兽研研究亮点
[内容请查看原文]
犬化疗耐药被重新定义为一种时间依赖性过程。
[内容请查看原文]
静态耐药标志物可能遗漏早期适应性转变。
[内容请查看原文]
应激信号传导与microRNA调控构成可检验的时间窗口。
[内容请查看原文]
提出了一种分子桥接机制，但尚未在犬肿瘤中得到验证。
[内容请查看原文]
验证需要时间进程实验和细胞命运测定。
[内容请查看原文]
引言：犬化疗耐药中的时间性问题
获得性化疗耐药仍是犬淋巴瘤和骨肉瘤治疗中的重要限制因素。在原发结节性DLBCL及相关外周结节性B细胞淋巴瘤队列中，基于CHOP的化疗方案可获得较高的初始缓解率，但缓解期往往有限，疾病进展仍然常见（Childress等，2018；Davies等，2018；Hawkes等，2024；Vos等，2019）。在犬骨肉瘤中，局部肿瘤控制联合辅助化疗较单纯局部控制可改善生存，但转移性进展仍然常见，长期生存并不普遍（Berg等，1997；Selvarajah & Kirpensteijn，2010）。这些疾病特异性临床模式表明，有意义的初始治疗反应之后仍可出现后续进展，但这些模式并未确定导致治疗失败的分子序列。因此，它们为本文提出的时间性假说提供了动机，而非证明。
大多数化疗耐药的分子研究可以理解地聚焦于耐药状态本身。这种方法已鉴定出重要机制，包括ATP结合盒药物外排、凋亡信号改变、DNA修复增强、表观遗传改变以及microRNA（miRNA）失调（Craig等，2019；Elshafie等，2023；Elshafie等，2024；Klopfleisch等，2016；Packeiser等，2023；Zandvliet & Teske，2015；Zorzan等，2022）。这些机制具有生物学和临床相关性，但它们通常描述的是已建立耐药的分子表型。它们并不总能解释一个初始有反应的肿瘤如何走向复发，或哪些治疗后早期事件决定肿瘤走向敏感还是耐药轨迹。
这一时间维度仍未得到充分解析。诊断性活检、耐药细胞系或复发肿瘤样本可以揭示选择压力发生之后耐药的表现，但这些快照对化疗暴露后事件序列提供的信息有限。在犬肿瘤学中，相对较少的研究在治疗后最初数小时和数天内对肿瘤或模型系统进行足够密集的采样，以确定早期应激反应是否可重复地先于并与后期调控变化相关联。因此，应激信号传导、miRNA调控与细胞命运决定之间可能重要的时间关系仍难以评估。
本文提出，犬DLBCL和骨肉瘤中的化疗耐药不仅应作为静态分子状态来研究，还应作为连接早期治疗诱导的应激信号传导与后期miRNA介导的命运调控的可检验序列来研究。其新颖性不在于将个别应激通路或miRNA鉴定为耐药相关标志物，而在于检验它们的相对时序和功能依赖性是否揭示一个初始有反应的肿瘤如何走向持续存在或复发。在此，“时间耦合失败”是指这一假定的早期反应与延迟反应之间协调性的丧失或时序错乱，而非仅仅指任一类标志物的水平异常。该模型旨在作为工作假说，而非已确立的犬类范式。其价值取决于它是否能产生可证伪的预测，同时清晰区分犬类证据与跨物种推断。
[内容请查看原文]
静态耐药标志物是必要的，但在时间维度上不完整
已确立的耐药机制不应被否定。通过ATP结合盒转运蛋白的药物外排、凋亡控制改变、DNA修复能力、表观遗传调控以及非编码RNA网络的变化均参与耐药癌细胞的行为（Craig等，2019；Elshafie等，2023；Elshafie等，2024；Klopfleisch等，2016；Packeiser等，2023；Zandvliet & Teske，2015；Zorzan等，2022）。在犬肿瘤学中，这些机制为描述肿瘤为何对化疗反应性降低提供了基本词汇。它们也仍具有临床相关性，因为复发肿瘤和耐药细胞系常表现出稳定的分子特征，这些特征可合理支持其在治疗压力下存活。
局限不在于这些标志物不重要，而在于它们通常是在耐药已经建立之后才被测量。复发淋巴结活检、与不良结局相关的诊断样本或实验筛选的耐药细胞系可以鉴定在耐药疾病中富集的分子。然而，这些方法解析化疗暴露后早期事件序列的能力有限。它们很少能显示耐药是始于应激反应不足、肿瘤抑制性调控维持失败、预先存在亚克隆的选择，还是这些过程的某种组合。
这一区分对犬弥漫性大B细胞淋巴瘤具有重要意义。多项研究已探讨了与犬淋巴瘤复发、耐药或早期治疗失败相关的基因表达或miRNA模式（Craig等，2019；Elshafie等，2023；Elshafie等，2024；Tani等，2021；Toyoda等，2024；Zandvliet等，2014）。这些研究具有重要价值，因为它们识别了可能与结局相关的候选通路和生物标志物。然而，其中大多数研究并非旨在探究治疗后早期信号变化是否与后期miRNA变化在时间上存在关联。因此，miRNA可能被解读为预后或复发的静态生物标志物，而非时间依赖性调控反应的潜在组成部分。
犬骨肉瘤同样需要类似的审慎态度。分子和miRNA研究已揭示抑癌性和致癌性miRNA在侵袭性行为、侵袭及预后中的作用（Gourbault & Llobat，2020；Lopez等，2018；Ludwig等，2024）。这些发现支持miRNA调控在犬癌症中的生物学相关性，但仅凭这些发现尚不能确定这些变化在化疗后何时出现，也不能确定它们是否与早期应激信号存在因果关联。因此，静态分子关联可以产生关于耐药的假说，但仅凭其本身无法建立时间机制。
本观点文章的核心论点是，静态模型和时间模型应被视为互补关系。静态标志物描述耐药疾病的分子架构；时间模型则探究该架构是如何形成的。对于犬化疗耐药研究而言，下一个有价值的问题不仅是哪些标志物在敏感状态与耐药状态之间存在差异，还包括早期治疗诱导的应激反应是否可重复地先于并关联于后期调控变化。若要确立机制依赖性，则还需要单独的扰动实验验证。
本文将犬DLBCL和骨肉瘤作为互补的测试情境一并讨论，并非将其视为在生物学或治疗上等同的肿瘤。DLBCL的讨论背景为基于CHOP方案的化疗及缓解后复发，而骨肉瘤的讨论背景为局部控制、辅助化疗及转移进展（Berg等，1997；Childress等，2018；Davies等，2018；Hawkes等，2024；Selvarajah & Kirpensteijn，2010；Vos等，2019）。因此，其理由是比较性的。同一时间性问题可在两种不同的治疗情境中进行检验，而相关动力学、分子读数、甚至所提出的偶联机制是否存在，可能因肿瘤类型而异。来自任一肿瘤类型的证据均不应被视为在另一肿瘤类型中的验证，也不应推广至犬癌症整体。
为区分直接证据与跨物种合理性及假说生成，下文讨论的证据按四个层级组织。第一，已确立的犬类证据指直接在犬肿瘤或犬源癌症系统中观察到的结果，不暗示时间或因果偶联。第二，哺乳动物/人类证据指在犬肿瘤学之外证明的机制，可确立生物学合理性，但不代表犬特异性治疗动力学。第三，理论推断指通过整合原本独立的证据流所提出的关系。第四，所提出的假说指需要在犬源系统中进行前瞻性检验的具体时间和因果预测。在此层级框架下，当前犬类证据支持选定的应激反应组分和miRNA失调作为独立观察结果，而有序的早期应激至延迟miRNA序列以及YAP–DDX17–Microprocessor偶联仍属于理论推断和所提出的假说，而非已确立的犬类机制。
[内容请查看原文]
一项提出的时间偶联假说
Within these two disease-specific test contexts, the temporal coupling concept is presented as a hypothesis generated from the evidence hierarchy above, not as a mechanism established in canine tumours. It asks whether acquired chemoresistance may involve two broad regulatory windows: an early stress-signalling response and a later miRNA-regulatory response. The hypothesis is not intended to replace established mechanisms of resistance, nor does it imply that all canine tumours follow a single kinetic programme. Its purpose is narrower: to provide a testable way of asking whether the timing and coordination of treatment-induced responses influence whether a tumour cell moves towards death, persistence, or later relapse. This proposed temporal structure, including one candidate YAP–DDX17 mechanistic bridge and the boundary between supported and unvalidated claims, is summarised in Fig. 1.
[图：用于研究犬DLBCL和骨肉瘤化疗耐药性的 proposed temporal coupling hypothesis。该模型将 ac]
为避免将时间顺序与机制混淆，需区分四个推断步骤。时间优先性仅要求早期应激反应发生在后期miRNA变化之前。时间关联性要求早期反应与后期反应在重复观察中呈现可重复的关系。机制依赖性要求扰动所提出的上游信号节点能够改变后期miRNA反应。因果偶联是更强的主张，需要汇聚性证据表明下游反应是通过所提出的通路介导的，而非仅仅发生在上游反应之后或与其平行发生。因此，仅凭时间优先性或时间关联性不足以被视为机制或因果偶联的证据。对于假说检验而言，下游miRNA反应缺失、显著延迟或时间上不一致，将被视为反对所提出时间序列的证据。可重复的有序反应将支持时间关联性，而将其认定为机制依赖性或因果偶联则需要下文所述的额外扰动实验和通路特异性证据。
第一个窗口是细胞毒性暴露后的早期应激信号反应。原则上，这一阶段可能涉及感知DNA损伤、细胞骨架应激、氧化应激或更广泛的细胞稳态丧失的通路。候选通路包括Hippo–YAP信号传导、p53通路活性以及p38/c-Jun N端激酶（JNK）应激激活激酶信号传导（Bao et al., 2011; Canovas & Nebreda, 2021; Liu et al., 2021; Lv & Zhou, 2023; Raj & Bam, 2019; Zhang et al., 2009）。在犬类系统中，该反应的部分要素比其他要素得到更有力的支持。例如，DNA损伤性治疗后的p53通路激活有直接的犬类实验支持，而化疗诱导的Hippo–YAP动力学和犬癌细胞中p38的时序仍更多属于推断。这一区别很重要：通路组分的保守性支持生物学合理性，但并不能证实犬肿瘤中会发生相同的时序、幅度或调控后果。
第二个窗口是延迟的miRNA调控反应。miRNA是这一后期阶段的有吸引力的候选分子，因为它们可以协同影响凋亡、增殖、侵袭和应激适应。在犬淋巴瘤和骨肉瘤中，多项研究已将miRNA表达模式与预后、复发、侵袭性行为或肿瘤生物学联系起来，包括miR-21、miR-34a和14q32相关miRNA（Craig et al., 2019; Elshafie et al., 2023; Elshafie et al., 2024; Gourbault & Llobat, 2020; Lopez et al., 2018; Ludwig et al., 2024）。这些数据支持miRNA调控在犬肿瘤中的生物学相关性，但大多数现有研究仍为横断面研究。它们尚未表明化疗后miRNA变化是否以可重复的时间顺序出现，或是否由更早的应激信号事件驱动。重要的是，成熟miRNA丰度的变化不应等同于miRNA生物生成的改变。成熟miRNA水平反映转录、前体加工、下游成熟和RNA周转的综合效应；因此，仅凭成熟miRNA谱分析无法确定是哪个调控步骤产生了所观察到的变化。关于生物生成改变的主张需要直接检测前体和成熟miRNA种类，而非仅检测成熟miRNA丰度（Mori et al., 2014; Wu, 2020; Xu et al., 2022; Yu & Guan, 2014）。
因此，该假说考虑的是两个候选反应阶段，而非已证实的调控序列。细胞毒性治疗可能诱导早期应激反应，暂时改变肿瘤细胞的调控状态。理论推断是，随后的miRNA反应可能有助于巩固细胞命运。然而，任何延迟的miRNA反应都可能反映情境依赖的调控网络，而非肿瘤抑制性miRNA与促生存miRNA之间的简单平衡。其功能后果可能因肿瘤类型、治疗、细胞状态以及多种miRNA及其靶标的综合活性而异。因此，miR-34a、选定的14q32 miRNA和miR-21等候选miRNA仅作为示例性读数呈现，其方向性和功能意义需要在每种犬类治疗情境中进行实证检验。
该模型还重新定义了候选生物标志物的解读方式。对耐药相关标志物的单次基线测量可能具有临床实用性，但无法确定是否发生了有序的治疗后关系，更无法确定是否存在机制依赖性。同样，复发肿瘤样本可以显示选择的分子终点，但不能显示产生该终点的序列。
[内容请查看原文]
YAP–DDX17–Microprocessor作为候选机制桥梁之一
该时序假说不依赖于早期和延迟反应阶段之间的单一分子桥梁。YAP–DDX17–Microprocessor在此被视为一种实验上可操作的候选机制，因为哺乳动物研究提供了Hippo–YAP信号传导影响miRNA生物生成的合理途径。不应将其解释为时序假说的必需组分或犬类癌症中已确立的通路。
在哺乳动物细胞中，YAP定位是Hippo通路活性的主要输出。当Hippo信号传导失活时，YAP可在核内积累并支持与生长和生存相关的转录程序。当上游Hippo激酶被激活时，YAP被磷酸化并滞留于细胞质，从而降低其核活性（Bao et al., 2011; Liu et al., 2021; Lv & Zhou, 2023）。另外，人类细胞研究已将YAP与通过DDX17（也称为p72）相互作用调控miRNA生物生成联系起来，DDX17是一种DEAD-box RNA解旋酶，在pri-miRNA加工过程中参与Microprocessor功能（Mori et al., 2014; Wu, 2020; Xu et al., 2022; Yu & Guan, 2014）。这为YAP定位与随后成熟miRNA丰度变化之间提供了生物学上合理的桥梁。
然而，成熟miRNA丰度的改变本身并不能作为Microprocessor活性改变的证据。所提出的YAP–DDX17机制特异地涉及pri-to-pre-miRNA加工过程。因此，机制性检验必须通过平行测量pri-miRNA、pre-miRNA和成熟miRNA种类，来区分miRNA转录的变化与前体加工的变化。与处理相关的pri-to-pre加工改变伴随成熟miRNA丰度的相应下游改变，将更符合Microprocessor加工的改变；而仅限于成熟miRNA丰度的改变则不能确立该机制。这些模式仍应谨慎解读，因为前体稳定性和下游加工也可影响RNA丰度。
该机制与犬化疗耐药的相关性仍未解决。核心Hippo通路是保守的，DDX17/Microprocessor生物学在哺乳动物中广泛保守，但仅凭保守性并不能证明化疗在犬肿瘤细胞中诱导相同的YAP定位动态或YAP–DDX17相互作用。也不能证明任何此类相互作用控制处理后miRNA变化的时间或方向。这些是实证问题。
[内容请查看原文]
什么证据会支持或反驳该假说？
所提出假说的价值取决于能否通过实验建立越来越强的推断水平。在检验机制依赖性之前，应证明时间先后顺序和关联性，而因果耦合应保留给通过所提出机制将上游和下游反应联系起来的证据。只有在这些细胞和检测层面的基础建立之后，才可考虑后续的临床层面。在实验层面，药物暴露和采样时间表应在有可用数据时以犬药代动力学背景为依据，或通过初步剂量–反应实验来论证，这些实验应保留足够多的活细胞以研究适应性反应，而不仅仅是急性细胞毒性崩溃。
药代动力学背景尤为重要，因为给药后的时钟时间并不等同于生物学暴露时间。相关的时间窗口将取决于药物、给药途径和持续时间、全身暴露、消除动力学、适用时的活性代谢物，以及肿瘤暴露的程度和持续性。因此，早期和延迟采样点应尽可能锚定于临床相关的暴露特征，而不是原样跨药物或肿瘤类型转移。在CHOP等多药方案中，归因更为复杂，因为序贯或重叠的药物暴露可能产生不同或相互作用的应激反应；因此，机制研究应检查单个药物，或在解释时间效应时明确考虑治疗顺序。
首先，应确立时间先后顺序和关联性。在暴露于药理学合理浓度和临床相关细胞毒性药物方案的犬源癌症系统中，早期应激信号读数应可重复地先于延迟的miRNA变化。候选早期读数可包括YAP定位或磷酸化、p53通路激活和应激激活激酶信号传导（Bao et al., 2011; Canovas & Nebreda, 2021; Liu et al., 2021; Lv & Zhou, 2023; Raj & Bam, 2019; Zhang et al., 2009）。后期读数可包括与犬淋巴瘤或骨肉瘤生物学相关的成熟miRNA，如miR-21、miR-34a或选定的14q32 miRNA（Craig et al., 2019; Elshafie et al., 2023; Elshafie et al., 2024; Gourbault & Llobat, 2020; Lopez et al., 2018; Ludwig et al., 2024），作为候选网络组分而非预定义的敏感性或耐药性标志物进行测量。确切的时间窗口应通过经验校准，而不是作为通用阈值强加。此类校准应考虑药物特异性暴露和药效学时间，因为相同的名义采样时间对不同药物可能代表不同的生物学阶段。除时间顺序外，还应检验早期反应的变化是否与后期miRNA反应具有可重复的关联。这些分子测量还应与正交的细胞命运读数配对，包括早期凋亡或DNA损伤检测以及后期活力或生长检测。此类发现将支持时间关联，但本身不能确立机制依赖性还是因果耦合。由于肿瘤细胞内在反应可能被肿瘤微环境修饰，时间进程实验应尽可能区分肿瘤细胞读数与免疫或基质区室产生的信号。补充的离体、共培养或细胞类型分辨的组织分析将有助于确定所观察到的时间变化是肿瘤细胞内在的还是由微环境相互作用塑造的。所提出假说的证据结构和最小检验总结于表1。表1. 所提出时间耦合假说的证据状态、不确定性和最小检验。
其次，机制依赖性应通过扰动实验来检验。如果假设Hippo–YAP活性影响后续miRNA反应，那么扰动该轴应在化疗后改变该反应。可通过操控上游Hippo活性、YAP定位或YAP功能来检验，随后配对测量应激标志物、pri-/pre-/成熟miRNA种类及细胞命运结局。这些RNA种类应被解读为不同的读数：成熟miRNA丰度反映最终测量的池，而pri-miRNA和pre-miRNA的变化提供关于前体加工的更直接信息。在检测性能允许的情况下，pri-、pre-和成熟种类之间的关系可能支持关于加工效率的推断，但不应被视为生物合成改变的独立证据。此类实验应验证预期通路调控已实现，并在可行时使用互补扰动以减少脱靶解读。在验证的上游扰动后出现可重复的下游变化将支持机制依赖性，但本身不能确立所提出的YAP–DDX17途径介导该效应。相反，如果尽管验证了Hippo–YAP扰动，延迟的miRNA反应仍未改变，则不支持对该轴的依赖性。
如果将YAP–DDX17–Microprocessor作为候选桥梁之一进行追踪，通过该特定途径的因果耦合将需要超越时间顺序和上游扰动的证据。人类细胞研究支持Hippo信号、YAP、DDX17/p72和Microprocessor活性之间的机制联系（Mori等，2014；Wu，2020；Xu等，2022；Yu & Guan，2014），但这不能在犬肿瘤中假定。第一个要求是证明犬淋巴瘤或骨肉瘤细胞中存在基础YAP–DDX17相互作用，并确定该相互作用在细胞毒性治疗后是否改变。第二个是证明扰动所提出的相互作用会改变pri-to-pre-miRNA加工和下游miRNA反应。在实验可行的情况下，拯救或互补扰动实验将进一步增强因果归因。直接YAP–DDX17相互作用而没有相应的下游加工效应将确立物理关联，但不足以证明因果耦合。因此，成熟miRNA丰度改变而没有pri-to-pre-miRNA加工改变的证据将不支持Microprocessor介导的解释。未能验证该途径将反对YAP–DDX17–Microprocessor作为相关介质，但本身不会反驳更广泛的时间假说。
最后，临床转化应被视为下游检验，而非起点。犬弥漫性大B细胞淋巴瘤的纵向采样最终可以探究早期应激读数和后期miRNA模式是否与首次缓解持续时间相关。然而，所提出的液体活检标志物，包括外泌体YAP相关测量，需要分析验证才能支持临床声明。阴性临床关联不一定反驳模型的每个细胞方面，但会削弱其作为生物标志物策略的价值。
[内容请查看原文]
讨论：意义、边界和未来方向
本Opinion Paper的直接意义是方法学的而非治疗性的。如果获得性化疗耐药部分由治疗后反应的时间决定，那么在这些犬肿瘤背景下的研究不应仅依赖基线样本、复发样本或长期暴露后选择的耐药细胞系。这些设计仍然有用，但它们不适合检测早期治疗诱导的转变。更强的方法将包括在犬细胞系、离体肿瘤材料或精心选择的临床研究中设定明确的早期和延迟治疗后时间点，并配对测量应激信号、miRNA表达和细胞命运结局。
因此，药代动力学和治疗背景异质性是时间假说的组成部分，而非外围技术考虑。名义上的6、24或48小时间隔在不同细胞毒性药物后可能具有不同的生物学意义，取决于全身和肿瘤暴露、药物持久性、活性代谢物和治疗顺序。这与犬DLBCL尤其相关，其中CHOP方案使肿瘤暴露于多种具有不同药理学特征的药物，也与骨肉瘤相关，其中相关时间反应在辅助治疗所用药物之间可能不同。因此，实验时间过程应具有药物特异性和暴露知情性，临床采样应相对于实际治疗方案而非仅按钟表时间解读。
犬弥漫性大B细胞淋巴瘤是本研究合理的首选临床背景。它具有清晰的治疗结构、对CHOP为基础方案的高初始缓解率，以及缓解持续时间和复发方面具有临床意义的变异（Childress等，2018；Davies等，2018；Hawkes等，2024；Vos等，2019）。骨肉瘤同样具有相关性，尤其适用于在犬源模型中研究治疗失败、侵袭性行为以及肿瘤抑制性miRNA（Berg等，1997；Gourbault & Llobat，2020；Lopez等，2018；Ludwig等，2024；Selvarajah & Kirpensteijn，2010）。然而，该模型不应在未经直接验证的情况下推广至所有犬类癌症。不同的肿瘤类型、药物、遗传背景和治疗方案可能产生不同的时间模式。因此，DLBCL和骨肉瘤应分别进行评估，两者之间的一致性发现将增强该框架的更广泛相关性，但不能确立一种通用的犬类癌症机制。这种异质性是自发性犬类癌症的固有特征。因此，任何有用的时间模型都必须对个体间变异性、治疗背景和预先存在的耐药亚克隆保持稳健。
模型的边界同样重要。YAP–DDX17–Microprocessor轴仅是一种候选机制，并非经过验证的犬类通路。更广泛的时间假说不要求该特定轴必须发挥作用。本文所使用的miRNA示例不应被解读为二元性的敏感–耐药轴。miR-34a、选定的14q32 miRNA和miR-21作为示例性候选分子纳入，是因为犬类研究已将它们与肿瘤生物学、预后、复发或侵袭性行为相关联（Craig等，2019；Elshafie等，2023；Elshafie等，2024；Gourbault & Llobat，2020；Lopez等，2018；Ludwig等，2024），而非因为当前证据已确立其中任何一个为化疗敏感性或耐药性的方向特异性标志物。额外的犬类特异性证据来自乳腺肿瘤，其中miR-18a、miR-18b和miR-21在恶性肿瘤中较对照组织上调，而miR-146b在良性和恶性肿瘤中均下调（Abbate等，2022）。随后一项研究将miR-21上调与IL-6和TNF-α表达增加以及犬乳腺 carcinoma中通过Ki67评估的更高增殖活性相关联（Abbate等，2023）。这些横断面发现拓宽了犬类肿瘤中生物学相关miRNA失调的证据，但并未确立化疗诱导的时间性调控或化疗耐药。单个miRNA可调控多个靶点，其功能效应可能随肿瘤类型、治疗和细胞状态而变化。因此，miRNA发现应被解读为情境依赖性调控模式，而非孤立的二元开关。同样，细胞外囊泡或液体活检读数在未来纵向研究中可能变得有用，但在分析性能、生物学来源及与结局的关联得到确立之前，不应将其视为经过验证的生物标志物。
肿瘤微环境是纯肿瘤细胞内在模型未能捕获的另一个重要生物学背景来源。在犬DLBCL中，转录研究已鉴定出与临床结局相关的免疫和炎症程序，包括T细胞、巨噬细胞和免疫检查点相关信号（Aresu等，2019）。犬骨肉瘤同样包含异质性免疫微环境，包括多种肿瘤相关巨噬细胞群体和预测的树突状细胞/T细胞相互作用（Ammons等，2024），并且巨噬细胞浸润与接受截肢和辅助卡铂治疗的犬的无病间期相关（Withers等，2019）。这些观察支持微环境的生物学相关性，但并未确立其驱动所提出的化疗后时间序列。细胞因子信号传导、免疫细胞组成、基质相互作用、缺氧和细胞外囊泡交换可能修饰早期应激反应或后期miRNA调控，也可能贡献在 bulk 肿瘤样本中检测到的信号。因此，未来的时间性研究应在实验可行的情况下尽可能区分肿瘤细胞内在变化与微环境来源的反应。
跨物种时间分辨证据为此类时间设计提供了有限的先例。在K562慢性髓性白血病细胞中，伊马替尼暴露与12小时早期Hippo相关转录反应相关，随后在48小时出现Hippo相关miRNA的延迟上调（Akbari-Ardabili等，2026）。然而，这些发现是相关性的，获取自人类细胞系模型，且未在原代CD34⁺ CML样本中重现为一致的Hippo–YAP转录组反应。因此，它们支持测试时间分离的Hippo和miRNA相关反应的可行性，但不支持所提出的机制在犬类肿瘤中的存在。
本观点文章并不主张时间耦合是化疗耐药的唯一解释。耐药性疾病可能通过克隆选择、预先存在的异质性、药物外排、凋亡受损、DNA修复改变、免疫与基质微环境影响、药代动力学变异或治疗异质性等途径产生。所提出的模型只有在其有助于将这些因素组织为可检验的时间性问题时才有用。因此，其近期价值不在于临床预测，而在于有纪律的假说生成：识别治疗反应何时开始失败、涉及哪些调控转变，以及哪些主张可在犬类系统中得到支持、修正或否定。
基金资助
本研究未获得来自公共、商业或非营利部门资助机构的任何专项拨款。
伦理声明
本文不涉及新的动物实验、人类参与者或私人拥有的动物数据。因此，无需伦理批准。
关于稿件准备过程中生成式人工智能和AI辅助技术的声明
在本工作的准备和修订过程中，作者使用ChatGPT（OpenAI）进行语言润色和编辑组织。使用该工具后，作者根据需要审查和编辑了内容，独立核实了科学主张和参考文献，并对发表文章的内容承担全部责任。
利益冲突声明
作者声明，他们没有已知的可能影响本文工作的竞争性经济利益或个人关系。
查看英文原文▾
Highlights
[内容请查看原文]
Canine chemoresistance is reframed as a time-dependent process.
[内容请查看原文]
Static resistance markers may miss early adaptive transitions.
[内容请查看原文]
Stress signalling and microRNA regulation form testable temporal windows.
[内容请查看原文]
One molecular bridge is proposed but remains unvalidated in canine tumours.
[内容请查看原文]
Validation requires time-course experiments and cell-fate measurements.
[内容请查看原文]
Introduction: temporal questions in canine chemoresistance
Acquired chemoresistance remains an important limitation in the treatment of canine lymphoma and osteosarcoma. In primary nodal DLBCL and related peripheral nodal B-cell lymphoma cohorts, CHOP-based chemotherapy produces high initial response rates, but remissions are often finite and disease progression remains common (Childress et al., 2018; Davies et al., 2018; Hawkes et al., 2024; Vos et al., 2019). In canine osteosarcoma, local tumour control combined with adjuvant chemotherapy improves survival relative to local control alone, yet metastatic progression remains common and long-term survival is uncommon (Berg et al., 1997; Selvarajah & Kirpensteijn, 2010). These disease-specific clinical patterns establish that meaningful initial treatment responses can be followed by later progression, but they do not identify the molecular sequence responsible for treatment failure. They therefore motivate, rather than demonstrate, the temporal hypothesis developed here.
Most molecular studies of chemoresistance have understandably focused on the resistant state itself. This approach has identified important mechanisms, including ATP-binding cassette drug efflux, altered apoptotic signalling, enhanced DNA repair, epigenetic change, and microRNA (miRNA) dysregulation (Craig et al., 2019; Elshafie et al., 2023; Elshafie et al., 2024; Klopfleisch et al., 2016; Packeiser et al., 2023; Zandvliet & Teske, 2015; Zorzan et al., 2022). These mechanisms are biologically and clinically relevant, but they often describe the molecular phenotype of established resistance. They do not always explain how an initially responsive tumour moves towards relapse, or which early post-treatment events determine whether a tumour follows a sensitive or resistant trajectory.
This temporal dimension remains poorly resolved. A diagnostic biopsy, a resistant cell line, or a relapsed tumour sample can reveal what resistance looks like after selection has occurred, but these snapshots provide limited information about the sequence of events that unfolded after chemotherapy exposure. In canine oncology, relatively few studies have sampled tumours or model systems densely enough during the first hours and days after treatment to determine whether early stress responses reproducibly precede and are associated with later regulatory changes. As a result, potentially important temporal relationships between stress signalling, miRNA regulation, and cell-fate commitment remain difficult to evaluate.
This Opinion Paper proposes that chemoresistance in canine DLBCL and osteosarcoma be studied not only as a static molecular state but also as a testable sequence linking early treatment-induced stress signalling to later miRNA-mediated fate regulation. The novelty does not lie in identifying individual stress pathways or miRNAs as resistance-associated markers. Rather, it lies in testing whether their relative timing and functional dependence reveal how an initially responsive tumour moves towards persistence or relapse. Here, “temporal coupling failure” refers to loss or mistiming of this proposed coordination between the early and delayed responses, rather than simply an abnormal level of either marker class. The model is intended as a working hypothesis, not an established canine paradigm. Its value depends on whether it generates falsifiable predictions while clearly distinguishing canine evidence from cross-species inference.
[内容请查看原文]
Static resistance markers are necessary but temporally incomplete
Established resistance mechanisms should not be dismissed. Drug efflux through ATP-binding cassette transporters, altered apoptotic control, DNA repair capacity, epigenetic regulation, and changes in non-coding RNA networks all contribute to the behaviour of resistant cancer cells (Craig et al., 2019; Elshafie et al., 2023; Elshafie et al., 2024; Klopfleisch et al., 2016; Packeiser et al., 2023; Zandvliet & Teske, 2015; Zorzan et al., 2022). In canine oncology, these mechanisms have provided an essential vocabulary for describing why a tumour may become less responsive to chemotherapy. They also remain clinically relevant, because relapsed tumours and drug-resistant cell lines often display stable molecular features that can plausibly support survival under treatment pressure.
The limitation is not that these markers are unimportant, but that they are usually measured after resistance is already established. A relapsed lymph node biopsy, a diagnostic sample associated with poor outcome, or an experimentally selected resistant cell line can identify molecules enriched in resistant disease. However, these approaches have limited ability to resolve the early sequence of events that occurs after chemotherapy exposure. They rarely show whether resistance begins with an inadequate stress response, a failure to sustain tumour-suppressive regulation, selection of a pre-existing subclone, or some combination of these processes.
This distinction is important for canine diffuse large B-cell lymphoma. Several studies have examined gene-expression or miRNA patterns associated with relapse, drug resistance, or early treatment failure in canine lymphoma (Craig et al., 2019; Elshafie et al., 2023; Elshafie et al., 2024; Tani et al., 2021; Toyoda et al., 2024; Zandvliet et al., 2014). These studies are valuable because they identify candidate pathways and biomarkers that may be linked to outcome. Yet most of them are not designed to ask whether early post-treatment signalling changes are temporally connected to later miRNA shifts. As a result, miRNAs may be interpreted as static biomarkers of prognosis or relapse, rather than as potential components of a time-dependent regulatory response.
A similar caution applies to canine osteosarcoma. Molecular and miRNA studies have implicated tumour-suppressive and oncogenic miRNAs in aggressive behaviour, invasion, and prognosis (Gourbault & Llobat, 2020; Lopez et al., 2018; Ludwig et al., 2024). These findings support the biological relevance of miRNA regulation in canine cancer, but they do not by themselves define when those changes arise after chemotherapy or whether they are causally linked to early stress signalling. Static molecular associations can therefore generate hypotheses about resistance, but they cannot alone establish a temporal mechanism.
The central point of this Opinion Paper is that static and temporal models should be treated as complementary. Static markers describe the molecular architecture of resistant disease; temporal models ask how that architecture is reached. For canine chemoresistance research, the next useful question is not only which markers differ between sensitive and resistant states, but whether early treatment-induced stress responses reproducibly precede and are associated with later regulatory changes. Mechanistic dependence would then require separate perturbational testing.
Canine DLBCL and osteosarcoma are considered together here as complementary test contexts, not as biologically or therapeutically equivalent tumours. DLBCL is considered in the setting of CHOP-based therapy and remission followed by relapse, whereas osteosarcoma is considered in the setting of local control, adjuvant chemotherapy, and metastatic progression (Berg et al., 1997; Childress et al., 2018; Davies et al., 2018; Hawkes et al., 2024; Selvarajah & Kirpensteijn, 2010; Vos et al., 2019). The rationale is therefore comparative. The same temporal question can be tested in two distinct treatment contexts, while the relevant kinetics, molecular readouts, and even the presence of the proposed coupling mechanism may differ by tumour type. Evidence from either tumour type should not be taken as validation in the other or extrapolated to canine cancer generally.
To distinguish direct evidence from cross-species plausibility and hypothesis generation, the evidence discussed below is organised into four levels. First, established canine evidence refers to observations demonstrated directly in canine tumours or dog-derived cancer systems, without implying temporal or causal coupling. Second, mammalian/human evidence refers to mechanisms demonstrated outside canine oncology that establish biological plausibility but not canine-specific treatment kinetics. Third, theoretical inference refers to relationships proposed by integrating otherwise separate evidence streams. Fourth, proposed hypotheses are the specific temporal and causal predictions that require prospective testing in dog-derived systems. Under this hierarchy, current canine evidence supports selected stress-response components and miRNA dysregulation as separate observations, whereas the ordered early-stress-to-delayed-miRNA sequence and the YAP–DDX17–Microprocessor coupling remain theoretical inferences and proposed hypotheses rather than established canine mechanisms.
[内容请查看原文]
A proposed temporal coupling hypothesis
Within these two disease-specific test contexts, the temporal coupling concept is presented as a hypothesis generated from the evidence hierarchy above, not as a mechanism established in canine tumours. It asks whether acquired chemoresistance may involve two broad regulatory windows: an early stress-signalling response and a later miRNA-regulatory response. The hypothesis is not intended to replace established mechanisms of resistance, nor does it imply that all canine tumours follow a single kinetic programme. Its purpose is narrower: to provide a testable way of asking whether the timing and coordination of treatment-induced responses influence whether a tumour cell moves towards death, persistence, or later relapse. This proposed temporal structure, including one candidate YAP–DDX17 mechanistic bridge and the boundary between supported and unvalidated claims, is summarised in Fig. 1.
[图：Proposed temporal coupling hypothesis for studying chemoresistance in canine DLBCL and osteosarcoma. The model frames ac]
To avoid conflating temporal order with mechanism, four inferential steps are distinguished. Temporal precedence requires only that an early stress response occur before a later miRNA change. Temporal association requires a reproducible relationship between the early and later responses across repeated observations. Mechanistic dependence requires perturbation of the proposed upstream signalling node to alter the later miRNA response. Causal coupling is a stronger claim and requires convergent evidence that the downstream response is mediated through the proposed pathway rather than merely occurring after, or in parallel with, the upstream response. Thus, temporal precedence or temporal association alone would not be considered evidence of mechanistic or causal coupling. For hypothesis testing, an absent, markedly delayed, or temporally inconsistent downstream miRNA response would be considered evidence against the proposed temporal sequence. A reproducibly ordered response would support temporal association, whereas designation as mechanistically dependent or causally coupled would require the additional perturbational and pathway-specific evidence described below.
The first window is an early stress-signalling response after cytotoxic exposure. In principle, this phase may involve pathways that sense DNA damage, cytoskeletal stress, oxidative stress, or broader loss of cellular homeostasis. Candidate pathways include Hippo–YAP signalling, p53 pathway activity, and p38/c-Jun N-terminal kinase (JNK) stress-activated kinase signalling (Bao et al., 2011; Canovas & Nebreda, 2021; Liu et al., 2021; Lv & Zhou, 2023; Raj & Bam, 2019; Zhang et al., 2009). In canine systems, some elements of this response are better supported than others. For example, p53 pathway activation after DNA-damaging treatment has direct canine experimental support, whereas chemotherapy-induced Hippo–YAP kinetics and p38 timing in canine cancer cells remain more inferential. This distinction matters: conservation of pathway components supports biological plausibility, but it does not establish that the same timing, amplitude, or regulatory consequences occur in canine tumours.
The second window is a delayed miRNA-regulatory response. miRNAs are attractive candidates for this later phase because they can coordinately influence apoptosis, proliferation, invasion, and stress adaptation. In canine lymphoma and osteosarcoma, several studies have linked miRNA expression patterns with prognosis, relapse, aggressive behaviour, or tumour biology, including miR-21, miR-34a, and 14q32-associated miRNAs (Craig et al., 2019; Elshafie et al., 2023; Elshafie et al., 2024; Gourbault & Llobat, 2020; Lopez et al., 2018; Ludwig et al., 2024). These data support the biological relevance of miRNA regulation in canine tumours, but most available studies remain cross-sectional. They do not yet show whether miRNA changes emerge after chemotherapy in a reproducible temporal order or whether they are driven by earlier stress-signalling events. Importantly, changes in mature miRNA abundance should not be equated with altered miRNA biogenesis. Mature miRNA levels reflect the combined effects of transcription, precursor processing, downstream maturation, and RNA turnover; mature-miRNA profiling alone therefore cannot identify which regulatory step produced an observed change. Claims concerning altered biogenesis require direct examination of precursor and mature miRNA species rather than mature miRNA abundance alone (Mori et al., 2014; Wu, 2020; Xu et al., 2022; Yu & Guan, 2014).
The hypothesis therefore considers two candidate response phases rather than a demonstrated regulatory sequence. Cytotoxic therapy may induce an early stress response that temporarily alters the regulatory state of the tumour cell. The theoretical inference is that a later miRNA response may then contribute to consolidation of cell fate. However, any delayed miRNA response is likely to reflect a context-dependent regulatory network rather than a simple balance between tumour-suppressive and pro-survival miRNAs. Its functional consequences may vary with tumour type, treatment, cellular state, and the combined activity of multiple miRNAs and their targets. Candidate miRNAs such as miR-34a, selected 14q32 miRNAs, and miR-21 are therefore presented only as illustrative readouts whose direction and functional significance require empirical testing in each canine treatment context.
This model also reframes how candidate biomarkers should be interpreted. A single baseline measurement of a resistance-associated marker may be clinically useful, but it cannot determine whether an ordered post-treatment relationship occurred, much less whether mechanistic dependence was present. Similarly, a relapsed tumour sample can show the molecular endpoint of selection, but not the sequence that produced it.
[内容请查看原文]
YAP–DDX17–Microprocessor as one candidate mechanistic bridge
The temporal hypothesis does not depend on a single molecular bridge between the early and delayed response phases. YAP–DDX17–Microprocessor is considered here as one experimentally tractable candidate mechanism because mammalian studies provide a plausible route by which Hippo–YAP signalling could influence miRNA biogenesis. It should not be interpreted as a required component of the temporal hypothesis or as an established pathway in canine cancer.
In mammalian cells, YAP localisation is a major output of Hippo pathway activity. When Hippo signalling is inactive, YAP can accumulate in the nucleus and support transcriptional programmes related to growth and survival. When upstream Hippo kinases are activated, YAP is phosphorylated and retained in the cytoplasm, reducing its nuclear activity (Bao et al., 2011; Liu et al., 2021; Lv & Zhou, 2023). Separately, human-cell studies have linked YAP to the regulation of miRNA biogenesis through interaction with DDX17, also known as p72, a DEAD-box RNA helicase that contributes to Microprocessor function during pri-miRNA processing (Mori et al., 2014; Wu, 2020; Xu et al., 2022; Yu & Guan, 2014). This provides a biologically plausible bridge between YAP localisation and later changes in mature miRNA abundance.
However, altered mature miRNA abundance is not itself evidence of altered Microprocessor activity. The proposed YAP–DDX17 mechanism concerns pri-to-pre-miRNA processing specifically. A mechanistic test must therefore distinguish changes in miRNA transcription from changes in precursor processing by measuring pri-miRNA, pre-miRNA, and mature miRNA species in parallel. A treatment-associated change in pri-to-pre processing accompanied by a corresponding downstream change in mature miRNA abundance would be more consistent with altered Microprocessor processing, whereas a change confined to mature miRNA abundance would not establish this mechanism. These patterns should still be interpreted cautiously because precursor stability and downstream processing can also affect RNA abundance.
The relevance of this mechanism to canine chemoresistance remains unresolved. The core Hippo pathway is conserved, and DDX17/Microprocessor biology is broadly conserved across mammals, but conservation alone does not prove that chemotherapy induces the same YAP localisation dynamics or YAP–DDX17 interaction in canine tumour cells. Nor does it prove that any such interaction controls the timing or direction of miRNA changes after treatment. These are empirical questions.
[内容请查看原文]
What evidence would support or refute the hypothesis?
The value of the proposed hypothesis depends on whether increasingly strong levels of inference can be established experimentally. Temporal precedence and association should be demonstrated before mechanistic dependence is tested, while causal coupling should be reserved for evidence linking the upstream and downstream responses through the proposed mechanism. A later clinical layer may be considered only after these cellular and assay-level foundations are established. At the experimental level, drug exposure and sampling schedules should be justified by canine pharmacokinetic context where available, or by pilot dose–response experiments that leave enough viable cells to study adaptive responses rather than only acute cytotoxic collapse.
Pharmacokinetic context is particularly important because clock time after drug administration is not equivalent to biological exposure time. The relevant temporal window will depend on the agent, route and duration of administration, systemic exposure, elimination kinetics, active metabolites where applicable, and the extent and persistence of tumour exposure. Accordingly, early and delayed sampling points should be anchored, where possible, to clinically relevant exposure profiles rather than transferred unchanged across drugs or tumour types. In multi-agent protocols such as CHOP, attribution is more complex because sequential or overlapping drug exposures may generate distinct or interacting stress responses; mechanistic studies should therefore examine individual agents or explicitly account for treatment sequence when interpreting temporal effects.
First, temporal precedence and association should be established. In dog-derived cancer systems exposed to pharmacologically justified concentrations and schedules of clinically relevant cytotoxic agents, early stress-signalling readouts should reproducibly precede delayed miRNA changes. Candidate early readouts may include YAP localisation or phosphorylation, p53 pathway activation, and stress-activated kinase signalling (Bao et al., 2011; Canovas & Nebreda, 2021; Liu et al., 2021; Lv & Zhou, 2023; Raj & Bam, 2019; Zhang et al., 2009). Later readouts may include mature miRNAs implicated in canine lymphoma or osteosarcoma biology, such as miR-21, miR-34a, or selected 14q32 miRNAs (Craig et al., 2019; Elshafie et al., 2023; Elshafie et al., 2024; Gourbault & Llobat, 2020; Lopez et al., 2018; Ludwig et al., 2024), measured as candidate network components rather than predefined markers of sensitivity or resistance. The exact time windows should be empirically calibrated rather than imposed as universal thresholds. Such calibration should consider drug-specific exposure and pharmacodynamic timing, because identical nominal sampling times may represent different biological phases for different agents. Beyond temporal ordering, variation in the early response should be tested for reproducible association with the later miRNA response. These molecular measurements should also be paired with orthogonal cell-fate readouts, including early apoptosis or DNA-damage assays and later viability or outgrowth assays. Such findings would support temporal association but would not, by themselves, establish mechanistic dependence or causal coupling. Because tumour-cell intrinsic responses may be modified by the tumour microenvironment, time-course experiments should, where feasible, distinguish neoplastic-cell readouts from signals arising in immune or stromal compartments. Complementary ex vivo, co-culture, or cell-type-resolved tissue analyses would help determine whether observed temporal changes are intrinsic to tumour cells or are shaped by microenvironmental interactions. The evidence structure and minimal tests for the proposed hypothesis are summarised in Table 1.
Table 1. Evidence status, uncertainties, and minimal tests for the proposed temporal coupling hypothesis.
| Model component | Current evidence status | Key uncertainty | Minimal test | Interpretation if not supported |
|---|---|---|---|---|
| Early stress-signalling window | Biologically plausible. Hippo–YAP, p53, and p38/JNK pathways are relevant to cancer stress responses; direct canine support is stronger for p53 pathway activity than for chemotherapy-induced Hippo–YAP kinetics. | Whether canine tumour cells show reproducible early stress-signalling kinetics under drug-specific, pharmacologically relevant exposure conditions. | Perform drug-specific time-course studies in dog-derived lymphoma and osteosarcoma systems using pharmacologically justified concentrations and exposure schedules; measure YAP localisation or phosphorylation, p53 pathway activity, and stress-activated kinase signalling after treatment. | Absence of reproducible early stress signalling would weaken the model and require reconsideration of the proposed initiating phase. |
| Delayed miRNA-regulatory window | Supported as a candidate regulatory layer. Canine lymphoma and osteosarcoma studies link miRNA patterns, including miR-21, miR-34a, and 14q32-associated miRNAs, with tumour biology, prognosis, relapse, or aggressive behaviour. These associations do not establish a binary sensitivity–resistance role for individual miRNAs. | Whether post-treatment miRNA changes emerge in a reproducible temporal order and whether changes in mature miRNA abundance reflect altered biogenesis or other transcriptional/post-transcriptional processes. | Measure mature miRNAs across early and delayed post-treatment time points; where altered biogenesis is proposed, quantify pri-, pre-, and mature miRNA species in parallel, ideally paired with cell-fate readouts. | Inconsistent mature-miRNA changes would weaken the proposed delayed regulatory phase; mature-miRNA changes alone would not establish altered biogenesis. |
| Tumour microenvironment context | Biologically relevant but not integrated into the temporal hypothesis experimentally. Canine DLBCL and osteosarcoma show immune and stromal heterogeneity associated with tumour biology or outcome ( Ammons et al., 2024 ; Aresu et al., 2019 ; Withers et al., 2019 ). | Whether immune or stromal compartments modify early stress signalling, delayed miRNA responses, or both, and whether bulk-tissue signals originate from neoplastic versus non-neoplastic cells. | Where feasible, complement tumour-cell assays with ex vivo, co-culture, or cell-type-resolved tissue analyses across matched time points, including relevant immune or stromal readouts. | Failure to account for microenvironmental contributions could confound attribution of temporal changes to tumour-cell-intrinsic mechanisms; absence of an effect in one model would not exclude context-dependent effects in vivo. |
| Temporal precedence and association | Conceptual basis of the hypothesis, but not yet established in canine cancer systems. | Whether early stress readouts reproducibly precede and are associated with later miRNA changes in canine tumour systems. | Use repeated time-course experiments with predefined sampling windows, then test associations between early stress readouts, later miRNA responses, and cell-fate outcomes. | Lack of reproducible temporal ordering or association would argue against the proposed biphasic relationship. |
| Hippo–YAP dependence of the miRNA response | Plausible but unvalidated in canine systems. Mammalian evidence supports Hippo–YAP regulation of miRNA biology, but canine chemotherapy-induced dependence has not been demonstrated. | Whether the delayed miRNA response depends on Hippo–YAP activity in canine tumour cells. | Perturb upstream Hippo activity or YAP localisation/function, verify pathway modulation, and measure delayed miRNA processing and mature miRNA abundance after chemotherapy. | If the delayed miRNA response is unchanged despite verified Hippo–YAP perturbation, mechanistic dependence on this axis would not be supported. |
| Candidate YAP–DDX17–Microprocessor-mediated coupling | Mechanistically plausible but not dog-validated. Human-cell studies support a YAP–DDX17/Microprocessor link, but this cannot be assumed in canine tumours. | Whether treatment-sensitive YAP–DDX17 interaction mediates downstream pri-to-pre-miRNA processing in canine tumour cells. | Demonstrate YAP–DDX17 interaction before and after cytotoxic exposure, perturb the proposed interaction, and measure pri-/pre-/mature miRNA processing; rescue or complementary perturbation should be used where feasible. | Interaction without a downstream processing effect would not establish causal coupling; absence of interaction or mediation would argue against this candidate mechanism but would not by itself refute the broader temporal hypothesis. |
| Clinical translation | Premature as a primary claim. Longitudinal clinical sampling in canine DLBCL could eventually test whether early stress readouts and delayed miRNA patterns associate with remission duration. | Whether cellular temporal patterns are detectable in clinically accessible samples and whether they add information beyond established clinical and pathological variables. | After cellular and assay validation, perform longitudinal studies with defined sampling times, standardised treatment context, and remission or relapse endpoints. | A negative clinical association would weaken biomarker utility, even if some cellular mechanisms remain biologically valid. |
Second, mechanistic dependence should be tested by perturbation. If Hippo–YAP activity is proposed to influence the later miRNA response, perturbing this axis should alter that response after chemotherapy. This could be tested by manipulating upstream Hippo activity, YAP localisation, or YAP function, followed by paired measurement of stress markers, pri-/pre-/mature miRNA species, and cell-fate outcomes. These RNA species should be interpreted as distinct readouts: mature-miRNA abundance reflects the final measured pool, whereas changes across pri-miRNA and pre-miRNA provide more direct information about precursor processing. Where assay performance permits, relationships between pri-, pre-, and mature species may support inference about processing efficiency, but should not be treated as standalone proof of altered biogenesis. Such experiments should verify that the intended pathway modulation was achieved and, where feasible, use complementary perturbations to reduce off-target interpretation. A reproducible downstream change following verified upstream perturbation would support mechanistic dependence, but would not by itself establish that the proposed YAP–DDX17 route mediates the effect. Conversely, if the delayed miRNA response is unchanged despite verified Hippo–YAP perturbation, dependence on this axis would not be supported.
If YAP–DDX17–Microprocessor is pursued as one candidate bridge, causal coupling through this specific route would require evidence beyond temporal ordering and upstream perturbation. Human-cell studies support a mechanistic link between Hippo signalling, YAP, DDX17/p72, and Microprocessor activity (Mori et al., 2014; Wu, 2020; Xu et al., 2022; Yu & Guan, 2014), but this cannot be assumed in canine tumours. The first requirement is to demonstrate a basal YAP–DDX17 interaction in canine lymphoma or osteosarcoma cells and determine whether that interaction changes after cytotoxic treatment. The second is to show that perturbation of the proposed interaction alters pri-to-pre-miRNA processing and the downstream miRNA response. Where experimentally feasible, rescue or complementary perturbation experiments would further strengthen causal attribution. Direct YAP–DDX17 interaction without a corresponding downstream processing effect would establish physical association but would not be sufficient to demonstrate causal coupling. Accordingly, a change in mature miRNA abundance without evidence of altered pri-to-pre-miRNA processing would not support a Microprocessor-mediated explanation. Failure to validate this route would argue against YAP–DDX17–Microprocessor as the relevant mediator, but would not by itself refute the broader temporal hypothesis.
Finally, clinical translation should be treated as a downstream test, not the starting point. Longitudinal sampling in canine diffuse large B-cell lymphoma could eventually ask whether early stress readouts and later miRNA patterns are associated with duration of first remission. However, proposed liquid-biopsy markers, including exosomal YAP-related measurements, require analytical validation before they can support clinical claims. A negative clinical association would not necessarily refute every cellular aspect of the model, but it would weaken its value as a biomarker strategy.
[内容请查看原文]
Discussion: implications, boundaries, and future directions
The immediate implication of this Opinion Paper is methodological rather than therapeutic. If acquired chemoresistance is partly shaped by the timing of post-treatment responses, then studies in these canine tumour contexts should not rely only on baseline samples, relapse samples, or drug-resistant cell lines selected after prolonged exposure. These designs remain useful, but they are poorly suited to detecting early treatment-induced transitions. A stronger approach would include defined early and delayed post-treatment time points in canine cell lines, ex vivo tumour material, or carefully selected clinical studies, with paired measurements of stress signalling, miRNA expression, and cell-fate outcome.
Pharmacokinetic and treatment-context heterogeneity are therefore integral to the temporal hypothesis rather than peripheral technical considerations. A nominal 6-, 24-, or 48-hour interval may have different biological meaning after different cytotoxic agents, depending on systemic and tumour exposure, drug persistence, active metabolites, and treatment sequence. This is particularly relevant to canine DLBCL, where CHOP protocols expose tumours to multiple agents with distinct pharmacological profiles, and to osteosarcoma, where the relevant temporal response may differ between drugs used in adjuvant treatment. Experimental time courses should therefore be drug-specific and exposure-informed, and clinical sampling should be interpreted relative to the actual treatment schedule rather than by clock time alone.
Canine diffuse large B-cell lymphoma is a logical first clinical context for this work. It has a clear therapeutic structure, high initial response rates to CHOP-based protocols, and clinically meaningful variation in remission duration and relapse (Childress et al., 2018; Davies et al., 2018; Hawkes et al., 2024; Vos et al., 2019). Osteosarcoma is also relevant, particularly for studying treatment failure, invasive behaviour, and tumour-suppressive miRNAs in dog-derived models (Berg et al., 1997; Gourbault & Llobat, 2020; Lopez et al., 2018; Ludwig et al., 2024; Selvarajah & Kirpensteijn, 2010). However, the model should not be generalised across all canine cancers without direct testing. Different tumour types, drugs, genetic backgrounds, and treatment schedules may produce different temporal patterns. Accordingly, DLBCL and osteosarcoma should be evaluated independently, and concordant findings across them would strengthen the broader relevance of the framework but would not establish a universal canine cancer mechanism. Such heterogeneity is inherent to spontaneous canine cancers. Any useful temporal model must therefore remain robust to inter-individual variability, treatment context, and pre-existing resistant subclones.
The boundaries of the model are equally important. The YAP–DDX17–Microprocessor axis is only one candidate mechanism and is not a validated canine pathway. The broader temporal hypothesis does not require this specific axis to be operative. The miRNA examples used in this article should not be interpreted as a binary sensitivity–resistance axis. miR-34a, selected 14q32 miRNAs, and miR-21 are included as illustrative candidates because canine studies have linked them to tumour biology, prognosis, relapse, or aggressive behaviour (Craig et al., 2019; Elshafie et al., 2023; Elshafie et al., 2024; Gourbault & Llobat, 2020; Lopez et al., 2018; Ludwig et al., 2024), not because current evidence establishes any of them as directionally specific markers of chemotherapy sensitivity or resistance. Additional canine-specific evidence comes from mammary tumours, in which miR-18a, miR-18b, and miR-21 were upregulated in malignant tumours compared with control tissues, whereas miR-146b was downregulated in both benign and malignant tumours (Abbate et al., 2022). A subsequent study linked miR-21 upregulation with increased IL-6 and TNF-α expression and with higher proliferative activity assessed by Ki67 in canine mammary carcinomas (Abbate et al., 2023). These cross-sectional findings broaden the evidence for biologically relevant miRNA dysregulation in canine tumours but do not establish chemotherapy-induced temporal regulation or chemoresistance. Individual miRNAs can regulate multiple targets, and their functional effects may vary with tumour type, treatment, and cellular state. miRNA findings should therefore be interpreted as context-dependent regulatory patterns rather than as isolated binary switches. Similarly, extracellular vesicle or liquid-biopsy readouts may become useful in future longitudinal studies, but they should not be treated as validated biomarkers until analytical performance, biological source, and association with outcome have been established.
The tumour microenvironment is another important source of biological context that is not captured by a purely tumour-cell-intrinsic model. In canine DLBCL, transcriptional studies have identified immune and inflammatory programmes, including T-cell-, macrophage-, and immune-checkpoint-related signals, associated with clinical outcome (Aresu et al., 2019). Canine osteosarcoma also contains a heterogeneous immune microenvironment, including multiple tumour-associated macrophage populations and predicted dendritic-cell/T-cell interactions (Ammons et al., 2024), and macrophage infiltration has been associated with disease-free interval in dogs treated with amputation and adjuvant carboplatin (Withers et al., 2019). These observations support the biological relevance of the microenvironment but do not establish that it drives the proposed post-chemotherapy temporal sequence. Cytokine signalling, immune-cell composition, stromal interactions, hypoxia, and extracellular-vesicle exchange could modify either the early stress response or later miRNA regulation and may also contribute signals detected in bulk tumour samples. Future temporal studies should therefore distinguish tumour-cell-intrinsic changes from microenvironment-derived responses wherever experimentally feasible.
Cross-species time-resolved evidence provides limited precedent for this type of temporal design. In K562 chronic myeloid leukaemia cells, imatinib exposure was associated with an early Hippo-related transcript response at 12 h followed by delayed upregulation of Hippo-associated miRNAs at 48 h (Akbari-Ardabili et al., 2026). However, these findings were correlative, were obtained in a human cell-line model, and were not reproduced as a uniform Hippo–YAP transcriptomic response in primary CD34⁺ CML samples. They therefore support the feasibility of testing temporally separated Hippo- and miRNA-related responses, but not the existence of the proposed mechanism in canine tumours.
This Opinion Paper does not argue that temporal coupling is the sole explanation for chemoresistance. Resistant disease may arise through clonal selection, pre-existing heterogeneity, drug efflux, impaired apoptosis, altered DNA repair, immune and stromal microenvironmental influences, pharmacokinetic variation, or treatment heterogeneity. The proposed model is useful only if it helps organise these factors into testable temporal questions. Its near-term value is therefore not clinical prediction, but disciplined hypothesis generation: identifying when treatment response begins to fail, which regulatory transitions are involved, and which claims can be supported, revised, or rejected in canine systems.
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Ethics statement
This article did not involve new animal experiments, human participants, or privately owned animal data. Therefore, ethics approval was not required.
Declaration of generative AI and AI-assisted technologies in the manuscript preparation process
During the preparation and revision of this work, the author used ChatGPT (OpenAI) for language refinement and editorial organisation. After using this tool, the author reviewed and edited the content as needed, independently verified the scientific claims and references, and takes full responsibility for the content of the published article.
Declaration of competing interest
The author declares that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.