the mIPSC experiment. ACSF was used without any supplements for recording excitatory paired-pulse ratios (PPRs). The PPRs were elicited by electrically stimulating the NAc (0.2-ms current pulses) at − 70 mV. They were calculated by dividing the second electric-evoked excitatory postsynaptic current (EPSC) by the first, with 50 ms intervals between the two stimulations.
Statistical analysis
All values were expressed as mean ± SEM. The Shapiro–Wilk normality test was utilized to assess data distribution. For normally distributed data in this study, the results of the two sets of data were subjected to the two-tailed Student’s t-test, while the remaining multiple sets of data were analyzed using the one-way analysis of variance with Tukey’s post hoc analysis. P < 0.05 was considered to be statistically significant. The sample sizes are similar to those reported in previous publications.
Results
Functional genomics identifies rs1950834, located in 14q21.1, as a potential functional risk variant for depression
Based on our previous findings, we first used bioinformatic analysis to predict if rs1950834 is a functional risk single nucleotide polymorphism (SNP). Firstly, we identified that rs1950834, located in the upstream promoter region of LRFN5 (Fig. A and B), was substantially enriched in signals associated with active transcription in human brain tissues, including H3k4me3, H3k4me1, and H3k27ac (Fig. C). Moreover, DNase-seq of human brain tissue and neurogenic cells consistently demonstrated that rs1950834 resides in an open chromatin region (Fig. D), supporting the fact that rs1950834 was located within a DNA region with transcription-promoting activity. Furthermore, ChIP-seq of human neurogenic cells showed that rs1950834 exhibited the strongest binding affinity to transcription factors RAD21 cohesin complex component (RAD21) and RNA polymerase II subunit A (POLR2A) (Fig. E). Finally, position weight matrix (PWM) analysis revealed that the A allele of rs1950834 exhibited reduced binding affinity to RAD21 (Fig. F) and POLR2A (Fig. G) compared to the G allele. These findings suggest that rs1950834 may function as a regulatory variant by modulating its binding to transcription factor RAD21 and POLR2A.
[Fig. 1] Functional genomics identified rs1950834 as a potential risk variant at depression risk locus 14q21.1. A LocusZoom plot showed that rs1950834 was located in a genomic region with multiple SNPs showing significant associations with MDD. B rs1950834 was located in the upstream promoter region of LRFN5. C ChIP-seq data showed that H3k4me3, H3k4me1, and H3k27ac preferentially bound to the genomic sequence containing rs1950834 in the human brain tissue. The heights of the colored graphs reflect the ChIP-seq signal intensities, and the location of rs1950834 is highlighted with the red line. D DNase-seq of human brain tissue, neurogenic cells, and cell lines consistently revealed that rs1950834 was located in the chromatin open region. E ChIP-seq of human neurogenic cells and cell lines showed that rs1950834 binds to transcription factors RAD21 and POLR2A. F and G PWM plots of rs1950834 binding to RAD21 and POLR2A, respectively. ChIP-Seq, Chromatin immunoprecipitation and sequencing; CTCF, CCCTC-binding factor; DNase, deoxyribonuclease; EZH2, enhancer of zeste 2 polycomb repressive complex 2 subunits; LRFN5, leucine-rich repeat and fibronectin type III domain containing 5; MDD, major depressive disorder; POLR2A, RNA polymerase II subunit A; PWM, position weight matrix; RAD21, RAD21 cohesin complex component; SMARCA4, SWI/SNF related BAF chromatin remodeling complex subunit ATPase 4
rs1950834 affects the binding of RAD21 and POLR2A which regulate the expression of LRFN5
To investigate the potential promoter activity of different alleles of rs1950834, we performed dual luciferase reporter gene assays, which can be used to evaluate the regulatory effect of specific DNA sequences. Inserted into the promoter region of firefly luciferase in the reporter vector, the impact of target DNA sequences on transcription can be detected by normalizing firefly luciferase activity with the control Renilla luciferase activity. The results indicated that the MDD risk rs1950834-A allele exhibited significantly reduced promoter activity compared to the rs1950834-G allele in both neuroblastoma SH-SY5Y and glioblastoma U251 cells (Fig. A).
[Fig. 2] rs1950834 regulated LRFN5expression by affecting the binding of RAD21 and POLR2A. A Reporter gene assays showed that different alleles of rs1950834 significantly affected luciferase activity. Compared with the A allele, the G allele of rs1950834 conferred significantly higher luciferase activity in both SH-SY5Y and U251 cells (N = 16 in each group). Two-tailed unpaired t-test. ***P < 0.001. Mean ± SEM. B–D rs1950834 affected the binding affinity of nuclear protein extracts. The quantification data (grey values, arbitrary units) of the probes/protein complexes were first quantified using Image J. The obtained values were then normalized to the total quantity of probes. B For rs1950834, the G allele showed stronger binding of 6.4 µg or 12.8 µg nuclear extracts than the A allele (N = 2 in each group). Two-tailed unpaired t-test. *P < 0.05. **P < 0.01. Mean ± SEM. C and D rs1950834 affected RAD21 and POLR2A binding affinity. Competitive experiments showed that the G allele had stronger RAD21 and POLR2A binding than the A allele. E Significant enrichment of RAD21 and POLR2A on genomic sequences containing rs1950834 by ChIP-qPCR. DNA templates from cross-linked chromatins of cells (with IgG immunoprecipitation) were used in the IgG group (N = 4 in each group). One-way ANOVA followed by Tukey’s post hoc test. **P < 0.01, ***P < 0.001. Mean ± SEM. F ChIP-sequencing analysis on the binding ability of the A or G allele to RAD21 and POLR2A in the whole-genome context in SK-N-AS cells. G Analysis of the GSE172425 dataset on the expression of Lrfn5 after knockdown of Rad21 by siRNA in the primary cultured mouse cortical neurons (N = 11 in the Control group, and N = 12 in the siRad21 group). Two-tailed unpaired t-test. **P < 0.01. Mean ± SEM. H–LLRFN5 expression was significantly decreased by shRNA lentivirus targeting both RAD21A and RAD21Bin SH-SY5Y and U251 cells. The reduced expression of RAD21 mRNA (shown in Additional file 1: Fig. S2) or protein was found both in SH-SY5Y and U251 cells, confirming the successful knockdown of RAD21 (N = 3 in each group). One-way ANOVA followed by Tukey’s post hoc test. **P < 0.01, ***P < 0.001. Mean ± SEM. M The GSE125413 dataset showed that Polr2a mutations decreasedLrfn5 expression in mature neurons of mice (N= 3 in each group). Two-tailed unpaired t-test. **P < 0.01. Mean ± SEM. N–R POLR2A knockdown in SH-SY5Y and U251 cells affected LRFN5expression significantly (N = 3 in each group). One-way ANOVA followed by Tukey’s post hoc test. *P < 0.05, ***P < 0.001. Mean ± SEM
Electrophoretic mobility shift assay (EMSA) is widely used to evaluate the interaction between DNA sequences and transcription factors [[55], [56]]. When the target DNA sequence binds a specific transcription factor, the protein-probe complexes should migrate more slowly than the corresponding free probes during electrophoresis. The EMSA results indicated that rs1950834 significantly altered binding affinities to nuclear proteins extracted from SH-SY5Y cells (Fig. B). The G allele of rs1950834 showed a higher affinity to transcription factors than the A allele (Fig. B). ChIP-seq data showed that rs1950834 loci were bound by RAD21 and POLR2A. Consistently, the EMSA results confirmed that rs1950834 affected RAD21 and POLR2A binding (Fig. C and D), as the band of the protein-probe complex appeared when RAD21 or POLR2A protein was added. Additionally, the binding capacity of the probe with the G allele was significantly higher than that of the A allele (Fig. C and D).
ChIP-qPCR and ChIP-DNA Sanger sequencing were conducted in SK-N-AS cells to test if different alleles of rs1950834 alter RAD21 and POLR2A binding affinity. ChIP-qPCR showed significant enrichment of RAD21 and POLR2A binding on the DNA region containing rs1950834 (Fig. E). ChIP-DNA Sanger sequencing further revealed that RAD21 and POLR2A preferred to bind the G allele than A allele at rs1950834 (Fig. F). These results demonstrated that different alleles of rs1950834 altered the binding affinity of RAD21 and POLR2A to the genomic sequence.
Previous studies have found a significant correlation between RAD21 or POLR2A and LRFN5 in the human brain tissue of healthy subjects [[57]]. Initial analysis of the transcriptomic dataset GSE172425 [[58]] shows that the knockdown of Rad21 using small interfering RNA (siRNA) resulted in decreased expression of Lrfn5 in mouse cortical neurons (Fig. G). Then, after short hairpin RNA (shRNA) lentivirus targeting both RAD21A and RAD21B were constructed and delivered into SH-SY5Y and U251 cells (Fig. H–J), significantly decreased LRFN5 mRNA expression (Additional file 1: Fig. S2A–D) and protein levels (Fig. H, K, L) as compared with the controls were observed. Similarly, another gene expression dataset (GSE125413) [[59]] showed that the mutation of Polr2a in mouse mature neurons inhibited Lrfn5 expression (Fig. M). Similarly, knockdown of the POLR2A by shRNA (Fig. N–P) led to significantly decreased LRFN5 mRNA (Additional file 1: Fig. S2E–H) or protein (Fig. N, Q, R) levels in both SH-SY5Y and U251 cells. These findings underscored the regulatory role of RAD21 and POLR2A on LRFN5 expression. Collectively, rs1950834 may influence the expression of LRFN5 by regulating the DNA binding affinity of RAD21 and POLR2A.
rs1950834 directly regulates LRFN5 expression
Then we further elucidated whether rs1950834 can directly regulate the expression of the target gene LRFN5. The cis-expression quantitative trait loci (cis-eQTL) analysis showed that only LRFN5 expression was significantly associated with rs1950834 among all the genes located within 1 Mb upstream and downstream of rs1950834 in BrainMeta v2 and PsychENCODE (Fig. A and B). Therefore, the subsequent research focused on LRFN5 as the target gene of rs1950834. Moreover, we explored whether the effects attributed to the unique rs1950834 genotype on LRFN5 expression levels, in which an analysis of human brain eQTLs found that the levels of LRFN5 expression were affected by gene dose efficacy (Additional file 1: Fig. S3A–D). The A allele of rs1950834 reduced LRFN5 expression compared to the G allele in different human brain regions (Additional file 1: Fig. S3A-D). Importantly, the knockout of the DNA segment containing rs1950834 via CRISPR-Cas9 gene editing led to a decrease in both LRFN5 mRNA expression (Fig. C and E) and protein level (Fig. D and F) in SH-SY5Y cells. Furthermore, to provide direct evidence that rs1950834 affected LRFN5 expression levels, we successfully utilized single-base editing technology to generate rs1950834-AA and rs1950834-GG genotypes in SK-N-AS cells (Fig. G). This approach directly revealed the decreased expression of LRFN5 mRNA (Fig. H and I) and protein (Fig. J–L) in the MDD risk-associated rs1950834-AA genotype compared to the rs1950834-GG genotype in both naive and dexamethasone (DEX)-stressed conditions.
[Fig. 3] rs1950834 regulated the expression of LRFN5 as validated in gene editing cells. A and B The eQTL analysis of rs1950834 and its surrounding genes within a 1 M bp window, e.g., 500 kb upstream and 500 kb downstream of rs1950834, respectively, in BrainMeta v2 and PsychENCODE eQTL dataset. Expression of LRFN5 showed the most significant associations with rs1950834. C Schematic diagram and validation of rs1950834 knockout. D–F rs1950834 knockout downregulated expression of LRFN5mRNA and protein level in SH-SY5Y cells (N = 3 in each group). Two-tailed unpaired t-test. **P < 0.01, ***P < 0.001. Mean ± SEM. G The successful construction of rs1950834-AA and rs1950834-GG genotypes in SK-N-AS cells. H and I LRFN5 mRNA expression levels were lower in the rs1950834-AA genotype compared to the rs1950834-GG genotype in both naive and DEX-stressed states (N = 3 in each group). *P < 0.05, ***P < 0.001. Mean ± SEM. J Western blot analysis showed that LRFN5 protein levels were lower in the rs1950834-AA genotype compared to the rs1950834-GG genotype in both K naïve and L DEX-stressed states, N= 6 in each group. **P < 0.01, ***P < 0.001. Mean ± SEM. M Differentially expressed genes (DEGs) (Log2|FC > 2|, FDR < 0.05) of bulk RNA-seq in the rs1950834-AA and rs1950834-GG genotypes in naive and DEX-stressed conditions. N GO analysis of the differentially expressed genes. O KEGG analysis of the differentially expressed genes. Data were analyzed using the two-tailed unpaired t-test and are expressed as the mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001
Then, utilizing the SK-N-AS cells of rs1950834-AA and rs1950834-GG genotypes, we performed RNA-seq for transcriptomic analysis to further elucidate the pathways regulated by rs1950834. In the naive state, the rs1950834-AA genotype exhibited 2087 upregulated genes and 2480 downregulated genes, resulting in a total of 4567 differentially expressed genes (DEGs) compared with the rs1950834-GG genotype (Fig. M) (Log2|FC|> 2, FDR < 0.05). Under DEX stress conditions, the rs1950834-AA genotype exhibited 2800 upregulated genes and 3191 downregulated genes, resulting in a total of 5991 DEGs compared with the rs1950834-GG genotype (Fig. M) (Log2|FC|> 2, FDR < 0.05). Finally, to understand shared biological processes between rs1950834 genotypes in the presence and absence of DEX stress, we selected the intersection of DEGs among AA vs GG and AA + DEX vs GG + DEX, resulting in a total of 3764 DEGs for subsequent GO and KEGG analyses (Fig. M). In GO-biological processes enrichment, DEGs were mainly associated with neuron projection morphogenesis, cell–cell adhesion, and axonogenesis. In GO-cellular components, it was associated with synaptic structures, while in GO-molecular functions, it was related to cell adhesion (Fig. N). KEGG analysis revealed enrichment in differentially expressed genes in synapses and adhesion pathways, including focal adhesion, axon guidance, and cell adhesion molecules (Fig. O). This provides valuable insights into elucidating the rs1950834-regulated interactions among synaptic-associated proteins of the target gene LRFN5.
Role of NAc neuron-derived LRFN5 in depression
As the A allele of rs1950834 reduced promoter activity of LRFN5 via regulating the DNA binding affinity of transcription factor RAD21 and RNA polymerase subunit POLR2A, it was necessary to investigate cell type and brain region specificity of LRFN5 downregulation in MDD. We first performed tissue-specificity expression analysis of LRFN5 using expression data from the GTEx (https://www.gtexportal.org/) (GTEx Consortium, 2013). LRFN5 is expressed in many human tissues. However, LRFN5 expression in brain tissues is relatively high compared with other tissues (Additional file 1: Fig. S4). Cell-type-specific expression analysis using the data from the Allen Brain Map (https://celltypes.brain-map.org/rnaseq/human_m1_10x) showed that LRFN5 exhibits predominant expression in neurons and low expression in glial cells (Additional file 1: Fig. S5).
We also examined the associations between LRFN5 and other psychiatric disorders (including attention deficit and hyperactivity disorder, anorexia nervosa, anxiety disorder, autism spectrum disorder, bipolar disorder, and schizophrenia) using published GWAS summary statistics [[60]–[65]]. Genetic variants near LRFN5 (within a ± 250 kb window upstream and downstream of the gene) did not show genome-wide significant associations (P < 5 × 10−8) with attention deficit and hyperactivity disorder (Additional file 1: Fig. S6A), anorexia nervosa (Additional file 1: Fig. S6B), anxiety disorder (Additional file 1: Fig. S6C), autism spectrum disorder (Additional file 1: Fig. S6D), bipolar disorder (Additional file 1: Fig. S6E), and schizophrenia (Additional file 1: Fig. S6F), implying that the association of LRFN5 is specific to MDD.
Then, we performed a meta-analysis of transcription data of MDD patients from the GEO database to identify LRFN5 differentially expressed brain regions in depressed patients (Additional file 1: Fig. S7A–H). The results demonstrated no significant differential expression of LRFN5 in the peripheral blood, frontal cortex, amygdala, and cingulate cortex between the MDD patients and the controls. Interestingly, LRFN5 was selectively decreased in the NAc of MDD patients (Additional file 1: Fig. S7H). As one of the emotion-related brain regions, NAc could be important for LRFN5 function in MDD [[66]].
Thereby, to validate the above results, we examined whether the differential expression of Lrfn5 was neuronal cell subtype-specific in the NAc of CUMS-induced depressive-like mice. As shown in Fig. A and B, mice with CUMS-induced depression-like phenotypes exhibited decreased sucrose preference and increased immobile time in the TST and FST compared to wildtype mice. Importantly, a significant reduction in LRFN5 protein expression in NeuN+ neurons was found in the NAc of depressed mice (Fig. C), but not those of GFAP+ astrocytes (Fig. D) and Iba1+ microglia (Fig. E). The expression of LRFN5 in neurons was negatively correlated with depression-like phenotypes in SPT and FST (Fig. F), suggesting that NAc neuron-derived LRFN5 played critical roles in depression pathogenesis.
[Fig. 4] NAc neuron-derived Lrfn5was involved in depression. A Flow chart of the animal experiment for CUMS-induced depressive-like behavior. B Compared with the controls, the CUMS group showed decreased sucrose preference (P < 0.01), increased immobility time of the tail suspension test (P < 0.01), and forced swimming test (P < 0.05), N = 7 in the Control and CUMS group. C–E Immunofluorescence of LRFN5 with NeuN, GFAP, and Iba1 in mouse NAc. Compared with the control group, the number of NeuN+LRFN5+ cells in the CUMS group were decreased (P < 0.01), while GFAP+LRFN5+ (P > 0.5) and Iba1+LRFN5+ (P > 0.5) had no significant difference in NAc of mice (N = 3 in each group). F The number of NeuN+LRFN5+ cells was correlated to the scores of SPT and FST (N = 6 in each group). G–J Immunofluorescence of LRFN5 in the DRD1 or DRD2 MSNs in three subregions of NAc, including core, medial, and lateral shell, respectively (G and H). The results showed that DRD1+LRFN5+ and DRD2+LRFN5+ MSNs were generally downregulated in the NAc core, media, and lateral shell of CUMS-stressed mice compared to controls without the differences in subregion and neuronal subtype specificity (I and J), N = 6 in each group. DAPI, 4′,6-diamidino-2-phenylindole; DRD1, dopamine receptor D1; DRD2, dopamine receptor D2; MSN, medium spiny neuron; FST, forced swimming test; GFAP, glial fibrillary acidic protein; Iba1, ionized calcium-binding adaptor molecule 1; IF, immunofluorescence; L, lateral; LRFN5, leucine-rich repeat and fibronectin type III domain containing 5; M, medial; NAc, nucleus accumbens; NeuN, neuronal nuclei antigen; WB, Western blot. Data were analyzed using the two-tailed unpaired t-test and are expressed as the mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001
The NAc consists of three subregions, namely the nucleus, the medial shell, and the lateral shell, and 95% of the NAc is composed of medium spiny neurons (MSNs) with dopamine receptor D1 (DRD1) and dopamine receptor D2 (DRD2) markers, respectively. Subsequently, we explored whether LRFN5 dysregulation specifically occurred in functional subregions and neuronal subtypes. The results showed that LRFN5 was broadly downregulated in DRD1+ and DRD2+ MSNs in the NAc core, and the medial and lateral shell of CUMS-stressed mice compared to controls (as shown in Fig. G–J). Taking together, in CUMS-induced depressive-like mice, neuron-derived LRFN5 protein were ubiquitously reduced in NAc, with no subregion or neuronal cell type specificity.
NAc neuron-derived LRFN5 plays an important role in depression by affecting synapse-related functions
To further validate whether the NAc neuron-derivedLRFN5 was involved in the occurrence of depression, we constructed Lrfn5 overexpression or knockdown mouse models via rAAV-injection into NAc (Fig. A and Additional file 1: Fig. S8). Western blot analysis showed significantly reduced LRFN5 protein levels in the knockdown mice, while significantly increased LRFN5 protein levels were in the overexpression mice as compared to rAAV controls (Fig. B), confirming the successful construction of Lrfn5 gain-of-function and loss-of-function models. Compared with the rAAV control group, the rAAV knockdown mice exhibited depression-like behavior indicated by decreased sucrose preference and increased immobility time for forced swimming (Fig. C). Furthermore, Lrfn5 knockdown exacerbated the CUMS-induced depressive phenotypes of the mice (Fig. D–F), while the Lrfn5 overexpressing mice exhibited increased resistance to CUMS (Fig. D–F), suggesting that NAc neuron-derived LRFN5 plays a critical protective role in depression and may serve as a potential therapeutic target.
[Fig. 5] NAc neuron-derived Lrfn5 participates in the occurrence of depression. A Flow chart of the animal experiment for knockdown or overexpression of neuron-derived Lrfn5 in NAc on CUMS-induced depressive-like behaviors. B Western blotting measures LRFN5 protein levels among control, control rAVV, shLrfn5, and OE-Lrfn5 group mice to verify successful construction (N = 6 in each group). One-way ANOVA followed by post hoc Tukey’s test, *P < 0.05, ** P < 0.01, *** P < 0.001, ns, no significant difference. Each bar represents the mean ± SEM. C The downregulation of NAc neuron-derived Lrfn5 expression induced depressive-like behaviors, while the upregulation of NAc neuron-derived Lrfn5had no effects (N = 26 in Control group, N = 27 in Control rAAV group, N = 23 in shLrfn5 group, N = 26 in OE-Lrfn5 group). One-way ANOVA followed by post hoc Tukey’s test, *P < 0.05, **P < 0.01, *** P < 0.001, ns, no significant difference. Each bar represents the mean ± SEM. D Flow chart of the animal experiment for CUMS-induced depressive-like behavior in knockdown or overexpression of neuron-derived Lrfn5 in NAc mice. E Western blotting of LRFN5 levels among each group of mice (N = 6 in each group). One-way ANOVA followed by post hoc Tukey’s test, *P < 0.05, **P < 0.01, ***P < 0.001, ns, no significant difference. Each bar represents the mean ± SEM. F Knockdown of NAc neuron-derived Lrfn5 of mice increased susceptibility to CUMS-induced depressive-like behaviors, while overexpression of NAc neuron-derived Lrfn5 of mice increased resistance to the above phenotype (N = 25 in Control + Control rAAV group, N = 20 in CUMS group, N = 21 in CUMS + Control rAAV group, N = 11 in CUMS + shLrfn5 group, N = 20 in CUMS + OE-Lrfn5 group). One-way ANOVA followed by post hoc Tukey’s test, *P < 0.05, **P < 0.01, ***P < 0.001, ns, no significant difference. Each bar represents the mean ± SEM. KD, knockdown; OE, overexpression
As LRFN5 affects synaptic development and function [[17], [18]] through adhesion molecules [[19]], we first examined the role ofLRFN5 on synaptic functions in the NAs. Compared with the rAAV control group, the CUMS rAAV group showed markedly decreased expression of synapse-associated proteins, including protein tyrosine phosphatase receptor type D (PTPRD), postsynaptic density protein 95 (PSD95), and synaptosome-associated protein 25 (SNAP25) (Fig. A), along with a reduced total dendritic spine density (Fig. C), the density (Fig. E) and the proportion of stubby spines (Fig. F) in the NAc. As the changes in spine structure indicate synaptic activity, the dysregulation of spine morphology, especially for the mature thin, stubby, and mushroom subtypes, underlies CUMS-induced depression-like behaviors and serves as a target for combined magnetic stimulation system treatment [[51], [52]]. Interestingly, stubby spines lack a clear distinction between head and neck, which allows the diffusion of elevated calcium transients at the synapse into the dendritic shaft [[67]]. These observations were further aggravated in the CUMS + shLrfn5 group (Fig. C–F). In contrast, the Lrfn5 overexpression in NAc neurons significantly reversed the CUMS-induced synaptic abnormities (Fig. C–F).
[Fig. 6] Synaptic structure and function influenced by NAc neuron-derived Lrfn5 in depression. A Western blotting with quantification of synaptic protein levels in the NAc in Control + Control rAAV, CUMS (non-rAAV injected), CUMS + Control rAAV, CUMS + sh-Lrfn5, andCUMS + OE-Lrfn5 (N = 6 in each group). Two-way ANOVA followed by Tukey’s post hoc test, *P < 0.05, **P < 0.01, ***P < 0.001. B–F Golgi-Cox stained dendritic spines and spine density quantification in the NAc in Control + Control rAAV, CUMS (non-rAAV injected), CUMS + Control rAAV, CUMS + sh-Lrfn5, and CUMS + OE-Lrfn5 (N = 22 in Control group, N = 22 in CUMS group, N = 20 in CUMS + Control rAAV group, N = 19 in CUMS + shLrfn5 group, N = 23 in CUMS + OE-Lrfn5 group). D The schematic diagram of different types of spines. Filopodia spines are > 2 µm in length; the maximum width of stubby spines is less than their length; long thin spines are 1–2 µm in length; thin spines are < 1 µm in length; mushroom spines have a head/neck diameter ratio > 1; branched spines have more than one spine head attached to same spine neck. C One-way ANOVA followed by post hoc Tukey’s test, *P < 0.05, **P < 0.01, ***P < 0.001, ns, no significant difference. E and F Two-way ANOVA followed by Tukey’s post hoc test. G Representative traces of mEPSCs from the Control group (black), CUMS group (red), and OE-Lrfn5 + CUMS group (blue). H Left: Average amplitudes of mEPSCs for the Control group (N = 10), CUMS group (N = 8), and CUMS + OE-Lrfn5 group (N = 8). One-way ANOVA followed by post hoc Tukey’s test, ns, no significant difference. Right: Average frequencies of mEPSCs for the Control group (N = 10), CUMS group (N = 8), and CUMS + OE-Lrfn5 group (N = 8). One-way ANOVA followed by post hoc Tukey’s test, **P < 0.01. Each bar represents the mean ± SEM. I Representative traces of mIPSCs from the Control group (blank), CUMS group (red), and OE-Lrfn5 + CUMS group (blue). J Left: Average amplitudes of mIPSCs for the Control group (N = 8), CUMS group (N = 8), and OE-CUMS + Lrfn5 group (N = 10). One-way ANOVA followed by post hoc Tukey’s test, ns, no significant difference. Right: Average frequencies of mIPSCs for the Control group (N = 8), CUMS group (N = 8), and CUMS + OE-Lrfn5 group (N = 10). One-way ANOVA followed by post hoc Tukey’s test, ns, no significant difference. Each bar represents the mean ± SEM. K and L, Example traces (K) and quantification (L) of the paired-pulse ratio of electrical stimulation-evoked EPSCs in NAc neurons (N = 9 in Control group, N = 9 in CUMS group, N = 11 in CUMS + OE-Lrfn5 group). One-way ANOVA followed by post hoc Tukey’s test, **P < 0.01, ***P < 0.001. Each bar represents the mean ± SEM. mEPSC, miniature excitatory postsynaptic current; mIPSC, miniature inhibitory postsynaptic current
To investigate the impact of Lrfn5 overexpression on CUMS-impaired synaptic transmission of median-spiny neurons, whole-cell patch clamp recordings were performed on acute NAc brain slices. We observed a significant reduction in the frequency, but not amplitude, of miniature excitatory postsynaptic currents (mEPSCs) in CUMS rAAV mice (Fig. G and H), indicating a decrease in α-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) receptor-mediated excitatory synaptic transmission. This reduction in mEPSC frequency was ameliorated byLrfn5 overexpression (Fig. G and H). We did not observe a significant difference in the frequency and amplitude of miniature inhibitory postsynaptic currents (mIPSCs) between the control rAAV group, CUMS rAAV group, and Lrfn5 overexpressing mice (CUMS + OE-Lrfn5 mice) (Fig. I and J). Next, we examined presynaptic function by measuring the paired-pulse ratio (PPR). The PPR was significantly increased in the CUMS rAAV mice, indicating a reduction of presynaptic release probability. This increase in PPR was reversed by Lrfn5 overexpression (Fig. K and L). These results suggest that impaired synaptic transmission in the NAc of CUMS mice could be rescued by Lrfn5 overexpression.
These findings suggested that NAc neuron-derived LRFN5 is a key molecule involved in depression through the maintenance of synaptic function and might serve as a potential molecular target for MDD interventions.
Discussion
Our findings have demonstrated for the first time that the MDD risk-associated rs1950834 allele A decreased the binding of transcription factor RAD21 and RNA polymerase subunit POLR2A, leading to reduced promoter activity and expression of the target gene LRFN5. This study revealed that LRFN5 expression was significantly reduced in the NAc of MDD patients. Knockdown of NAc neuron-derived Lrfn5 induced depression-like behavior and further exacerbated the phenotype caused by CUMS in mice, while Lrfn5 overexpression induced resistance to CUMS. Furthermore, the study illustrated the potential mechanism of decreased NAc neuron-derived LRFN5 protein in depression via synaptic dysfunction. These findings offered new insights and evidence for the diagnosis, molecular targeted intervention, and precise neuromodulation of MDD.
This study comprehensively used bioinformatic tools, knockout or knock-in cell lines, and the depression-like mouse model. We successfully constructed SK-N-AS cells with rs1950834 AA and GG genotypes by CRISPR-Cas9-mediated single-base editing for the first time, which verified the bioinformatics findings that the role of rs1950834 is in regulating the target gene expression. Transcriptomic datasets were used to identify NAc as the key emotional brain region responsible for the dysregulation of LRFN5 in depression, which were further validated for the contribution to depression by rAAV-mediated gene overexpression or knockdown. Subregional and neuron subtype specificity for Lrfn5 downregulation was precisely characterized in CUMS-induced depression-like mice. The contribution of Lrfn5 to neuron structure and function was systematically validated by measuring synaptic proteins, Golgi staining, and electrophysiological analysis. Collectively, the utilization of state-of-the-art methodologies improves the reliability of experimental results.
The most important contribution of this study is that it provides systematic and direct evidence that MDD risk-associated allele A of rs1950834 exhibited reduced binding affinity with RAD21 and POLR2A, leading to reduced promoter activity and expression of LRFN5. The interpretation of abundant risk loci for MDD identified by GWAS has become crucial in the post-GWAS era [[7]]. In the previous study, we identified rs1950834 as one of the 34 MDD risk SNPs that disrupt the binding of 15 TFs by integrating the ChIP-Seq and PWM data. Luciferase reporter assay showed higher luciferase activities of the G allele than the A allele of rs1950834 in HEK293T, SH-SY5Y, and SN-K-SH cells. However, no further investigation of the rs1950834 SNP was performed in that study [[10]]. Here, we elucidated that RAD21 and POLR2A were enriched in the DNA sequence flanking rs1950834, and targeted knockdown of these transcription factors reduced the expression of target genes LRFN5. In particular, we successfully established homozygous rs1950834-AA and rs1950834-GG genotypic SK-N-AS cells via single-base editing technology for the first time. We directly elucidated that the risk allele AA reducedLRFN5expression compared to the protective allele GG, even in the stress state of DEX intervention. Consequently, our results robustly support the notion that rs1950834 is a functional variant for MDD via regulating target gene expression.
Another important finding of this study is that LRFN5 was selectively decreased in the NAc of MDD patients by meta-analysis of transcription data of MDD patients from the GEO database. Accordingly, LRFN5 expression was reduced in NAc neurons in CUMS-induced depression-like behavior mice, without the specificity of subregion and neuron subtype. Targeted knockdown of neuron-derived Lrfn5 in NAc induced depression-like behaviors in mice and further aggravated the CUMS-induced phenotypes. On the contrary, Lrfn5 overexpression in neurons can confer resistance to CUMS. The NAc is a component of the basal ganglia system and acts as a critical structure for reward and motivation-related behavior [[68], [69]]. Growing evidence supports that the NAc is a central brain region involved in affective circuits [[70]], contributing to the pathophysiology and neuronal-circuit impairment of MDD [[71]–[73]]. Importantly, precise neuromodulation of NAc, such as deep brain stimulation, has been shown to play a significant antidepressant role [[74], [75]]. Taken together, the results indicated that the reduction of NAc neuron-derived Lrfn5 leads to chronic stress-induced depression, which provides empirical evidence for precisely targeted antidepressants. On the other side, it should be noted that LRFN5 was also highly expressed in several brain regions, including amygdala, anterior cingulate cortex (BA24), caudate (basal ganglia), cerebellar hemisphere, cerebellum, cortex, frontal cortex (BA9), hippocampus, hypothalamus, putamen (basal ganglia), spinal cord (cervical c-1) and substantia nigra (Additional file 1: Fig. S3). Although this study is based on the findings of the human meta-analysis above, it first focuses on the NAs to reveal that the target gene is involved in chronic stress-induced depression in mice; we could not rule out the possibility that LRFN5 contributes to MDD pathogenesis in other emotion-related brain regions. It should also be noted that examining multiple regions in meta-analysis increases the risk of false positives. Thereby, it is necessary for further detailed exploration.
This study also elucidated that the knockdown of Lrfn5 in NAc neurons exacerbated CUMS-induced synaptic damage and depression-like behaviors in mice, while overexpression of Lrfn5 in NAc neurons reversed these phenotypes. LRFNs, also known as synaptic adhesion-like molecules (SALMs), are members of a family of synaptic adhesion molecules consisting of five known members: LRFN1-5 [[16], [76], [77]]. LRFN5 contains six leucine-rich repeats flanked by N-terminal and C-terminal cysteine-rich capping domains, an Ig domain, and a fibronectin III domain in the extracellular region, followed by a single transmembrane domain. LRFN5 was enriched in the synaptic fractions and could form complexes with PSD95 at excitatory synapses [[18]]. Knockdown of Lrfn5 in cultured mouse hippocampal neurons inhibited the number of both excitatory and inhibitory synapses. Consistently, the frequencies and amplitudes of mEPSCs and mIPSCs were also reduced in Lrfn5 knockdown neurons. Postsynaptic LRFN5 promotes synapse development and regulates AMPA-type glutamate receptor-mediated synaptic strength by trans-synaptically interacting with presynaptic leukocyte common antigen-related receptor protein tyrosine phosphatases [[18], [20]]. Importantly, we observed CUMS significantly reduced the frequency, but not amplitude, of mEPSCs in mice, indicating a decrease in AMPA receptor-mediated excitatory synaptic transmission. This reduction in mEPSCs frequency was ameliorated by Lrfn5 overexpression. These findings consistently confirm that LRFN5 can be used as a molecular target to regulate synaptic function and MDD treatment, and provide evidence for NAc as a target area for neuromodulation.
Several limitations remain for this study. First, the discrepancy in the genomic location of rs1950834 in humans (14q21.1) and that of mouse Lrfn5 (12qC1) complicactes the identification of the corresponding position of rs1950834 in mice. Consequently, the rs1950834 site-specific single-base edited mice could not be generated. Second, the absence of Lrfn5 knockout or conditional knockout mice may have limited our ability to thoroughly elucidate the mechanisms underlying the protective role of Lrfn5 on synaptic plasticity in depression. Third, it is necessary to note that LRFN5 in other brain regions may also contribute to the pathogenesis of depressive disorder, together with NAc. For example, Levey et al. reported on the potential involvement of LRFN5 in the frontal cortex in the pathogenesis of MDD [[6], [15]]. Therefore, further empirical investigations are required to validate these findings. In the future, we will also test whether LRFN5 is differentially expressed in neuron-derived exosomes of peripheral blood from MDD patients with a large sample size to establish a reliable diagnostic biomarker.
Conclusions
The study comprehensively elucidates the molecular regulatory mechanisms of MDD risk-associated variant rs1950834 at the single nucleotide level. The investigation has identified LRFN5 as the target gene of rs1950834 and has suggested that NAc neuron-derived LRFN5 was involved in depression by maintaining synapse-related functions. These findings provide valuable molecular and biological insights into the genetic regulatory mechanisms of MDD, offering a framework to elucidate functional variants at reported MDD risk loci. This study may broaden the boundaries of our understanding of MDD and provide new avenues for the development of novel therapeutic interventions.
Supplementary Information
Acknowledgements
Not applicable.
Abbreviations
Authors’ contributions
Z. Z., X. L., and Y. K. were responsible for the overall design and guidance of the experiment, and Z. Z. and Y. K. were responsible for the writing and revision of draft articles. D. L., Q. B., and T. Z. completed the experimental work, and D. L. partly participated in the draft preparation in the beginning. Y. L. guided graduate students X. C., X. D., and J. W. through the in vitro experiments and bioinformatics analysis, respectively, and Y. L. participated in draft writing, charting, and revision. Y. S. helps D. L. to build animal models. Under the guidance of Y. Z., T. Z., and S. J. collaborated to complete the model animal electrophysiology-related experiments and documentation. A. Z. generated the figures and tables under the guidance of Profs Z. Z., Y. K., and Y. L. All authors read and approved the manuscript.
Funding
This study was supported by China Science and Technology Innovation 2030—Major Project (2022ZD0211701, 2022ZD0211704 and 2021ZD0200700), National Natural Science Foundation of China (82471552, 82130042, 82371532, U2102205 and 82301690), Shenzhen Science and Technology Serial Funds (GJHZ20210705141400002, KCXFZ20211020164543006, JCYJ20220818101615033, ZDSYS20220606100606014, and KQTD20221101093608028), Postdoctoral Fellowship Program of CPSF (GZB20230136), and Open Program of Henan Key Laboratory of Biological Psychiatry (ZDSYS2022001).
Data availability
The data that support the findings of this study are available in ENCODE, PsychENCODE, The data that support the findings of this study are available from publicly accessible resources, including the Ensemble, https://www.ensembl.org/index.html; PsychENCODE, UCSC genome browser, https://genome.ucsc.edu/; SZDB, http://www.szdb.org.cn/index.html, RegulomeDB, https://regulomedb.org/regulome-search/; BrainSpan atlas, http://www.brainspan.org/; 3D Interaction Viewer and database (3DIV), https://kobic.kr/3divv1/; UCSC cell Browser https://cells.ucsc.edu/; the comprehensive human Super-Enhancer database (SEdb), and Human Protein Atlas, https://www.proteinatlas.org/. The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. The data of high-throughput RNA sequencing performed in this study have been deposited in the GEO database (reference number: GSE296333).
Declarations
Ethics approval and consent to participate
The use of animals in this experiment was approved by the Animal Experimentation Ethics Committee of Southeast University (20220104003 and 20221014001).
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Contributor Information
Yan Kong, Email: kongyancn@163.com.
Xiong-Jian Luo, Email: luoxiongjian@seu.edu.cn.
Zhijun Zhang, Email: janemengzhang@vip.163.com.
References
Associated Data
Supplementary Materials
Data Availability Statement
The data that support the findings of this study are available in ENCODE, PsychENCODE, The data that support the findings of this study are available from publicly accessible resources, including the Ensemble, https://www.ensembl.org/index.html; PsychENCODE, UCSC genome browser, https://genome.ucsc.edu/; SZDB, http://www.szdb.org.cn/index.html, RegulomeDB, https://regulomedb.org/regulome-search/; BrainSpan atlas, http://www.brainspan.org/; 3D Interaction Viewer and database (3DIV), https://kobic.kr/3divv1/; UCSC cell Browser https://cells.ucsc.edu/; the comprehensive human Super-Enhancer database (SEdb), and Human Protein Atlas, https://www.proteinatlas.org/. The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. The data of high-throughput RNA sequencing performed in this study have been deposited in the GEO database (reference number: GSE296333).