高产奶牛在泌乳早期遇到代谢挑战。通常,β-羟基丁酸酯(BHB),利用在特定时间点测量的基于预定阈值来诊断奶牛的代谢状态。然而,在泌乳早期,BHB是高度动态的,并且在其时间分布中存在高度的个体间变异性。这可能会限制单一测量和基于阈值的诊断的有效性,可能会导致将代谢状态与生产和生殖结果联系起来的报告中的差异。这项研究深入研究了BHB的轨迹检查,以揭示牛间的变异并确定潜在的代谢组。我们从2项观察性研究中收集了一组数据,这些研究涉及来自多胎荷斯坦弗里斯奶牛的195次泌乳。数据集包括BHB的测量,NEFA,和在第3、6、9和21天收集的血液样本中的胰岛素在牛奶中(DIM),以及每周测定乳脂肪中的乳成分和脂肪酸(FA)比例。在这两个实验中,在泌乳的第一个月,每天记录产奶量(MY)和采食量。我们使用血液BHB的轨迹探索了个体间和个体内代谢反应的变化,并基于这种变化评估了不同代谢组的存在。为此,我们采用了生长混合模型(GMM),一种轨迹聚类技术。我们的发现揭示了奶牛之间不同代谢反应的新见解,包括轨迹模式和血液BHB浓度的大小。具体来说,我们确定了3个潜在的代谢组:“QuiBHB”簇(≈10%)显示出比其他簇更高的初始BHB浓度,在d9达到峰值(平均最大BHB为2.4mM),然后在d21下降;\“SloBHB\”簇(≈23%)开始于较低的BHB浓度,逐渐增加至第9天,并在第21天达到最高BHB浓度(实验期结束时血清BHB为1.6mM);而\“LoBHB\”簇(≈67%)以最低血清BHB浓度(血清BHB<0.75mM)开始,在整个采样期间保持相对稳定。值得注意的是,3个代谢组表现出显著的生理差异,在血液中NEFA和胰岛素浓度明显。与LoBHB奶牛相比,QuiBHB和SloBHB奶牛表现出更高的NEFA和更低的胰岛素浓度。有趣的是,这些代谢差异在泌乳的第一个月扩展到MY和DMI.与SloBHB和LoBHB母牛相比,在QuiBHB母牛中观察到的BHB浓度升高与较低的MDI和MY相关。因此,这些动物被认为是代谢受损。相反,SloBHB奶牛显示出较高的MY以及增加的MI,因此,升高的BHB可能表明这些奶牛的适应性反应。QuiBHB奶牛还显示出较高比例的不饱和脂肪酸(UFA),单不饱和脂肪酸(MUFA),泌乳第一周期间牛奶中的总C18:1FA。使用这些FA和测试日变量对QuiBHB奶牛的预测导致中等预测准确性(ROCAUC>0.7)。鉴于开发预测模型的样本量有限,以及同一周样本之间的DIM变化,结果表明模型的预测潜力和模型优化的空间。总之,根据泌乳早期血液BHB的轨迹,可以确定不同的奶牛代谢组。
High-yielding dairy cows encounter metabolic challenges in early lactation. Typically, β-hydroxybutyrate (BHB), measured at a specific time point is employed to diagnose the metabolic status of cows based on a predetermined threshold. However, in early lactation, BHB is highly dynamic, and there is high interindividual variability in its time profile. This could limit the effectiveness of the single measurement and threshold-based diagnosis probably contributing to the disparities in reports linking metabolic status with productive and reproductive outcomes. This research delves into the examination of the trajectories of BHB to unveil inter-cow variations and identify latent metabolic groups. We compiled a data set from 2 observational studies involving a total of 195 lactations from multiparous Holstein Friesian cows. The data set encompasses measurements of BHB, NEFA, and insulin from blood samples collected at 3, 6, 9, and 21 d in milk (DIM), along with weekly determinations of milk composition and fatty acids (FA) proportions in milk fat. In both experiments, milk yield (MY) and feed intake were recorded daily during the first month of lactation. We explored interindividual and intraindividual variations in metabolic responses using the trajectories of blood BHB and evaluated the presence of distinct metabolic groups based on such variations. For this purpose, we employed the growth mixture model (GMM), a trajectory clustering technique. Our findings unveil novel insights into the diverse metabolic responses among cows, encompassing both trajectory patterns and the magnitude of blood BHB concentrations. Specifically, we identified 3 latent metabolic groups: the \"QuiBHB\" cluster (≈10%) exhibited a higher initial BHB concentration than other clusters, peaking on d 9 (average maximum BHB of 2.4 mM) and then declining by d 21; the \"SloBHB\" cluster (≈23%) started with a lower BHB concentration, gradually increasing until d 9, and at the highest BHB concentration at d 21 (1.6 mM serum BHB at the end of the experimental period); and the \"LoBHB\" cluster (≈67%) began with the lowest serum BHB concentration (serum BHB <0.75 mM), remaining relatively stable throughout the sampling period. Notably, the 3 metabolic groups exhibited significant physiological disparities, evident in blood NEFA and insulin concentrations. The QuiBHB and SloBHB cows exhibited higher NEFA and lower insulin concentrations as compared with the LoBHB cows. Interestingly, these metabolic differences extended to MY and DMI during the first month of lactation. The elevated BHB concentrations observed in QuiBHB cows were linked with lower DMI and MY as compared with SloBHB and LoBHB cows. Accordingly, these animals were considered metabolically impaired. Conversely, SloBHB cows displayed higher MY along with increased DMI, and thus the elevated BHB might be indicative of an adaptive response for these cows. The QuiBHB cows also displayed higher proportions of unsaturated FA (UFA), monounsaturated FA (MUFA), and total C18:1 FA in milk during the first week of lactation. Prediction of the QuiBHB cows using these FA and test day variables resulted in moderate predictive accuracy (ROCAUC > 0.7). Given the limited sample size for the development of prediction models, and the variation in DIM among samples in the same week, the result is indicative of the predictive potential of the model and room for model optimization. In summary, distinct metabolic groups of cows could be identified based on the trajectories of blood BHB in early lactation.