Association between body condition profiles, milk production and reproduction performance in Holstein and Normande cows

Body-condition dynamics are known to affect the different steps of reproduction in cattle (cyclicity, estrus expression, fertilization, embryo development). This has led to a widespread idea that there is an ideal-target optimal body condition, but no clear profile has yet been identified. Here we investigated the relationships between body condition score (BCS) profiles and reproductive performance in dairy cows. Data were from Holstein or Normande herds in 6 French experimental farms. In the Holstein breed, we discriminated 4 BCS profiles based on combining BCS at calving ( Low : around 2.6 points, or High : around 3.3 points) with BCS loss after calving ( Moderate (M) : ≤ 1.0 points, or Severe (S) : > 1.0 points). The Low-M profile mostly included multiparous cows with higher milk yield and lower reproductive performance than cows in the 3 other profiles. Low-M cows that experienced abnormal ovarian activity had lower reproductive performance than their profile-mates. Moreover, 67% of Low-M cows kept the same profile at the following lactation. The High-S profile mostly included primiparous cows with lower milk yield and higher reproductive performance than cows in other profiles. In High-S cows, higher milk yields correlated to higher risk of failure to calf on first insemination. Moreover, 38% of High-S cows kept the same profile at the following lactation, and none changed to Low-M . The other 2 BCS profiles ( Low-S and High-M ) were intermediate in terms of milk yield and reproductive performance. In Normande, we discriminated 3 BCS profiles based on combining BCS at calving ( Low : around 2.6 points, or High : around 3.5 points) with BCS loss after calving ( Flat (F) : flat with no loss, Moderate (M) : around 0.5 points, or Severe (S) : around 1.0 point). The Low-M and High-S profiles included cows with similar performance, even though High-S -profile cows showed better but not significantly different milk yield and reproduction performance. The High-F profile included cows that were more likely to experience abnormal ovarian activity and fail at first insemination than cows in other profiles. More than 50% of Normande cows with 2 successive lactations kept in the same BCS profile at the next lactation. Even though a low BCS at calving combined with severe BCS loss (more than 1 point) after calving was found to in - crease reproductive failure, there was no evidence of an optimal BCS profile for reproduction in dairy cows, and reproductive success or failure is multifactorial.


INTRODUCTION
In the early 2000s, several studies showed a worrying decline in the ability of dairy cows to reproduce: the delay to resumption of ovarian activity had increased, the proportion of animals with cyclicity abnormalities was higher, the duration and intensity of estruses had decreased, and fertility had deteriorated (Royal et al., 2000;Friggens et al., 2010).However, over the past 10 years, reproductive performance has tended to improve across the Holstein breed.In France, calving interval decreased in the Holstein breed but increased in the Normande breed, with less than a 10-d difference between the 2 breeds (Bidan et al., 2019;reproscope.fr).Intrabreed reproductive performance varies widely among animals and among farms, suggesting that reproductive performance is multifactorial at both animal level and farm level.
Many factors impact fertility (for review, see Walsh et al., 2011).The nutritional status and metabolism of females (in relation to milk yield and body condition) before or during breeding and during early gestation have substantial effects on follicular growth and oocyte maturation in the ovary (Britt, 1991;Scaramuzzi et al., 2011) and thus on subsequent embryonic development (Britt, 1992;Diskin and Morris, 2008) and uterine environment.
Several literature reviews and meta-analyses point to an ideal body condition score (BCS) profile to preserve reproductive performance (Royal et al., 2000;Chagas et al., 2007;Bedere et al., 2018).Changes in body fat reserves affect cyclicity, fertilization, and early embryonic development, and the dynamics and levels of milk production affect estrus and late embryonic development (Garnsworthy, 2006;Cutullic et al., 2012;Bedere et al., 2018).Roche et al. (2009) noted that the BCS of cows at calving, the nadir BCS, and the postpartum BCS loss are associated with differences in milk production, reproduction, and health.Carvalho et al. (2014) showed that cows that maintained or gain body condition at the beginning of lactation had better pregnancy at first time artificial insemination (AI).Recently, the concept of a high fertility cycle as emerged (Middleton et al., 2019;Fricke et al., 2023), which stipulates that "maintaining a cycle of pregnancy before 130 DIM may reduce the amount of body condition lost after the next parturition, enhance subsequent pregnancies per AI, and reduce the possibility of early pregnancy loss." A field study in 10 French commercial Holstein herds highlighted that certain trajectories of body condition dynamics (especially between 0 and 30 d or 0-60 d postpartum) and milk production profiles were associated with contrasting reproductive performances (time to breeding, ovarian cyclicity profile, risk for non-fertilization/early embryonic mortality; (Freret et al., 2005;Dubois et al., 2006;Ponsart et al., 2006).
We hypothesized that (1) cows could be discriminated according to their body condition profile, mainly defined by BCS at calving and BCS loss during early lactation; (2) differences for milk production, ovarian activity and reproduction performance could be highlighted between BCS profiles, and (3) for a given BCS profile, individual factors such as milk production, parity, calving condition, postpartum ovarian activity, days in milk at AI, BCS at calving, nadir BCS and BCS at AI, would be associated with reproductive performance.Therefore, the objective of this study was to assess the relationships between individual body condition profiles, milk yields and reproductive performances in both Holstein and Normande dairy cows to assess whether BCS profiles can reliably predict reproductive performances.

MATERIALS AND METHODS
Data concerning individual milk yield, milk composition, BCS, reproduction and calving conditions were sourced from 6 French experimental farms and covered 2 breeds (Holstein and Normande).Three of the farms used milk progesterone profiles to track postpartum ovarian activity (Cutullic et al., 2011).Only data from cows with at least one insemination and 5 BCS records (scored on a 0-5 scale; Bazin et al., 1984) from 7 d before calving to 210 DIM were included in the statistical analysis.

Description of the farm systems (breeding season and feeding system)
The 6 farms were characterized by cow-herd breed, breeding season and season length, type of feed system based on the main forage distributed (100% grass-based, pasture and maize/grass silage or no pasture), and frequency of body condition scoring (Figure 1).All farms used seasonal breeding with a fixed calendar date for the start and end of the breeding season.Number of cows and lactations and average performances per farm are given in Table 1.
Le Pin experimental farm.Data came from an experiment conducted from 2006 to 2019 at the INRAE's 'Le Pin' dairy research farm in Gouffern-en-Auge, France (Bedere et al., 2017).
Cows received either a "Low" or "High" diet.The "Low" diet (for Holstein (n = 91) and Normande cows (n = 95) managed together) was based on pasture grazing and forage if necessary supplemented with minerals only during the outdoor period and on grass silage and haylage supplemented with minerals only during the indoor period.The "High" diet (for Holstein (n = 90) and Normande cows (n = 93) managed together) was based on pasture grazing and forage if necessary supplemented with concentrate (4 kg/cow/day) and minerals during the outdoor period and on maize silage and dehydrated alfalfa supplemented with concentrate and minerals during the indoor period.All cows were inseminated during a 3-mo breeding season running from mid-March to mid-June.BCS was recorded monthly every year.
Statistical analysis here considered the Le Pin "High" and Le Pin "Low" cows as 2 different herds.
Méjusseaume experimental farm.Data came from an experiment conducted from 2014 to 2016 at the INRAE's ' 'Méjusseaume' dairy research farm (Le Rheu, France) as part of the Deffilait project (ANR-15-CE20-0014, France).The data set consisted of 115 Holstein cows housed in a freestall barn and receiving the same diet from calving to drying off (Fischer et al., 2018).The total mixed ration (TMR) was based on maize silage and concentrate with mineral supplementation.All cows were inseminated during a 4-mo breeding season running from late November to late March.BCS was recorded monthly every year.
Orcival experimental farm.Data came from an experimental conducted at the INRAE's experimental dairy research farm in Orcival (Orcival, France) between 2001 and 2006.Diet was based on pasture grazing and forage if necessary supplemented with concentrate and minerals during the outdoor period and hay and wilted grass silage supplemented with concentrate and minerals during the indoor period.In this farm, Holstein cows (n = 52) were inseminated during a 4-mo breeding season running from early January to late April.BCS was recorded monthly every year.
Trévarez experimental farm.Data came from several experiments conducted at the Chambre d'Agriculture's ' 'Trévarez' dairy research farm (Saint-Goazec, France) between 2001 and 2018.All cows were Holstein breed.Diet was based on pasture (grazed more or less intensively depending on the experiment) and forage if necessary supplemented with concentrate and minerals (varying between experiments) during the outdoor period and on maize and grass silage supplemented with concentrate and minerals (varying between experiments) during the indoor period.BCS was recorded monthly every year.
The Trévarez "Spring" cows (n = 184) were inseminated during a 4-mo breeding season running from early or late May (depending on the year) to early or late September (depending on the year).The Trévarez "Autumn" cows (n = 249) were inseminated during a 4-mo breeding season running from late November to late March.Statistical analysis considered the Trévarez "Spring" and "Autumn" cows as 2 different herds.
La Blanche Maison experimental farm.Data came from an experiment conducted at the 'La Blanche Maison' dairy research farm (Pont-Hébert, France) from 2012 to 2016.The data set consisted of 56 Normande-breed 56 cows fed a diet based on pasture grazing and forage if necessary supplemented with concentrate and minerals during the outdoor period and maize or grass silage and hay supplemented with concentrate and minerals during the indoor period.The cows were distributed over 2 4-mo breeding periods: 36 cows were inseminated from mid-June to mid-October and 20 cows were inseminated from mid-December (around the 15th) to mid-April (around the 15th).Despite 2 contrasting feeding systems and 2 breeding seasons, the limited number of cows per treatment meant that they could not be considered as different herds.BCS was recorded monthly from calving to dryoff every year.Les Trinottières experimental farm.Data came from several experiments conducted at the Chambre d'Agriculture's ' 'Les Trinottières' dairy research farm (Montreuil-sur-Loir, France) from 2012 to 2018.All cows were Holstein breed.The farm hosted trials on different levels and types of concentrate supplementation, but diets were all based on pasture grazing and maize silage supplemented with concentrate and minerals during the outdoor period and maize and grass silage supplemented with concentrate and minerals during the indoor period.All cows were inseminated during a 5-mo breeding season running from late November to late March.BCS has was recorded every 2 weeks from September to June every year.

Collection of data on milk yield, milk composition and body condition score
In total, we had 1,685 confirmed lactations from 948 Holstein cows and 482 confirmed lactations from 244 Normande cows.For each lactation, all farms provided data on 305-d milk yield, 305-d fat yield, 305-d protein yield, and milk yield at peak.Average 305-d fat content and 305-d protein content were computed by dividing 305-d fat yield or 305-d protein yield by 305-d milk yield (same units, i.e., kg) and then were expressed as %.Number of lactations and average milk yield performance per farm are described in Table 1.
A weekly BCS was obtained from calving to 210 d in milk (DIM) using an interpolation spline created from on-farm BCS records (R Core Team, 2023).To homogenize the BCS data set and minimize the influence of between-farm operator variability, weekly BCS were centered within each farm and then scaled by adding the overall mean BCS (all farms combined).Based on this weekly BCS, we computed BCS at calving, nadir BCS, BCS at first AI, BCS at conception, BCS loss between calving and nadir, and DIM at nadir BCS.

Collection of data on reproductive performance
Based on the dates of artificial insemination (AI) and calving, we determined whether cows became pregnant or not, and which AI resulted in conception, enabling us to compute reproductive indicators for each lactation, such as DIM at first AI, DIM at conception, interval between start of the breeding season and first AI, interval between start of the breeding season and conception, interval between calvings, and number of AI per cow.DIM at the start of the breeding season was also calculated.Calving conditions and uterine health events were recorded as either problem-free calving or complicated calving (dystocia, cesarean, metritis, vaginal infection, or uterine prolapse).Average reproduction performances per farm are described in Table 1.

Monitoring of postpartum ovarian activity
Postpartum progesterone profiles were determined for 721 lactations distributed among 3 of the farms (Méjusseaume, Les Trinottières, Le Pin).Morning milk samples were collected every Monday, Wednesday and Friday from calving to 5 weeks after the end of the breeding season in Le Pin, every Tuesday and Thursday from calving to 2 weeks after the second pregnancy diagnosis in Méjusseaume, and every Tuesday and Thursday from calving to 90 d postpartum in Les Trinottières.Progesterone was measured in milk using a commercially available ELISA kit (Milk Progesterone ELISA, Ridgeway Science Ltd., England).The coefficient of variation between assays on 5ng/ml control samples ranged from 7% to 15% between years and experimental farms.Based on the method developed by Cutullic et al. (2011), physiological intervals were calculated for each luteal phase.Briefly, 2 milk progesterone thresholds were defined to separate basal progesterone levels (below the first threshold) from a luteal phase (above the second threshold) and an uncertain status between the 2 thresholds.The first threshold was defined as the value below which 95% of progesterone concentrations during estrus were recorded (it ranged from 0.1 to 1.5 ng/ml in the data sets).The second threshold was defined as the first quartile of all progesterone concentrations recorded above the first threshold (it ranged from 2.0 to 6.4 ng/ml in the data sets).A luteal phase was considered to begin when at least 2 consecutive progesterone concentrations were above the first threshold and at least one was above the second threshold.Furthermore, a luteal phase was considered to end when at least 1 progesterone concentration fell below the first threshold.Finally, commencement of luteal activity (CLA, time from calving to first luteal phase onset), cycle length (time between 2 luteal phase onsets), luteal phase length and inter-luteal interval were calculated.A luteal phase exceeding 25 d was considered as a prolonged luteal phase (PLP).Ovulation was considered to be delayed if the inter-luteal interval exceeded 12 d.Based on these definitions, progesterone profiles were classified as (i) Normal, (ii) PLP (when at least one PLP was observed), (iii) Delayed (if CLA >60 d), (iv) Interrupted (when at least one ovulation of rank ≥2 was delayed) and (v) Disordered (when luteal activity appeared irregular but could not be assigned to another abnormality class).

Determination of BCS profiles and concordance of BCS profiles between 2 successive lactations
Within-breed BCS profiles were determined using principal component analysis (PCA) on 5 BCS variables (BCS at calving, BCS at 28 DIM, BCS at 56 DIM, BCS at 98 DIM, and BCS at 210 DIM) and the 4 between-BCSvariable differences between these stages.Hierarchical cluster analysis (HCA) analysis was performed on the PCA axes.The number of profiles was based on the criteria of inertia gain and number of cows per profile.These analyses were performed using the FactoMineR package (Lê et al., 2008).Distribution of cows by farms, parity (primiparous vs multiparous) and calving conditions (0 = problem-free calving and 1 = calving with health issues) between BCS profiles was tested using a Pearson's Chisquared test.We then computed the Kappa coefficient (vcd package; Meyer et al., 2006) to study the concordance of BCS profiles from one lactation to the next for cows with successive lactations.A Sankey diagram was used to represent the flow of cow BCS profiles between 2 successive lactations, using the ggplot2 package (Wickham et al., 2016), the ggsankey package (Sjoberg, 2021) and the dplyr package (Wickham et al., 2019)).

Relationships between milk production, reproduction, and BCS profiles
Relationships between milk production, reproduction, and BCS profiles were analyzed for each breed separately.Within-breed, the relationship between BCS profile and animal performance was estimated using a linear mixed model for continuous variables and a generalized linear mixed model for binary variables with a binomial distribution (lme4 package; Bates et al., 2009).The final models were: where Y ij = 305-d milk yield, peak milk yield, 305-d fat plus protein, average 305-d fat content, average 305-d protein content for lactation i; BCS j = BCS profile (j = number of the BCS profile cluster) for cow-lactation i; herd_year = random effect of the combination of herd and year (see Table 1); e ij = residual effect where V ijk = DIM at first insemination, interval between start of the breeding season and first AI, DIM at conception, interval between start of the breeding season and conception, calving interval as a continuous variable, and in-calf at first insemination as a binary variable (0 = failed to calve and 1 = calved again) for cow-lactation i; BCS j = BCS profile cluster (j = number of the BCS profile cluster); Ccondition k = calving condition (0 = problemfree calving and 1 = calving with health issues); β 1 = linear regression coefficient on DIM at start of the breeding season centered within BCS profile (DIM_BS centered ); herd_year = random effect of the combination of herd and year (see Table 1); e ijk = residual effect.Tukey's test was used to perform post hoc comparisons for continuous variables.For binary variables, odds ratio was calculated as the exponent of the associated model estimate coefficient for the variable.
For each breed separately, within each BCS profile, the relationships between some cows' characteristics and calving at first insemination were estimated.First, the following individual variables were considered: 305-d milk yield, 305-d protein content, 305-d fat content, milk yield at peak, 305-d fat plus protein, BCS at calving, days in milk at the start of the breeding season centered within BCS profile, days in milk at first insemination, BCS at first insemination, days in milk at nadir BCS, nadir BCS, calving condition (0 = problem-free calving and 1 = calving with health issues) and parity (primiparous vs multiparous).Correlation coefficients between continuous variables were estimated previously, and when the coefficient was superior to 0.85, only one variable was kept.Hence, nadir BCS, milk yield at peak, and 305-d fat plus protein were removed from the initial model, the final model being: were compared with cows without progesterone profile for the main variables (i.e., milk yield, fat and protein contents, calving rate at first insemination, overall calving rate, etc.) using a t-test for quantitative variables or a Pearson's Chi-squared tests for binary variables.The relationship between the BCS profile and log (CLA) was estimated using a linear mixed model and between the BCS profile and type of progesterone profile was estimated using a generalized linear mixed model with a binomial distribution, with the lme4 package.The final models were: where Y ijk = log (CLA) or type of progesterone profile (normal, PLP, delayed) for cow-lactation i; BCS j = BCS profile (j = number of the BCS profile cluster); herd_year = random effect of the combination of herd and year (see Table 1); Ccondition k = Ccalving condition (0 = problemfree calving and 1 = calving with health issues); Parity = primiparous vs multiparous; e ijk = residual effect.
Within each BCS profile, we performed an anova on the previous individual variables to test the effect of success at first insemination.We also used a Pearson's Chi-squared test to test the relationship between parity (primiparous or multiparous), calving conditions (0 = problem-free calving and 1 = calving with health issues), the progesterone profile (normal or not) and success at first insemination.

Description of BCS profiles and concordance of BCS profiles between 2 successive lactations
Four BCS profiles were identified for the Holstein breed and 3 for the Normande breed (Figure 2).For Holstein (Table 2), the profiles (named Low-M, Low-S, High-M, and High-S) were distinguished according to BCS at calving (Low: a BCS at calving of around 2.60, and High: a BCS at calving around 3.30) and slope of body-condition loss after calving (moderate, named 'M', for less than 1.0 points, and severe, named 'S', for more than 1.0 points between BCS at calving and nadir BCS).Table 2 reports the number of lactations in each BCS profile per farm.Distribution of cows between farms was different according to BCS profiles (P value < 0.001).The cows from Méjusseaume farm were mostly in the Low-M profile.The cows from the other farms followed a more balanced distribution across the 4 profiles, even although there very few Low-S-profile Le Pin "High" cows and very few High-M-profile Le Pin "Low" cows.The proportion of multiparous cows was higher in the Low-S (73%) and Low-M (60%) profiles than in the High-S (43%) and High-M (39%) profiles (P value < 0.001).
The concordance of a BCS profile through successive lactations is illustrated in a Sankey diagram (Figure 3).The concordance of a BCS profile between 2 successive lactations was low, with a Kappa coefficient of 0.22 (Pvalue < 0.001).Holstein cows (Figure 3a) in the Low-S profile in lactation n mostly (more than 2/3 of cows) kept the same BCS profile in lactation n+1.Around 40% of Holstein cows in the Low-M profile in lactation n had a higher BCS at calving in lactation n+1, and around 63% of Holstein cows in the High-S profile in lactation n had a lower BCS at calving in lactation n+1.Cows in the High-M profile in lactation n were evenly distributed between High-M, High-S, and Low-M profiles in lactation n+1.
For Normande (Table 2), the first profile (named High-F) was characterized by a high BCS at calving (around 3.6) and no body-condition loss, the second profile (named Low-M) was characterized by a low BCS at calving (around 2.6) and a moderate body-condition loss at the beginning of lactation, and the third profile (named High-S) was characterized by a high BCS at calving (around 3.4) and a slow and steady body condition loss after calving).Number of lactations was lower in the High-F profile than in the other High-S and Low-M profiles.The proportion of multiparous cows was a little bit higher in the High-S (66%) and Low-M (62%) profiles than in the High-F profile (55%; P value = 0.01).Although all 3 profiles encompassed cows from each farm (Table 2), distribution of cows between farms was different according to BCS profiles (P value < 0.001).The Low-M profile included mostly Le Pin "Low cows" (59% of cows in this profile) and the High-F profile included mostly Le Pin "High" cows (70% of cows in this profile).
The Kappa coefficient of concordance on a BCS profile between 2 successive lactations was 0.33 (P value < 0.001).Normande cows whose lactations were qualified as Low-M mostly stayed Low-M during the following lactation (63%; Figure 3b).Half of the Normande cows whose lactations were qualified as High-S or High-F stayed High-S or High-F during the following lactation.

Relationships between milk production, reproduction, and BCS profiles
Milk yield and milk composition.For Holstein (Table 3), Low-S-profile cows had the highest 305-d milk yield and the lowest 305-d fat and protein content whereas High-M-profile cows had the lowest 305-d milk yield and the highest 305-d fat and protein content (P value < 0.001).The Low-M and High-S BCS profiles had intermediate values.For Normande (Table 4), High-F-  0.3) and tended to be lower in profiles with higher BCS at calving for the Holstein breed (P value = 0.1).In both breeds (Table 3 and Table 4), the occurrence of such issues was negatively associated with calving rate after first insemination (P value = 0.02 for both breeds) or overall calving rate (P value = 0.01 in the Holstein breed and P value < 0.001 in the Normande breed) and with DIM at first insemination (i.e., longer calving to 1st AI interval for cows with complicated calving and/or uterine health issues; P value < 0.001 in the Holstein breed and P value = 0.002 in the Normande breed).
For Holstein (Table 3), DIM at start of the breeding season (interval between calving and the start of the breeding season centered within BCS profile) was associated with all measures of reproductive performance.Calving rate at first insemination and overall calving rate were better for cows with a longer DIM at the start of the breeding season (odds ratio >1, P value = 0.004 and P value < 0.001, respectively).Increasing DIM at start of the breeding season by 1 d reduced the interval between the start of the breeding season and first insemination by 0.5 d (P value < 0.001) and reduced the interval between start of the breeding season and conception by 0.4 d (P value < 0.001).For Normande (Table 4), overall calving rate was better for cows with a longer DIM at the start of the breeding season (odds ratio > 1, P value < 0.001).Increasing DIM at start of the breeding season by 1 d reduced the interval between start of the breeding season and first insemination by 0.5 d (P value < 0.001) and reduced the interval between start of the breeding season and conception by 0.3 d (P value < 0.001).
For Holstein (Table 3), Low-S-profile cows had the longest DIM at first insemination (+4.2 d compared with the High-S profile, P value = 0.01) and the lowest overall calving rate (−9 points compared with the Low-M profile, P value = 0.03).Low-M-profile cows had the longest DIM at conception and the longest calving interval (+6 d for both compared with High-S profile, P value = 0.05).Calving rate at first insemination did not differ between BCS profiles (P value >0.05).
For Normande (Table 4), Low-M-profile cows had the longest DIM at first insemination (+10 d compared with High-S and High-F profiles, P value < 0.001), the longest DIM at conception and the longest calving interval (+9 d for both compared with the High-S profile, P value = 0.008).High-F-profile cows had lower calving rate at first insemination (−14 points compared with High-Sprofile cows, P value = 0.05).Overall calving rate did not differ between BCS profiles (P value > 0.05).
Concerning ovarian activity (based on data from 3 farms), for Holstein (Table 3), progesterone profiles were available for 63% of cows in the Low-M profile but less than 40% of cows in the other BCS profiles.Compared with cows without a progesterone profile, cows monitored for ovarian activity had a significantly higher 305d milk yield (+945 kg, P value < 0.001) and significantly      Calving rate after first insemination and overall calving rate were significantly lower for cows with versus without a progesterone profile (34% vs 46% (P value <0.001) and 69% vs 76% (P value <0.001), respectively).There were no significant between-profile differences in calving-to-first-insemination interval, calving-to-conception internal, and calving interval (P value > 0.05).
CLA was significantly shorter in multiparous compared with primiparous Holstein cows (−2.4 d; P value < 0.001).The occurrence of complicated calving and/or uterine health issues was negatively associated with the normality of progesterone profiles: cows with complicated calving and/or uterine health issues were more likely to have an abnormal progesterone profile (OR = 1.7 [IC = 1.2-2.6],P value < 0.01).The occurrence of such issues was also associated with PLP-type progesterone profiles: cows with complicated calving and/or uterine health issues were more likely to have a PLP progesterone profile (OR = 2.1 [IC = 1.3 -3.5], P value < 0.01) compared with cows without complicated calving and/or no uterine health issues.Low-S-profile, Low-M-profile and High-S-profile cows started their luteal activity significantly later than High-M-profile cows (+8.5, +5.3 and +4.5 d on average (P value < 0.001), respectively).Low-S-profile and High-S-profile cows had significantly lower normal progesterone profiles than High-M-profile cows (−23 and −14 points (P value = 0.005), respectively).Low-Sprofile cows had higher delayed ovulation profiles than High-M-profile cows (+11 points; P value < 0.001).
For Normande (Table 4), 86% of cows had progesterone profiles, all coming from Le Pin farm.Focusing on milk yield and reproduction performance, cows monitored for ovarian activity were not different from cows without progesterone profiles (P value > 0.05).Commencement of luteal activity was significantly shorter in multiparous than primiparous Normande cows (−3.7 d; P value = 0.003).Primiparous cows had a significantly higher proportion of delayed ovulation profiles than multiparous cows (OR = 1.7 [IC = 1.2-2.5],P value = 0.001).Like for Holstein, Normande cows with complicated calving and/or uterine health issues were more likely to have an abnormal progesterone profile (OR = 2.6 [IC = 1.5-4.5],P value < 0.001) and were also more likely to have a PLP-type progesterone profile (OR = 3.9 [IC = 2.0-7.5],P value < 0.001).There were no significant betweenprofile differences in commencement of luteal activity (P value > 0.05).Proportion of normal progesterone profiles was significantly lower in the High-F profile than in the Low-M profile (−20 points, P value = 0.01).

Relationship between individual cow characteristics and calving rate after first insemination within each BCS profile
Table 5 presents the associations between individual Holstein cow characteristics and calving rate after first insemination within each of the 4 BCS profiles.In the Low-S profile, only DIM at first insemination was associated with calving rate at first insemination (P value = 0.01).On average, cows in-calf at first insemination had a shorter DIM at first insemination than cows not in-calf (−6 d (P value = 0.03), Table 6).The frequency of cows with a normal progesterone profile was higher for cows in-calf at first insemination than for cows not in-calf (59% vs 40%, respectively; P value = 0.05).In the Low-M profile, 305-d milk yield, DIM at first insemination, and DIM at nadir BCS were associated with calving rate at first insemination (P value < 0.05).On average, cows that succeeded at first insemination had shorter DIM at first insemination (−5 d, P value = 0.02), and a shorter DIM at nadir BCS (−8 d, P value = 0.05) than cows that failed (Table 6).For cows with progesterone profiles, normal ovarian activity was not associated with calving rate at first insemination (60% for cows that failed and 65% for cows that succeeded; P value > 0.05).In the High-S profile, only 305-d milk yield was associated with calving rate at first insemination (P value < 0.001).On average, cows that failed had a higher 305-d milk yield than cows that succeeded (+485 kg, P value < 0.001, Table 6).For cows with progesterone profiles, normal ovarian activity was not associated with calving rate at first insemination (54% for cows that failed and 57% for cows that succeeded; P value > 0.05).In the High-M profile, 305-d milk yield (P value = 0.02), 305-d fat content (P value = 0.003) and 305-d protein content (P value = 0.007) were associated with calving rate at first insemination.On average, cows that failed produced more 305-d milk yield (+587 kg, P value = 0.004), and tend to have less protein content (−0.04%,P value = 0.06) than cows that succeeded (Table 6).The frequency of cows with a normal progesterone profile was higher for cows that calved after first insemination than for cows that failed (82% vs 61%, respectively; P value = 0.02).
Table 7 presents the associations between individual Normande cow characteristics and calving rate at first insemination within each of the 3 BCS profiles.In the Low-M profile, none of the characteristics tested were associated with calving rate at first insemination (P value > 0.05).For cows with progesterone profiles, the frequency of cows with a normal progesterone profile was higher for cows that succeeded than for cows that failed (88% vs 73%, respectively; P value = 0.02).In the High-S profile, none of the characteristics tested were associated with calving rate at first insemination (P value DEZETTER et al.: Association between body… > 0.05).For cows with progesterone profiles, normal ovarian activity was not associated with calving rate at first insemination (67% for cows that failed and 78% for cows that succeeded; P value > 0.05).In the High-F profile, 305-d protein content (P value = 0.005) and DIM at first insemination (P value = 0.03) were associated with calving rate at first insemination.On average, cows that failed tend to produce more 305-d protein content (+0.08%;P value = 0.08) and had a shorter DIM at first insemination (−17 d; P value < 0.001) than cows that succeeded.For cows with progesterone profiles, normal ovarian activity was not associated with calving rate at first insemination (56% for cows that failed and 64% for cows that succeeded; P value > 0.05).

DISCUSSION
The 4 BCS profiles identified in Holstein-breed cows are similar to the 4 profiles identified in a previous study in 10 French commercial herds (Ponsart et al., 2006).Here, based on an PCA followed by a HCA on BCS at different DIM, we found that Holstein cows are classified in the same both categories of BCS at calving (≥3 or < 3 points) than in studies involving direct categorization (Barletta et al., 2017;Stevenson and Atanasov, 2022).As in studies involving direct categorization of BCS loss by experts (Carvalho et al., 2014;Barletta et al., 2017;Manríquez et al., 2021), we found 2 types of BCS loss after calving (severe loss or moderate loss).However, the maintaining or gaining type of BCS change was only find in the Normande breed.One hypothesis is that the proportion of Holstein cows that maintain or gain BCS after calving was too low to be discriminated by the PCA and HCA.
In both Normande and Holstein breeds, BCS profiles with a high BCS at calving and low or no BCS loss counted a high proportion of primiparous cows.Several studies have shown that primiparous cows tend to have a higher BCS and mobilize body reserves less than multiparous cows (Berry et al., 2006;Ponsart et al., 2006;Roche et al., 2007a).
In Holstein, cows in the Low-S-profile cows were mainly primiparous cows with greater 305-d milk yield and lower 305-d fat and protein contents than High-Mprofile cows.This result is consistent with the fact that multiparous cows tend to produce more milk than primiparous cows (Wood, 1967;Hansen et al., 2006), and that high milk yield is associated with greater BCS loss in early lactation (Berry et al., 2007;Roche et al., 2007b).In Normande, a high BCS at calving combined with no BCS loss post-calving was associated with higher 305-d milk fat and protein contents.Roche et al. (2007b) found that milk fat content was positively correlated with increasing BCS at calving and nadir BCS, and that milk protein content was positively associated with nadir BCS and negatively associated with BCS loss.
In both breeds, multiparous cows resumed ovarian activity earlier than primiparous cows which is consistent with many studies (Horan et al., 2005;Tanaka et al., 2008;Bedere et al., 2017).As in several studies (Opsomer et al., 2000;Petersson et al., 2006) that showed an association between postpartum endometritis or metritis and delayed ovarian cycle or prolonged luteal phase, we found that complications occurring at calving and/or uterine health issues were associated with ovarian cycles abnormalities.
Results concerning a higher probability of having abnormal ovarian cyclicity, especially a prolonged luteal phase, for High-F-profile Normande cows compared with the other 2 BCS profiles were consistent with results obtained by Bedere et al. (2017) using data from Le Pin farm.However, compared with some results found on high-yielding Holstein cows (Cutullic et al., 2012;Kafi et al., 2012), the occurrence of a prolonged luteal phase was not associated with high BCS loss during the first 60 d postpartum.In Holstein, Bedere et al. (2018) showed that there was a quadratic relationship between BCS at calving and time to resumption of cyclicity, in particular very low BCS at calving (below 2.5) delays resumption of cyclicity.Results found in Holstein cows appear to fit the left and bottom parts of the parabola (i.e., increasing BCS at calving up to 3.1 points reduces time to resume ovarian activity), while the results in Normande cows appear to fit the right part of the parabola (i.e., increasing BCS at calving to over 3.5 points increases time to resume luteal activity).In addition, the findings of the association between the Loss-S-profile and the probability of having abnormal ovarian cyclicity (mostly linked to delayed resumption of cyclicity) than cows in the other BCS profile are consistent with a previous study in 10 French commercial herds where frequency of prolonged luteal phase and delayed resumption of ovarian cyclicity were associated with a high BCS loss (≥1.5 points) during the first 60 d postpartum (Freret et al., 2005).Moreover, Santos et al. (2004) showed in Holstein cows that when BCS at 70 d postpartum is less than 2.5 points, the proportion of cows in anovulation is greater than 30%.
In the Normande breed, contrary to some findings in the Holstein breed (Middleton et al., 2019;Fricke et al., 2023), normal ovarian cyclicity and calving rate after first insemination was negatively associated with a High-F profile.One hypothesis is that because of the time-sequence of reproductive steps (Darwash et al., 1997;Gautam et al., 2010), the failure to have a normal ovarian cycle compromises the success of pregnancy at first insemination.On the other hand, High-S-profile Normande cows managed to produce the highest amount of milk while also maintaining good reproductive per-  formance.However, the milk-yield level of these cows was very low compared with Holstein cows in the abovementioned studies.
In Holstein, our results are consistent with some previous observations showing that BCS loss and higher milk yield are associated with a lower rate of conception at first insemination (Bedere et al., 2018).However, several studies (Carvalho et al., 2014;Barletta et al., 2017;Middleton et al., 2019) showed that for high-producing Holstein cows involved in synchronization programs, cows with good reproductive performance are also cows with no BCS change during the first 30 d in milk, better health and better milk production.Here, our results showed that Low-M-profile cows managed to have a relative high milk production (compared with the average milk production in our database), while also maintaining a moderate BCS loss and good reproductive performance.They also have a better chance to stay in the same BCS profile the following lactation which might be in accordance with the concept of the high-fertility cycle (Middleton et al., 2019).
Within BCS profiles, the success or failure of calving after first insemination appeared to be multifactorial, as it differed between BCS profiles and no main association emerged to explain it.In the Low-S profile, our findings supports the hypothesis that cows in-calf at first insemination have better health and hence shorter calving interval and better expression of estrus (Ismael et al., 2015;Manríquez et al., 2021).Indeed, Holstein cows in-calf at first insemination appeared to have fewer ovarian activity disorders and shorter DIM at first insemination than cows not in-calf at first insemination.Plus, even though the occurrence of complications at calving and/or uterine health issues was not significantly associated with calving rate at first insemination, the proportion of cows in-calf at first insemination that encountered complication at calving was lower than for cows not in-calf.
For Low-M profile Holstein cows, our results differ from those of Carvalho et al. (2014) as calving rate after first insemination was associated with a lower 305-d milk yield, shorter DIM at nadir BCS, and a longer DIM at first insemination.In this profile, reaching nadir BCS earlier gives better chances of conceiving.Roche et al. (2007c) found that reproductive success is associated with nadir BCS, level of BCS loss between calving and nadir, and rate of body weight gain after the start of the breeding season.For High-S-profile and High-M-profile Holstein cows, our results also differ from those of Carvalho et al. (2014) as calving rate after first insemination was associated with a lower 305-d milk yield.As this profile was also characterized by a high proportion of primiparous cows (57%), one explanation could be that the energy requirements for growth and milk production were higher for primiparous cows than for the other cows, which penalized reproductive success at first insemination (Lucy, 2001;Inchaisri et al., 2010).Indeed, 305-d milk yield and milk protein content are indicators of energy status (de Vries and Veerkamp, 2000), hence primiparous cows with a lower 305-d milk yield and a higher milk protein content might be in a better state of energy balance to achieve growth and reproduction and maintain a correct BCS profile.Furthermore, it could be helpful to study the relationships between lactation curve profiles and BCS profiles in addition to total and maximum milk production, as a previous field study in 10 commercial Holstein herds highlighted an impact of lactation curve profiles on calving-to-first-insemination interval and on late embryonic mortality rate (Dubois et al., 2006).

DEZETTER et al.: Association between body…
Table 7. Estimates and odds ratios for the relationships between cow characteristics and calving rate after first insemination within each BCS profile for the Normande breed  In the Normande breed results within BCS profiles are in line with the association between BCS profile and reproduction.Indeed, in Normande, contrary to what we observed in Holstein, cows with what could be a zero or positive energy status (i.e., cows with no mobilization of body reserves and a low yield of high-protein milk) were more likely to have complicated calving and/or uterine health issues (in the High-F profile, the proportion of complicated calving and/or uterine health issues was 24% for cows not in-calf at first insemination vs 9% of cows in-calf), to have abnormal ovarian cyclicity, and to fail at first insemination, although they were ultimately not significantly different on overall calving rate.
Finally, BCS profile may vary through successive lactations.However, cows with the lowest BCS at calving were more likely to stay in the same BCS profile.This is consistent with the results of Ponsart et al. (2006) on dairy cows and Macé et al. (2019) on ewes.When BCS at calving is low, limiting the mobilization of body reserves makes it possible to preserve reproductive performance and achieve a BCS gain at the end of lactation, which can ultimately lead to a higher BCS at the next calving.Cows with a high BCS at calving and a strong BCS loss tended to be in a BCS profile with a low BCS at the following calving, which may negatively impact their future reproductive performance, especially for primiparous cows.Low-S-profile cows tended to prioritize milk yield rather than preserving body condition, and mainly stayed in the same BCS profile between successive lactations.However, the number of successive lactations may be reduced due to culling for reproductive failure.

CONCLUSION
The objective of this study was to investigate whether there is an optimal body condition profile for reproduction in dairy cows, using a very large data set from 6 experimental farms.We identified 4 BCS profiles in Holstein-breed cows and 3 in Normande-breed cows.All the BCS profiles featured cows that managed to be in-calf under a seasonal reproductive period, showing that there might not be an optimal BCS profile.However, some BCS profiles were more at risk for ovarian activity and fertility.In the Holstein, the 'leanest' profile corresponded to the most productive cows (+800 kg of 305-d milk on average compared with cows in the profile with the best body condition).Cows in the Low-S profile also had a higher risk of postpartum cyclicity abnormalities and reproductive failures.In the Normande breed, contrary to the Holstein breed, the 'leanest' profile corresponded to the least productive cows, and these cows had good reproductive performances.On the other hand, cows in the profile without no BCS loss had an increased  risk of having abnormal ovarian cyclicity and not being in-calf at first AI.However, between-profile differences in reproductive performance remained small, which confirms that reproductive failures are multifactorial and hard to predict based on BCS profiles alone.Nevertheless, there evidence suggests there is potential to improve reproductive by controlling the BCS of Holstein cows with a low BCS at calving and a high loss of BCS post-calving and by controlling the BCS of Normande cows with a high BCS at calving and no BCS loss.Managing BCS dynamics in cows with at-risk BCS profiles at the end of lactation could be a way to evolve those cows toward a less risky BCS profile during the following lactation.

Figure 1 .
Figure 1.Schematic representation of the calving, breeding and feeding periods in the 6 experimental farms after one insemination for cow-lactation i; protein content; DIM_AI1 = DIM at first insemination; DIM_BS centered = DIM at the start of the breeding season centered within BCS profile; BCS calving = BCS at calving; BCS AI1 = BCS at first insemination; DIM_BCS nadir = DIM at nadir BCS; Ccondition j = calving condition (0 = problem-free calving and 1 = calving with health issues); Parity = primiparous vs multiparous; β 1-8 = linear regression coefficients on continuous variables; e ij = residual effect.Correlation coefficients between continuous variables were estimated previously, and when the coefficient was superior to 0.85, only one variable was kept.Hence, nadir BCS, milk yield at peak, and 305-d fat plus protein were removed from the initial model.For cows with a progesterone profile, we log-transformed CLA in days.Cows with a progesterone profile DEZETTER et al.: Association between body…

Figure 2 .
Figure 2. Average BCS profiles found in the Holstein (a) and the Normande (b) breed.

Figure 3 .
Figure 3. Concordance of BCS profile between 2 successive lactations in the Holstein breed (a) and in the Normande breed (b).

Table 3 .
Association between BCS profiles, interval between calving and start of the breeding season, calving conditions, or parity and and milk production or reproduction of Holstein cows (least squares means (SEM), frequencies, and estimated odd ratios (95% confidence intervals)) at 1st AI = days in milk at first insemination, BS_AI1 interval = interval between start of the breeding season and first insemination, DIM at conception = days in milk at conception, BS_conception interval = interval between start of the breeding season and conception, CLA = commencement of luteal activity, DIM_BS centered = days in milk at the start of the breeding season centered within BCS profile; Ccondition = calving condition (1 = complicated calving and/or uterine health issues, 0 = problem-free calving); Parity = primiparous vs multiparous.
at 1st AI = days in milk at first insemination, BS_AI1 interval = interval between start of the breeding season and first insemination, DIM at conception = days in milk at conception, BS_conception interval = interval between start of the breeding season and conception, CLA = commencement of luteal activity, DIM_BS centered = days in milk at the start of the breeding season centered within BCS profile; Ccondition = calving condition (1 = complicated calving and/or uterine health issues, 0 = problem-free calving); Parity = primiparous vs multiparous.lower 305-d fat and protein contents (−0.22% (P value < 0.001) and −0.07%(P value = 0.004), respectively).

Class
DEZETTER et al.: Association between body… Table 5.Estimates and odds ratios for the relationships between cow characteristics and calving rate after first insemination within each BCS profile for the Holstein breed Factor 1 at the start of the breeding season centered within BCS profile; DIM_AI1 = DIM at first insemination; BCS calving = BCS at calving; BCS AI1 = BCS at first insemination; DIM_BCS nadir = DIM at nadir BCS; C condition = calving condition (1 = complicated calving and/or uterine health issues, 0 = problem-free calving).NS: p value > 0.10.
DEZETTER et al.: Association between body… DEZETTER et al.: Association between body…

Table 2 .
Number of lactations per breed, BCS profile and farm; description of the BCS profiles of Holstein and Normande cows according to BCS at typical stages (mean and standard deviation) c BCS calving = BCS at calving; BCS 28DIM = BCS at 28 DIM; BCS 56DIM = BCS at 56 DIM; BCS 98DIM = BCS at 98 DIM; BCS 210DIM = BCS at 210 DIM; BCS nadir = nadir BCS; BCS AI1 = BCS at first insemination.a-d Different superscript letters signal significantly different means between BCS profiles (P < 0.05, Tukey's pairwise comparison).

Table 4 .
Association between BCS profiles, interval between calving and start of the breeding season, calving conditions, or parity and milk production or reproduction of Normande cows (least squares means (SEM), frequencies, and estimated odd ratios (95% confidence intervals))

Table 6 .
Characteristics (means ± SD) of cows in-calf or not in-calf at first insemination within each BCS profile for the Holstein breed Variable 1 1 DIM_BS centered = DIM at the start of the breeding season centered intra BCS profile; DIM_AI1 = DIM at first insemination; BCS calving = BCS at calving; BCS AI1 = BCS at first insemination; DIM_BCS nadir = DIM at nadir BCS; C condition1 = complicated calving and/or uterine health issues.

Table 8 .
Characteristics (means ± SD) of cows in-calf or not in-calf at first insemination within each BCS profile for the Normande breed Variable 1 DIM_BS centered = DIM at the start of the breeding season centered intra BCS profile; DIM_AI1 = DIM at first insemination; BCS calving = BCS at calving; BCS AI1 = BCS at first insemination; DIM_BCS nadir = DIM at nadir BCS; C condition1 = complicated calving and/or uterine health issues.