All procedures were approved by the Animal Use and Care Committee of the University of Wisconsin – Madison (A006270-R01). The calves were born at the Blaine Dairy Cattle Center, Arlington, Wisconsin. Following a 1-week adaption period, 36 female Holstein calves (40 ± 5.42 kg) were paired by birthweight (Bw) and placed on 1 of 2 treatments intended to create mammary gland growth differences that could be detected by ultrasound.

The high (H) nutritional value diet consisted of milk replacer (Cow’s Match ColdFront Protein Blend, Land O Lakes; 27% CP, 20% Fat) fed 1 gallon twice daily and ad libitum starter grain (18% CP guaranteed analysis; UW Calf Starter – Medicated Rum/Clar,Vita Plus, Lake Mills Feed and Grain Inc., Lake Mills, WI). The low (L) nutritional value diet consisted of milk replacer (Herd Maker Protein Blend, Land O Lakes; 22% CP, 15% Fat) fed 2 quarts twice daily. Dairy calves were raised in individual calf hutches and L calves were pair-fed starter grain based on consumption by their paired H calf. The starter refusal was weighed daily, and the L calves were fed the amount of starter the H calf consumed the previous day. Calves were gradually weaned from milk replacer beginning at 6 weeks and completely weaned by 7 weeks of age. At 8 weeks of age, the calves were transitioned to ad libitum grower grain (15% CP guaranteed analysis; Vita Plus, Lake Mills Feed and Grain Inc., Lake Mills, WI). At 12 weeks of age, the heifers were moved to the Marshfield Agricultural Research Station, Stratford, WI, and transitioned from grower grain to standard total mixed ration (TMR). The animals were first transitioned to a light TMR diet for 4 to 8 weeks (45% haylage, 27.7% ground shell corn, 16.6% corn silage, 8.8% soybean meal, and 1.3% vitamins and minerals). They were then transitioned to a medium TMR diet until 12 months of age (49% haylage, 42.7% corn silage, 4.0% soybean meal, 2.8% whey, 0.8% vitamins and minerals, and 0.4% urea). At 12 months of age the animals are transitioned to the farm’s breeding diet (44.1% haylage, 32.4% corn silage, 18.1% urea 1.3% soybean meal, and 3% vitamin and minerals).

Mammary gland biopsies were performed on heifers at 10, 26, 39, and 52 weeks of age. Ultrasound (Mindray Z5 Ultrasound, Mindray 65C15EA 6.5 MHz Micro-Convex Ultrasound Transducer) was used to determine the location of the biopsy. A scalpel was used to create a 1 to 2 cm slit through the skin and the capsule was dissected to allow for tissue cores to be taken with Integra Miltex disposable biopsy punches. The 2 mm punch was used at 10 weeks and the 6 mm punch was used at 26, 39, and 52 weeks of age. The tissue size varied based on age and animal. At 10 weeks of age, the tissue collected was approximately 2 mm by 5 mm. At 26 weeks of age, the tissue cores were approximately 6 mm by 1 to 2 cm, and beyond 26 weeks of age, the tissue was approximately 6 mm by 2 to 3 cm. Tissue was rinsed with saline and fixed for 24 h in 10% buffered formalin and then transferred to 70% ethanol. The tissue samples were then sent to the University of Wisconsin – Madison Veterinary School of Medicine to be embedded, sectioned, and stained with hematoxylin and eosin.

Coccygeal blood samples were collected at each biopsy as well as weekly from 8 to 14 months of age to establish cyclicity. Progesterone concentrations were determined by radioimmunoassay (MP Biomedicals).

Hematoxylin and eosin-stained sections were imaged at 10 × magnification (Basler Ace 5.0 MP, Zeiss Axio Vert A1) and annotated using QuPath, an open-source software package for digital pathology image analysis [3]. In total, 72,201 objects were annotated from 132 whole-mount images. Using QuPath, the following features were annotated and calculated from whole-mount images of the histology sections: percentage of ductal tissue in the whole-mount image, average ductal area, and average maximal ductal diameter, percentage of adipose tissue in the whole-mount image (Fig. 1).

Annotation of an ultrasound image and histology image. Image A demonstrates annotation an area of parenchyma (pink), an area of fat pad (orange), and average echogenicity, which was averaged between 10 circles. Image B demonstrates annotation of the histology images in which adipose tissue is annotated in red and ductal structures are annotated in purple. The white bar in the ultrasound images measures 1 cm and the black bar in the histology images measures 500 μm

Ultrasound videos were collected prior to biopsies and frames were extracted using ffmpeg [39]. One frame from each animal at each timepoint was chosen for analysis by QuPath. Average area of parenchyma tissue and fat pad as well as average echogenicity were calculated at 10 weeks of age (Fig. 1, Table 3). Because tissue boundaries are less clear after 10 weeks average echogenicity was used to quantify parenchymal tissue density in ultrasound images at 26, 39, and 52 weeks of age. Adipose tissue appears hyperechoic or brighter, compared to parenchymal tissue which appears more hypoechoic or darker (Fig. 1). Average echogenicity was calculated by averaging the echogenicity of 10 circles placed on the image as depicted in Fig. 1.

A linear model (R version 4.2.1, stats package version 4.2.1) including birthweight as a covariate was used to evaluate the effect of diet on area, circularity, solidity, perimeter and maximum and minimum diameter of parenchyma as well as on area, circularity, solidity, maximum and minimum diameter, mean echogenicity, echogenicity standard deviation, minimum and maximum, and perimeter of the fat pad at 10 weeks of age. Circularity measures the roundness of an object and solidity gives a measurement of the compactness of an object. A circularity value of 1 indicates a perfect circle and a perfectly convex shape has a solidity of 1. A linear mixed model (lme4 package version 1.1–30) was used to analyze weight, average duct area, average duct max diameter, total duct area / whole tissue, total adipose area / whole tissue, average echogenicity, echogenicity standard deviation, minimum and maximum echogenicity at all timepoints, which can be described as follows:

$$_=\upmu +_+_+_+ _}_+_+}_$$

where yijkl represents the response variable of interest, µ is the model intercept, Di is the fixed effect of the ith diet (high and low), Wj is the fixed effect of the jth week (10, 26, 39 and 52), Bw is the effect of the birthweight as a covariate, Al is the random effect of animal, and eijkl is the independent identically distributed normal error. For all models, residual analysis was performed to verify the model assumptions of normality and homogeneous variances. Pearson correlations (corrr version 0.4.4) was used to assess the relationship between average daily gain and histological and ultrasound variables at each week (e.g., 10, 26, 39 and 52).