Animl (2012), 6:3, pp 476 487 & The Animl Consortium 2011 doi:10.1017/s175173111100173x niml Towrds etter understnding of the respective effects of milk yield nd ody condition dynmics on reproduction in Holstein diry cows E. Cutullic 1,2,3, L. Dely 1,2, Y. Gllrd 4 nd C. Disenhus 1,2-1 INRA, UMR1080 Diry Production, 35590 Sint-Gilles, Frnce; 2 Agrocmpus-Ouest, UMR1080 Diry Production, 35000 Rennes, Frnce; 3 Bern University of Applied Sciences, Swiss College of Agriculture, 3052 Zollikofen, Switzerlnd; 4 INRA, UE326 Domine Expérimentl du Pin-u-Hrs, 61310 Exmes, Frnce (Received 22 July 2010; Accepted 1 August 2011; First pulished online 29 Septemer 2011) The overll reproductive performnce hs decresed over the lst decdes, involving chnges in cyclicity, oestrous ehviour nd fertility. High milk yield (MY), low ody condition score (BCS) nd lrge ody condition (BC) loss hve een identified s risk fctors. However, these effects re often confounded, s high MY nd ody lipid reserve moiliztion re correlted. The im of this study ws to evlute the respective effects of MY nd BC on post-prtum ovrin cyclicity, oestrus nd fertility of Holstein cows. This study provides novel informtion, s MY nd BC chnge were uncoupled in the overll dtset tht included 98 lcttions nd milk progesterone profiles. Cows were ssigned to two feeding-level groups: high feed, which chieved high MY nd moderte BC loss throughout lcttion (8410 kg, 21.17 unit from clving to ndir BCS), nd low feed, which limited MY nd triggered lrge BC loss (5719 kg, 21.54 unit). MY nd BC hd different effects t different stges of the reproductive process. Cyclicity s well s non-fertiliztion nd erly emryo mortlity were minly driven y ody lipid reserves, wheres oestrous ehviour nd lte emryo mortlity were relted to MY. The results point to possile uncoupling etween cyclicity, oestrus nd erly nd lte emryo survivl llowing compenstion long the reproductive process nd leding to similr finl reproductive performnce. In compct clving systems, which require high pregnncy rtes within short period, higher MY strtegies pper unsuitle even where BCS is mintined, owing to depressed oestrous ehviour nd proly incresed lte emryo mortlity, which delys rereeding. Similrly, strtegies tht compromise cyclicity nd fertility y excessively low BCS re unsuitle. Keywords: diry cows, reproduction, progesterone, milk yield, ody condition Implictions The ojective of this study ws to improve our understnding of the respective effects of milk yield nd ody condition on post-prtum ovrin ctivity, oestrus nd fertility in Holstein diry cows, wheres the originlity of the dtset lies in n uncoupling of these two highly correlted fctors. Cyclicity nderlyemryosurvivlpperedtoeminlydriveny ody lipid reserves, wheres oestrus nd lte emryo survivl ppered to e minly driven y milk production. These different reltionships should e tken into ccount y physiologists to improve iologicl knowledge of diry cows reproduction nd y zootechnicins to optimize reeding strtegies. Introduction Reproductive performnce hs decresed over the lst 5 decdes in diry cows s milk production incresed - E-mil: ctherine.disenhus@grocmpus-ouest.fr (Rodriguez-Mrtinez et l., 2008). All the stges of the reproductive process hve een ffected, with incresed ovrin normlities (Royl et l., 2000), depressed oestrus durtion nd intensity (Vn Eerdenurg et l., 1996; Kerrt nd Disenhus, 2004), depressed AI success (Brt et l., 2005) nd n increse in oth erly nd lte emryonic deth (Diskin et l., 2006; Humlot et l., 2009). Adverse effects of high milk yield (MY) were oserved in mny studies. High MY hs een reported to dely first ovultion (Petersson et l., 2006), to increse the risk of prolonged lutel phses (PLP; Opsomer et l., 2000), to reduce stnding oestrus durtion (Lopez et l., 2004) nd to depress fertility (Mckey et l., 2007) y incresing lte emryo mortlity (Grimrd et l., 2006). Some of these effects my e ttriuted to genetics, s genetic correltions etween MY nd severl reproductive trits hve een estlished (Boichrd et l., 2002; Royl et l., 2002). More generlly, ody condition score (BCS) nd ody condition (BC) loss resulting from the negtive energy lnce fter 476
Milk nd ody condition effects on reproduction clving re thought to explin these dverse effects of high MY (Cutullic, 2010; Friggens et l., 2010). However, in most studies, MY nd BC loss effects re confounded. Comprison of high-fed nimls (high milk, moderte BC loss) with low-fed nimls (low milk, lrge BC loss) cn help to distinguish these effects, nd Cutullic et l. (2011), using this design, oserved tht lmost ll the stges of the reproductive process ppered to e sensitive to feeding level in Holstein cows. Working on the sme Holstein dtset, this study ims to quntify the respective effects of MY nd BCS t ech stge of the reproductive process. Mteril nd methods Experimentl design nd dtset From 2006 to 2008, reed 3 feeding system experiment ws conducted on the INRA experimentl frm in Le Pin-u- Hrs (Normndy, 48.448N, 0.098E). The results per reed nd feeding system re reported in nother pper (Cutullic et l., 2011). This tril is limited to Holstein diry cows. Ech yer, 34 Holstein cows were eqully distriuted within two winter compct clving systems. At the eginning of the experiment, systems were lnced for prity, clving dte, genetic indices for MY, ft nd protein content, BCS, BW nd previous rnk of successful insemintion. For ech trit, the estimted genetic indices were evluted using Best Liner Unised Prediction niml model especilly developed for this experiment y Lrroque nd Boichrd (personl communiction). The model included the sire nd grndsire s genetic evlution, the dm s performnce over three lcttions, the clssicl fixed environmentl effects nd the feeding tretment. Cows remined in their feeding group until culling (not pregnnt, helth prolem or deth). Replcement primiprous cows were lnced for genetic indices, BCS, BW nd rnk of successful insemintion. The yerly clendr experimentl design is reported in Figure 1. The two feeding systems were grss-sed: high feed system (llowing high MY for grss-sed system nd limiting BC loss) nd low feed system (without ny concentrte, limiting MY nd triggering lrge BC loss). The cows were milked twice dily t 0630 nd 1600 h. Individul MYs were recorded y flow meters (Mettron, Westfli, Germny). Both morning nd evening milk smples were collected nd nlysed for ft nd protein content three times week using n infrred nlyser (MilkoScn TM, Foss Electric, Hillerd, Denmrk). From clving to 5 weeks fter the end of the reeding period, morning milk smples were dditionlly collected on Mondy, Wednesdy nd Fridy, stored t 2188C without preservtive nd ssyed for progesterone (P 4 ) using commercil ELISA kits (yers 1 to 2: Ovucheck Milk, Biovet, Cnd; yer 3: Milk Progesterone ELISA, Ridgewy Science Ltd, Englnd). From P 4 profiles, commencement of lutel ctivity (C-LA), interovultory intervls (ioi, estimte of cycle length), lutel phse lengths (lut) nd interlutel intervls (ili) were determined (for detils, see Cutullic et l., 2011). The first ovultion ws considered to e delyed if C-LA.50 dys. The lutel phse following the n-th ovultion ws considered to e PLP if lut n. 25 dys. The n-th ovultion (n > 2) ws considered to e delyed if ili n21. 12 dys. The cycle following the n-th ovultion ws considered to e short cycle if ioi n <15 dys. Post-prtum (pp) P 4 profiles were clssified s norml, PLP profile (if t lest one PLP ws oserved), delyed (if C-LA.50 dys), interruption (if t lest one ovultion of rnk >2 ws delyed) or disordered (if lutel ctivity ppered irregulr ut without ny of the previously indicted normlities). Cudl vein lood smples were tken t 20 nd 60 dys in milk nd t first insemintion to ssy non-esterified ftty cids nd glucose plsm concentrtions, using enzymtic colorimetry (Kone Instruments Corportion, Espoo, Finlnd). BCS ws evluted monthly, lwys y the sme two experienced ssessors. It ws scled from 0 to 5, from extremely thin to extremely ft, in increments of 0.25 (Bzin et l., 1984). During the entire reeding seson nd the 4 previous weeks, experienced stff oserved the cows t lest 5 times dy nd ehviours were noted down on stndrdized crd. An oestrus detection id ws used in 2007 nd 2008 (Kmr Het Mount Detector, KAMAR Inc., Stemot Springs, CO, USA). For nlyses, oestri detected y stnding to e mounted (specific ehviour) were distinguished from oestri detected y proceptive nd generl signs (mounting, sniffing, licking, chin-resting, restless, cler mucous dischrge, het mount detector only). This informtion ws linked with P 4 profiles: (i) to clssify oestri s true if corresponding to n ovultion period or flse nd (ii) to determine the ovultion detection rte. For ech undetected ovultion, tht is, not mtching with detected oestrus, n expected dte of oestrus ws Clving period Breeding period 13 weeks J F M A M J J A S O N D Stock feeding Psture feeding (turn out 64 ± 31 dys in milk) High MS:A:C 55:15:30 rottionl grzing (totl re 0.35 h / cow) + 4 kg C GS + 5 kg MS + 4 kg C / GS* Low GS:HL 50:50 rottionl grzing (totl re 0.60 h / cow) + 0 kg C GS / GS* Figure 1 Yerly clendr experimentl design of the feeding system experiment conducted in the INRA experimentl frm Le Pin-u-Hrs, compring high feeding level system with low feeding level system. MS, A, C, GS, HL: mize silge, dehydrted lflf pellets, concentrte, grss silge, hylge. Slt nd minerl were included in the concentrte of high-fed cows nd were dditionlly fed to low-fed cows. GS*: during the dry period, cows were fed d liitum with grss silge only. Htched period: if the grss offered ws insufficient, cows received dditionl MS (high) or GS (low) feeding (up to 8 kg/dy), mostly in the high group, which hd 1.7 smller mximl grzing re. 477
Cutullic, Dely, Gllrd nd Disenhus clculted. For ech ovultion, we considered tht there ws nother cow in oestrus t the sme time if there ws t lest one other cow in true oestrus on the sme or the previous dy. During the reeding period, rtificil insemintion (AI) ws systemticlly used on spontneous oestri fter 42 dys in milk. By comining progesterone ssys nd ultrsonogrphy t 40 nd 70 dys fter insemintion, AI results were clssified into four ctegories ccording to Humlot (2001): non-fertiliztion or erly emryo mortlity (low P 4 efore 25 dys fter AI); lte emryo mortlity (high P 4 t 25 dys, low efore 50 dys); foetl deth or ortion (ultrsonogrphy positive twice or P 4 high more thn 50 dys nd ultrsonogrphy positive once); nd reclving. In sttisticl nlyses, the emryo survivl ctegory included the two ctegories reclving nd foetl deth or ortion following insemintion. All the mjor helth prolems were recorded throughout lcttion with focus on reproduction prolems (clving difficulties, cesren, retined plcent, metritis, etc.). In the cse of cesren, retined plcent or norml involution, prostglndin nlogue injections (cloprostenol, Intervet-Schering Plough, Beucouzé, Frnce) were used. Severe metritis (purulent nd foul-smelling vginl dischrge) ws treted with n intruterine infusion (Intervet- Schering Plough, Beucouzé, Frnce) nd cloprostenol injection. Anoestrous cows received no specific tretment. Sttisticl nlyses MY nd BCS reltionship. For 98 lcttions, the effects of feeding level on MY nd BCS vriles were investigted using vrince covrince nlysis. The models included the effects of lcttion numer (ctegorized s first, second, third or more), feeding level, yer, first-order interctions with feeding level nd MY index s covrite for MY vriles (Model MY-1). The reltionship etween BC chnge from clving to 120 dys in milk (DBC 120 ) nd pek MY ws investigted using vrince covrince nlysis. A first model included the fctors lcttion numer (first, second, third or more), yer, BCS t clving, feeding level nd pek MY, nd lso the interction etween these two vriles (Model MY-2). A second model did not include feeding level nd its interction with pek MY (Model MY-3). Vriles influencing reproductive outcome. The effects of independent vriles on reproductive outcome were ssessed using forwrd stepwise regression pproch. Unless stted otherwise, only vriles tht significntly improved the model were included (P, 0.05). At ech step, chnges in the P-vlues of type III ANOVA tles were checked to detect confounding effects nd we checked in the finl models tht significnt vriles remined significnt regrdless of which sttisticl unit of the popultion ws removed. The investigted outcome vriles relting to cyclicity, oestrus nd fertility stges re listed in Tle 1. They were nlysed in chronologicl mnner in Model CYC-1-2-3, Model OES-1-2 nd Model FER-1-2. Continuous output vriles were nlysed with liner models nd ctegoricl output vriles with logistic regressions, either inomil (two output ctegories) or multinomil (three output ctegories). Overll, three kinds of independent vriles were investigted in these models: generl vriles nd lcttion-level vriles for ll models, nd event-level Tle 1 Investigted reproductive outcome vriles, their relted popultion nd rw chrcteristics (Models CYC, OES, FER) Outcome vriles Model n Popultion Men 6 s.d. or ctegories % Cyclicity Log e (C-LA) CYC-1 98 All lcttions 3.53 6 0.55 (i.e. 34.1 dys) PLP profile CYC-2 98 All lcttions PLP (23%) v. not Cycle type CYC-3 189 Ovultions of rnks 1 to 4 not followed y Norml (72%) v. PLP (14%) v. short (14%) n AI, <135 dys in milk, not elonging to disordered profile nd not followed y delyed ovultion 1 Oestrus Ovultion detection OES-1-1is 255 Ovultions of rnks ove 2 2 Detected (69%) v. not Ovultion detection y OES-2 241 Ovultions of rnks ove 2 nd with Detected (44%) v. not stnding oestrus known ehviour 2,3 Fertility Emryo survivl FER-1 128 First nd second AIs performed on true oestri 4 Non-fert. erly emr. m. (39%) v. lte emryo mortlity (20%) v. emryo survivl (41%) Reclving FER-2-2is 128 First nd second AIs performed on true oestri 4 Reclving (37%) v. not C-LA 5 commencement of lutel ctivity; PLP 5 prolonged lutel phse; AI 5 rtificil insemintion; non-fert. erly emr. m. 5 non-fertiliztion or erly emryo mortlity. 1 Ovrin ctivity interruption nd disordered profiles were too scrce (n 5 3 nd 1) to e specificlly nlysed. 2 First post-prtum ovultions re known to e hrdly detectle (17% detected v. 69% for higher rnks in this study, P, 0.001) nd were not included in ovultion detection nlyses. 3 14 detected ovultions with unknown ehviour were excluded. 4 17 insemintions performed on flse oestri were excluded. 478
Milk nd ody condition effects on reproduction vriles only for specific models (Tle 2). Log e (C-LA) ws lso tested for the occurrence of PLP profiles (Model CYC-2), s erly resumption of lutel ctivity is often ssocited with PLP (Opsomer et l., 2000). For ovultion detection models (Model OES-1-2), oservtion period (efore or during the reeding period nd the 4 previous weeks) ws included in the model to ccount for possile oservtion ises. For insemintion result models (Model FER-1-2), prity nd clving AI intervl (, or >50 dys fter clving) were lso included in the model considering their known influence on fertility. Owing to the first results, we dditionlly tested the effects of the prmeters of fitted Wood lcttion curves over the first 14 weeks of lcttion ðmy ¼ week exp cweek Þ on insemintion results (Model FER-1). Feedinglevel effect ws not directly included in these models (CYC, OES, FER), s its effect is strongly correlted with the effects of BCS nd MY vriles. In ddition, for norml, PLP nd delyed profiles, MY, BCS nd BC chnge curves were compred using vrince nlysis; models included prity nd profile ctegory (Model CYCBCS). Linking cyclicity, oestrus nd fertility. In order to test the potentil dditionl effects of cyclicity on oestrus nd of cyclicity or oestrus on fertility, the following vriles were investigted y including them in the two finl models predicting ovultion detection (Model OES-1) nd insemintion results (Model FER-2): log e (C-LA), norml v. norml P 4 profile, norml v. PLP v. other P 4 profile, previous cycle length <15 v..15 dys, previous lutel phse <25 v..25 dys, current ili <12 v..12 dys, stnding oestrus v. not. The new finl models were lelled Model OES-1is nd Model FER-2is, respectively (Tle 1). Tle 2 Independent vriles included in sttisticl nlyses to study their effectiveness to predict reproductive outcome vriles (Models CYC, OES, FER) Generl vriles investigted in Models CYC, OES nd FER Prity Primiprous/multiprous Yer 2006/2007/2008 Genitl helth prolem t/following clving (severe metritis, plcent retention, vginl infection) Yes/no Lcttion-level vriles investigted in Models CYC, OES nd FER Genetic MY index, MP content index, MF content index MYindex y MY MY (or FCMY) over weeks 1 to 3, 1 to 6, 1 to 14, 4 to 6, 7 to 14, t pek MY 3 y MY 4to6 y MY mx Milk content MP content over weeks 1 to 3, 1 to 6, t ndir MP 3 y MP min MP/MF rtio over weeks 1 to 3, 1 to 6, t ndir MP/MF 3 y MP/MF min BCS/BC chnge (DBC) BCS 30 dys efore clving, t clving, 30, 60, 90, 120 dys fter clving, t ndir BCS 230, BCS 0 y BCS min BC chnge from clving to 230, 30, 60, 90, 120 dys, or to BCS min (cn e BCS 0 ) DBC 230, DBC 30 y DBC mx BC chnge from 30 to 90, 60 to 120 dys fter clving DBC 30 to 90, DBC 90 to 120 Metolic sttus NEFA nd glucose plsm concentrtions t 20 nd 60 d fter clving NEFA 20, NEFA 60 y Event-level vriles investigted in Models CYC-3, OES nd FER only Dys fter clving Psture ccess Yes/no (indoors only) Ovultion rnk First/second or lter Oservtion during the reeding period or the 4 previous weeks Yes (during)/no (efore) MY MY (or FCMY) over week(s) 22 to0,21 to11, 0, 0 to 12 reltive to event MY week 22 to12 y MY dynmic MY (or FCMY) slope weeks 22 to0,22 to12, 21 to11, 0 to 12 MYslope week 22 to12 y Milk content MP content over weeks 22 to0,21 to11, 0 to 12 MPC week 22 to12 y MP/MF rtio over weeks 22 to0,21 to11, 0 to 12 MP/MF week 22 to12 y BCS/BC chnge BCS t event (liner interpoltion) BCS event BC chnge from clving to event DBC event BC dynmic BC chnge per 30 dys from 230 1 to 130, 230 1 to 0, 215 1 to 115, 0 to 130 reltive event (dy 0) BCslope event 230 to 130 y Metolic sttus NEFA nd glucose plsm concentrtions t first nd second insemintions NEFA AI, glucose AI MY 5 milk yield (kg/dy); MP 5 milk protein content (g/kg); MF 5 milk ft content (g/kg); FCMY 5 ft corrected milk yield (kg/dy, milk ft content 5 40 g/kg); BCS 5 ody condition score (0 to 5 scle); BC 5 ody condition; NEFA 5 non-esterified ftty cids. 1 Left censored t clving to void smoothing of the BC loss tht occurs just fter clving: for 2 to 1 cse, the vlue for n event t time t is [BCS t t1 BCS t mx(0, t2)]/[t1 mx(0, t2)]330 to hve chnge expressed s BCS unit per 30 dys. 479
Cutullic, Dely, Gllrd nd Disenhus Sttisticl softwre nd outputs. Anlyses were performed using the R sttisticl softwre (R Development Core Tem, 2009; functions lm, glm nd multinom for liner models, inomil logistic regressions nd multinomil logistic regressions, respectively). When predicting continuous output vrile, liner model results re expressed s lest-squred mens for ctegoricl independent vriles or s slopes for continuous independent vriles. When predicting ctegoricl output vriles with either two or three ctegories, the results re expressed s odds rtios for ctegoricl independent vriles (i.e. exp(coefficient of the liner predictor)) or s trnsformed odds rtios for continuous independent vriles. The trnsformed odds rtios reflect the effect of 1 s.d. increse in the independent vrile insted of the effect of 1 unit increse (i.e. exp(coefficient of the liner predictor) s.d., with s.d. eing clculted on the smple popultion, insted of exp(coefficient of the liner predictor) 1 ). This trnsformtion simplifies effect comprisons, s, for exmple, 1 unit increse in pek MY is not comprle with 1 unit increse in BCS t clving. Odds rtios interprettion remins the sme: n odds rtio.1 indictes tht n increse in the independent vrile increses the risk of the occurrence of specific ctegory rther thn the occurrence of the reference ctegory, wheres n odds rtio,1 indictes tht n increse in the independent vrile decreses this risk. For exmple, in inomil logistic regression predicting the occurrence of either ctegory B or reference ctegory A, n odds rtio.1 indictes n increse in the risk of occurrence of ctegory B rther thn ctegory A. In multinomil logistic regression predicting the occurrence of ctegory C, ctegory B or reference ctegory A, n odds rtio (C v.a).1 indictes n increse in the risk of occurrence of ctegory C rther thn ctegory A nd n odds rtio (B v.a).1 indictes n increse in the risk of occurrence of ctegory B rther thn ctegory A. In these multinomil models, the reference ctegory is thus importnt. To improve the understnding of the results of multinomil logistic regressions, the proilities of ech ctegory were dditionlly clculted nd plotted long the independent vrile, so tht pro(a) 1 pro(b) 1 pro(c) 5 1tech given vlue of the independent vrile. Results nd discussion The link etween MY nd BC dynmic ws disrupted in the entire dtset Cows in the low feed lost more BC, fster nd for longer period thn those in high feed lthough they produced much less milk (Tle 3; MY nd BCS curves re reported y Cutullic et l., 2011). The negtive reltionship etween MY nd BC chnge ws oserved within high feed (Figure 2). In low feed, BC loss ws importnt irrespective of the MY. When the feeding level nd the interction etween feeding level nd MY re not included in the model, MY hs no significnt effect on BC loss: in the entire popultion, the link etween MY nd BC chnge ws disrupted. In studies investigting the reltionships etween reproduction nd MY or BC loss, these two fctors re often confounded. Our Tle 3 Adjusted MY nd BCS (0 to 5 scle) for Holstein cows suject to either high or low feeding level during lcttion (Model MY-1) Feeding Model High Low r.s.e. R 2 P feeding No. of lcttions 50 48 44 weeks MY (kg) 8410 5719 851 79% *** Pek MY (kg/dy) 41.5 29.3 4.3 79% *** BCS t clving 2.98 2.96 0.42 41% 0.85 BCS t ndir 1.81 1.42 0.50 33% *** MY 5 milk yield; BCS 5 ody condition score; r.s.e. 5 residul stndrd error. ***, **, *, - P, 0.001, 0.01, 0.05, 0.10. BC chnge from clving to 120 dys 1 0-1 -2-3 - 0.012 P = 0.43 + 0.012 P = 0.13-0.052 P < 0.001 10 20 30 40 50 60 Pek milk yield (kg/dy) Figure 2 Body condition (BC) chnge from clving to 120 dys post prtum (ody condition score (BCS) on 0 to 5 scle) ccording to the pek milk yield (MY) for Holstein cows of the high (K, n 5 50) nd low (J, n 5 48) feeding level groups. The solid lines represent the predicted regressions within feeding level when ccounting for lcttion numer, yer, BCS t clving, feeding level nd pek MY, nd the interction etween these two vriles (R 2 5 51%, r.s.e. 5 0.46). The dshed line represents the predicted regression when not ccounting for feeding level nd its interction with pek MY (R 2 5 29%, r.s.e. 5 0.55). Regression slopes nd corresponding P-vlues re given on the plot. design prtilly disrupted the reltionship etween these two fctors nd, therefore, provided n opportunity to disentngle the respective effects of MY nd BC t ech reproductive step in Holstein diry cows. Owing to the welldocumented reltionships etween reproduction nd oth MY nd BCS (Roche et l., 2009; Friggens et l., 2010), we hve ssumed in our study tht reproductive performnce is minly explined y the expressed MY nd BCS level or chnge, lthough we cnnot exclude ny direct feeding tretment or genetic effect. Pp cyclicity ws linked to BCS nd BC chnge vriles C-LA ws minly relted to BCS in erly lcttion. C-LA hs een extensively studied in diry cows, s fr too lte first ovultion delys reeding time. The reltionship etween C-LA nd prity, genitl helth prolem nd BCS pp is 480
Milk nd ody condition effects on reproduction C-LA (dys) 80 60 40 20 0 0 1 2 3 4 BCS t 30 dys pp Figure 3 Predicted commencement of lutel ctivity (C-LA) ccording to the ody condition score (BCS) t 30 dys post prtum for Holstein primiprous ( &) nd multiprous (KJ) cows with (- - -) or without ( ) genitl helth prolem t or following clving (n 5 98). The corresponding equtions re log e (C-LA) 5 4.39 6 0.17 if primiprous/multiprous 6 0.13 if genitl helth prolem/no prolem 0.3523BCS 30 (P 5 0.002, P 5 0.027, P, 0.001, r.s.e. 5 0.49, R 2 5 25%; vrile definitions re in Tle 2). consistently oserved in the literture (Opsomer et l., 2000; Roche et l., 2009), wheres Figure 3 illustrtes tht higher BCS 30 is ssocited with shorter intervl to C-LA (Model CYC-1, slope 20.35, i.e. 20.22 log e dys per 0.63 unit BCS increse (1 s.d.), P, 0.001). Our nlysis reveled tht the protein content t ndir could lso e included in the model s higher protein content t ndir ws ssocited with shorter C-LA (slope 20.054, i.e. 20.10 log e dys per 1.9 g/kg increse (1 s.d., men 26.5 g/kg), P 5 0.043), wheres MY vriles did not ffect the model t ll (P. 0.5). All pp BCS vriles hd greter effect thn BC chnge vriles. Among those vriles, DBC 30 ws the most predictive one when BCS 30 ws not included in the model. A shorter C-LA ws ssocited with reduced BC loss (slope 20.44, i.e. 20.18 log e dys per 0.40 unit BC chnge increse (1 s.d., men 20.81 unit), P, 0.001, R 2 5 21%). In this model, the prity effect ws no longer significnt (P 5 0.17), s primiprous cows hd oth longer C-LA nd greter BC loss thn multiprous cows. Among the metolic vriles, only the plsm glucose concentrtion t 20 dys in milk hd significnt effect when BCS 30 ws not included in the model (slope 20.019, i.e. 20.12 log e dys per 6.2 mg/dl glucose concentrtion increse (1 s.d., men 58.2 mg/dl), P 5 0.044, R 2 5 17%). Therefore, lthough pp BCS vriles were more predictive thn BC chnge vriles, the influence of negtive energy lnce (Butler et l., 1981) ws oserved once more, s lrge BC chnge within 30 dys pp, low milk protein content t ndir nd low plsm glucose concentrtion t 20 dys pp lso dely the resumption of lutel ctivity. The comined effects of oth the level nd the chnge of ody lipid reserves my e expected, s reported in eef cows y Wright et l. (1992). A lrge BC loss my not prevent the cow from ovulting if the current BCS is still high, nmely if the nte-prtum BCS ws high enough, nd is consistent with the concept of n endocrine memory (Chgs et l., 2007). However, in our dtset, including n interction etween BC chnge from clving to 30 dys pp nd ctegorized BCS t clving (. v.,men within prity) did not improve the prediction of C-LA (results re not reported). Moreover, BCS t 30 dys pp does not explin the delyed resumption of lutel ctivity for primiprous cows wheres BC chnge does. In comprison with multiprous cows, BCS my not represent the sme mount of ody lipid reserves for primiprous cows where prolonged uterine involution nd the requirements to grow could lso ffect lutel ctivity independent of BCS sttus. In conclusion, the respective influences of the level nd dynmic of ody lipid reserves on C-LA remin to e quntified. PLP profiles were minly relted to C-LA nd BC chnge. PLP, which dely reeding y preventing the occurrence of ovultion nd lso y complicting the understnding of cows cycles for the frmer, hs ecome mjor normlity in Holstein cows (Royl et l., 2000). Once ccounted for log e (C-LA) in logistic regression Model CYC-2, the est predictive vriles for these norml profiles were DBC 60, DBC 90 nd DBC 120 (P 5 0.002, 0.007 nd 0.002, respectively). The sooner lutel ctivity commenced nd the deeper the DBC 60, the greter the proility of hving t lest one PLP pp (Tle 4). Whtever the model, neither MY vriles, nor prity or genitl helth prolem t/following clving hd significnt effect (P. 0.2). In other words, cows resuming lutel ctivity erly nd moilizing reserves for long time hd PLP (lsting more thn 25 dys). According to the previously discussed predictors of C-LA, oth high BCS t clving nd intermedite BC chnge from clving to 30 dys pp re proly required to mintin BCS t 30 dys pp nd to ovulte erly. This result is supported y the different BCS curves oserved for delyed-, norml- nd PLP-clssified cows (Figure 4). Delyed cows were thinner thn norml cows fter clving. PLP cows tended to hve higher BCS efore clving thn delyed cows, ut reched similr low BC y 60 dys in milk, owing to lrge nd prolonged BC loss. Conversely, MY curves did not differ etween the P 4 profile ctegories. A higher BCS t clving for PLP cows is often oserved (Pushpkumr et l., 2003; Tylor et l., 2003; Shresth et l., 2005) ut is more rrely significnt (Disenhus et l., 2002). This suggests nrrow optimum for BCS t clving: too thin nd too ft re known to dely ovultion, wheres slightly over-conditioned nimls re lso susceptile to n incresed risk of PLP. Metritis is often considered to e mjor risk fctor of PLP (Opsomer et l., 2000); however, the inclusion of genitl helth prolems t clving did not improve models, even when distinguishing metritis from other events. Interestingly, ll of the 15 cows tht encountered severe metritis were well clssified using 0.25 cut-off proility (.0.25 is PLP predicted) in the finl model predicting the PLP profile with regrd to log e (C-LA) nd DBC 60 : nine norml nd four PLP profiles. Metritis nd PLP my hve similr risk fctors, 481
Cutullic, Dely, Gllrd nd Disenhus Tle 4 ORs nd 95% CIs for the inomil logistic regression model used to predict the occurrence of PLP profile in 98 progesterone profiles of Holstein cows (decrese in the residul devince, P 5 5.8E25; Model CYC-2) Risk fctor P (.x 2 ) Men 6 s.d. OR 1 (PLP v. no PLP) 95% CI 1 Log e (C-LA),0.001 3.53 6 0.55 0.271** 0.118 to 0.620 BC chnge from clving to 60 dys pp 0.002 21.08 6 0.54 0.406** 0.213 to 0.776 OR 5 odds rtio; CI 5 confidence intervl; PLP 5 prolonged lutel phse; C-LA 5 commencement of lutel ctivity; pp 5 post-prtum. The intercept of the model is 5.01. ***, **, *, - OR is significntly or tends to e different from 1 (P, 0.001, 0.01, 0.05, 0.10). 1 For continuous vriles, ORs re expressed s n increse in 1 s.d. insted of 1 unit (i.e. exp(coefficient of the liner predictor) s.d. ). BCS () 3.0 2.0 1.0-30 0 30 60 90 120 Dys fter clving () () Δ BC 0-30 0 30 60 90 120-1 -2 Dys fter clving Figure 4 Curves of the fitted ody condition score (BCS, 0 to 5 scle) nd ody condition chnge from clving (DBC) of Holstein cows ccording to the post-prtum (pp) progesterone profile ctegory (J, norml; m, prolonged lutel phse; n, delyed; n 5 53, 23 nd 18, respectively). Models lso ccounted for prity. At the sme pp intervl, vlues with common letter re not significntly different (P. 0.05) nd vlues with common letter in rckets tend to e different (P, 0.10). ut the elongtion of the lutel phse cn proly occur without ny uterine disorder. In 2 3 2 nutritionl experimentl design, Vill-Godoy et l. (1990) induced overfttening or norml BC in diry heifers nd then restricted their feeding or not. The first nd second lutel phses were lengthened y 3 nd 5 dys only for ft-restricted heifers. Thus, comintion of lrge ody lipid reserves t clving nd strong energy deficit in erly lcttion my lengthen lutel phses. More precisely, the ongoing cycle type ws relted to the BC dynmic t the time of ovultion. The finl model used to predict the occurrence of PLP cycle, short cycle (<15 dys) or norml cycle following ovultion (Model CYC-3) included ovultion rnk (first or lter), BC slope t ovultion nd BC chnge from clving to ovultion. Irrespective of the ovultion rnk, severely declining BC round ovultion incresed the proility of PLP (Figure 5, Tle 5). Conversely, slightly declining or n incresing BC round ovultion incresed the proility of short cycle, ut t first ovultion only, s short cycles re scrce nywy t susequent ovultions (only 2% of ovultions of rnks 2-3-4 were followed y short cycle). The proility of short cycle ws ll the more high s BC chnge from clving to first ovultion ws negtive. Of the PLP events, hlf occur t first ovultion, ut cn occur up to the fourth ovultion. According to our results, ovultions tht occur while moiliztion is still ongoing re more likely to e followed y PLP. Although requiring further investigtion, our results indicte tht this BC chnge effect remins even when ccounting for dys fter clving. The fct tht short cycles conversely occur while the cow is gining BC close to ovultion is of interest. To our knowledge, there is no literture investigting the reltionship etween BC chnge nd the reliztion of short nd long lutel phses t first ovultion. Butler et l. (1981) considered tht trnsitory P 4 elevtion my occur when the cow returns to positive energy lnce, ut concluded tht it ws not relted to lutel ctivity. The proility of short cycle increses s BC chnge from clving to ovultion widens, suggesting possile crry-over effect of the negtive energy lnce period on folliculr ctivity, ut only t first ovultion. The risk of short cycles lso increses s dys in milk increse; however, prt of this effect my relte to genetics, s C-LA nd the length of the first lutel phse re negtively correlted geneticlly (Royl et l., 2002). Overll, these BC effects on ovrin ctivity re consistent with the much etter ovrin ctivity in reeds with higher BCS throughout lcttion nd smoother BC chnge, such s the Normnde reed (Cutullic et l., 2011). Oestrus detection relied on current MY Oestrus detection is key element of the reproductive performnce, especilly in compct clving systems. Theoreticlly, during 12-week reeding period, cows only hve one to four ovultions ccording to their clving dte. In the finl Model OES-1 used to predict ovultion detection, the odds of eing detected in oestrus were multiplied y 3.0 when t 482
Milk nd ody condition effects on reproduction Figure 5 Proilities of the occurrence of either norml cycle ( J ), short cycle ( & ) or prolonged lutel phse cycle (- -J- -), predicted t first ovultion (Ov.1) nd second third fourth ovultions (Ov.2-3-4) not followed y n insemintion in 189 cycles of Holstein cows, ccording to the ody condition (BC) slope from 230 to 130 dys reltive to the dy of ovultion (left censored t clving), predicted for BC chnge from clving to ovultion of 20.83 (Men), 21.35 (Men21 s.d.) or 20.31 (Men 1 1 s.d.) for the short cycles ctegory. Model chrcteristics re given in Tle 5. Tle 5 ORs nd 95% CIs for the multinomil logistic regression model used to predict the occurrence of norml cycle, short cycle or PLP cycle following ovultion in Holstein cows. Prediction relied on 189 ovultions of rnks 1 to 4, occurring efore 135 dys in milk, not preceding delyed ovultion of rnk ove 2, not elonging to disorder clssified progesterone profile nd not followed y n insemintion (decrese in the residul devince, P 5 7.6E212; Model CYC-3) Risk fctor P (.x 2 ) Ctegory % Men 6 s.d. OR 1 PLP v. OR 1 short v. norml cycle 95% CI 1 norml cycle 95% CI 1 Ovultion rnk,0.001 2-3-4 52% 1 1 First 48% 1.37 0.42 to 4.54 78.3*** 13.1 to 469 BC slope 2,0.001 20.32 6 0.30 0.47* 0.26 to 0.84 2.52** 1.38 to 4.62 DBC clving to ovultion 0.012 20.83 6 0.52 1.11 0.63 to 1.97 0.47** 0.27 to 0.82 PLP 5 prolonged lutel phse; OR 5 odds rtio; CI 5 confidence intervl; BCS 5 ody condition score; BC 5 ody condition. Intercepts of the model re 22.54 for PLP nd 22.78 for the short cycle; the norml cycle is the reference ctegory. ***, **, *, - OR is significntly or tends to e different from 1 (P, 0.001, 0.01, 0.05, 0.10). 1 For continuous vriles, ORs re expressed s n increse in 1 s.d. insted of 1 unit (i.e. exp(coefficient of the liner predictor) s.d. ). 2 BC slope from 230 to 130 dys reltive to the dy of ovultion (left censored t clving), expressed in BCS unit per 30 dys. lest one other cow ws in oestrus, were divided y 1.7 for n increse in MY y 7.8 kg/dy (1 s.d.) nd were multiplied y 1.6 for n increse in BCS t ovultion y 0.66 unit (1 s.d.; Tle 6). Wheres ll MY vriles hd significnt effect (P, 0.001), MY dynmic nd generl vriles hd no effect. It ws similr for the proility of detecting ovultion y stnding oestrus (Model OES-2), except tht BCS t ovultion hd no significnt effect (P. 0.4; Tle 6). Thus, s oserved in our previous work (Cutullic et l., 2009), n increse in MY t ovultion depresses oestrous ehviour independent of BC chnge. This deleterious effect of MY is consistent with descriptive studies tht possily confound MY nd BC loss effects (Lopez et l., 2004; López-Gtius et l., 2005). However, too low BCS my lso suppress oestrous ehviour, possily y ltering ovrin ctivity. Indeed, when we investigted the possile links etween cyclicity vriles nd ovultion detection (Model OES-1is), we oserved tht delyed ovultions were scrcely detected. Among the vriles relted to the previous cycle, only hving current delyed ovultion could ffect the model previously reported in Tle 6, s none of the six delyed ovultions were detected. When ccounting for this vrile in the model, the effect of BCS t ovultion nrrowed (odds rtio 1.8, 95% CI 1.1 to 3.0). The results suggest tht threshold my exist, under which oestrous ehviour is compromised nd ove which oestrous ehviour is relted to MY. Cutullic et l. (2009) oserved positive effect of losing more BC on oestrus signs; however, the winter totl mixed rtion low-feeding level ws higher nd cows did not rech s low BCS s in this work. In tht study, the 2 3 2 fctoril design (high v. low feeding in winter 3 high v. low feeding t grzing, llowing cows to switch from high-to-low or low-to-high MY) suggested n instntneous MY effect. This my help in chieving strtegies to improve oestrus detection round the reeding period. Physiologiclly, sed on stnding oestrus durtion quntifictions (i.e. ovultions without stnding oestrus were excluded), Wiltnk et l. (2006) suggested tht this deleterious effect of MY could e ecuse of incresed 17-oestrdiol nd P 4 heptic clernce rtes, owing to 483
Cutullic, Dely, Gllrd nd Disenhus Tle 6 ORs nd 95% CIs for the two inomil logistic regression models used to predict ovultion detection (n 5 255; Model OES-1) nd ovultion detection y stnding oestrus (n 5 241; Model OES-2) in ovultions of rnks >2 of Holstein cows (decrese in residul devince, P 5 1.6E27 nd P 5 2.4E26) Risk fctors P (.x 2 ) Detection Ctegory % Men 6 s.d. OR 1 95% CI 1 P (.x 2 ) Detection y stnding oestrus Ctegory % Men 6 s.d. OR 1 95% CI 1 Herdmte in oestrus,0.001 0.027 Asence 35% 1 37% 1 Presence 65% 3.01*** 1.66 to 5.48 63% 1.91* 1.07 to 3.40 MY ovultion week,0.001 29.8 6 7.8 0.57*** 0.43 to 0.77,0.001 29.6 6 7.8 0.59*** 0.44 to 0.79 BCS t ovultion 0.001 1.99 6 0.66 1.64** 1.20 to 2.25 Oservtion period 0.427 0.005 Before 9% 1 9% 1 During 91% 1.52 0.54 to 4.27 91% 4.92* 1.36 to 17.7 OR 5 odds rtio; CI 5 confidence intervl; MY 5 milk yield; BCS 5 ody condition score. Intercepts re 0.46 for the detection model nd 20.15 for the detection y stnding oestrus model. ***, **, *, - OR is significntly or tends to e different from 1 (P, 0.001, 0.01, 0.05, 0.10)., Not included in the finl model. 1 For continuous vriles, ORs re expressed s n increse in 1 s.d. insted of 1 unit (i.e. exp(coefficient of the liner predictor) s.d. ). higher dry mtter intke. However, if ssumed to e the only effect, the dry mtter hypothesis is inconsistent with oth the similr ovultion detection rte nd the similr proportion of oestri detected y stnding ehviour, which we oserved for Normnde nd Holstein cows (Cutullic et l., 2011). Indeed, Holstein cows hve higher dry mtter intkes thn Normnde cows in grss-sed feeding systems (Dillon et l., 2003) nd should therefore hve poorer oestrus expression. Such dry mtter effect should thus e considered within reed. We my lso wonder whether low BC cn enhnce oestrdiol production (Cutullic et l., 2009), possily y reducing the inhiitory effect of leptin on steroidogenesis (Spicer, 2001). Direct links etween the mmmry glnd, ovries nd rin my lso exist, ut they re still poorly understood. While investigting possile links etween cyclicity nd oestrus (Model OES-1is), we lso oserved tht the odds of ovultion detection were divided y 2.6 (95% CI 1.2 to 5.5) for cows with PLP profile compred with cows with norml profile. Thus, PLP cows were more t risk of hving undetected ovultions, regrdless of whether or not these ovultions followed the PLP itself. This result is consistent with the concomitnt increse in the PLP frequency (Royl et l., 2000) nd the decrese in oestrus durtion nd intensity (Vn Eerdenurg et l., 1996) over the lst decdes in Holstein cows. PLP occurrence ppers to e relted to ody lipid reserves, wheres oestrus ehviour is ssocited with MY. This suggests tht some effects on oestrus re not ccounted for y the MY level nd requires further investigtions. Emryo survivl ws linked to oth BCS nd MY Reclving requires fertiliztion, erly emryo survivl, ut lso lte emryo survivl nd foetl survivl. Both erly emryo mortlity nd lte emryo mortlity re thought to e the mjor risk fctors for pregnncy filure in temperte climte, in vrying proportions ccording to regions, cow types nd mngement strtegies (Diskin et l., 2006; Humlot et l., 2009). An increse in BCS min significntly decresed the risk of non-fertiliztion or erly emryo mortlity rther thn emryo survivl t first nd second insemintions (Tle 7, Model FER-1). However, n increse in BCS min only tended to increse the reclving rte (Tle 7, Model FER-2). In fct, in our study, lte emryo mortlity did not pper to e BCS relted, ut rther MY relted. Indeed, n increse in the slope of MY round ovultion (i.e. fltter MY slope round ovultion, s MY slopes re mostly negtive) decresed the risk of lte emryo mortlity rther thn emryo survivl. Significnt effects were lso found for most of MY vriles t the eginning of lcttion (P, 0.05), suggesting link to the shpe of the lcttion curve. Indeed, the Wood curve persistency prmeter c hd significnt effect on insemintion results (P 5 0.003). In order to illustrte this effect, four lowpersistency curves nd four high-persistency curves (within the rnge of our dtset) were drwn nd the corresponding predicted proilities of insemintion results were clculted (Figure 6). Regrdless of the MY level, cows with low persistency fter pek were more likely to encounter lte emryo mortlity, wheres cows with flt curves were more likely to encounter non-fertiliztion or erly emryo mortlity. In the finl Model FER-1 (non-fertiliztion or erly emryo mortlity v. lte emryo mortlity v. emryo survivl), glucose lso tended to hve n effect while replcing BCS min (P 5 0.065, n 5 122, six missing vlues). An increse in glucose concentrtion t insemintion y 5.7 mg/dl (1 s.d., men 62.8 mg/dl) divided y 2.0 the odds of non-fertiliztion or erly emryo mortlity rther thn lte emryo mortlity (95% CI 1.1 to 3.7). Our results suggest tht low BCS increses the risk of erly emryo mortlity, even though MY is low, wheres the effects of pek lcttion increse the risks of lte emryo mortlity (Figure 6) even when correct BCS is mintined. MY nd BCS my thus ffect oocyte nd emryo development t different stges. 484
Milk nd ody condition effects on reproduction Tle 7 ORs nd 95% CIs for the two logistic regression models used to predict either the occurrence of nf-eem v. lem v. emryo survivl (Model FER-1, multinomil) or the occurrence of reclving (Model FER-2, inomil) following 128 rtificil insemintions of rnks 1 nd 2 of Holstein cows (decrese in the residul devince, P 5 0.005 nd 0.085) Model FER-1 Model FER-2 Ctegory % Men 6 s.d. P (.x 2 ) OR 1 nf-eem v. survivl 95% CI 1 OR 1 lem v. survivl 95% CI 1 P (.x 2 ) OR 1 reclve v. not 95% CI 1 Risk fctors Prity 0.845 0.704 Primiprous 41% 1 1 1 Multiprous 59% 1.24 0.52 to 2.98 1.28 0.44 to 3.75 0.86 0.39 to 1.90 Clving AI 0.100 0.091,50 dys 17% 1 1 1 >50 dys 83% 0.38 0.12 to 1.27 2.13 0.22 to 20.59 2.87 0.75 to 10.9 MY slope 1,2 20.071 6 0.141 0.006 0.94 0.62 to 1.42 0.43** 0.24 to 0.77 BCSmin 1 1.70 6 0.55 0.098 0.63* 0.40 to 0.99 0.95 0.56 to 1.62 0.089 1.40-0.94 to 2.08 OR 5 odds rtio; CI 5 confidence intervl; nf-eem 5 non-fertiliztion erly emryo mortlity; lem 5 lte emryo mortlity; MY 5 milk yield. ***, **, *, - OR is significntly or tends to e different from 1 (P, 0.001, 0.01, 0.05, 0.10)., Not included in the finl model. Intercepts of Model FER-1 re 2.00 for the nf-eem ctegory nd 22.06 for the lem ctegory; the intercept of Model FER-2 is 22.45. 1 For continuous vriles, ORs re expressed s n increse in 1 s.d. insted of 1 unit (i.e. exp(coefficient of the liner predictor) s.d. ). 2 Milk yield slope over weeks 22 to12 reltive to the week of ovultion, expressed in kg/dy. A negtive energy lnce pp is widely considered to ffect oocyte nd emryo developmentl competence in high-yielding diry cows (Leroy et l., 2008) nd low BCS or lrge BC chnge re often reported s mjor risk fctors of infertility (Roche et l., 2007). Little is known out the potentil direct deleterious effects of current MY or lcttion curve shpe on lte emryo survivl, which my e relted to over-exposure to metolic or glctopoietic hormones. In field study, Grimrd et l. (2006) previously reported deleterious effect of high pek MY on emryo foetl survivl. This shpe effect my explin the reduced lte emryo mortlity rtes in reeds with fltter lcttion curves, such s the Normnde reed (Cutullic et l., 2011). However, further investigtion is needed to confirm this result. Similrly, the conflicting impct of plsm glucose concentrtions on erly nd lte emryo development remins to e confirmed. While investigting the possile links etween cyclicity, oestrus nd fertility (Model FER-2is), it ppered tht fertility t first nd second insemintions ws not considerly influenced y previous steps. The Model FER-2 lredy included prity, clving AI intervl,50 dys, MYslope oestrus week 2toþ2 nd BCS min (Tle 7). Log e (C-LA) tended to hve n effect when BCS min ws removed from the model (P 5 0.082). An increse in Log e (C-LA) y 0.47 (1 s.d., men 3.49) multiplied the odds of non-fertiliztion or erly emryo mortlity rther thn emryo survivl y 1.6 (95% CI 1.0 to 2.5). The occurrence of previous short cycle lso tended to improve the model (P 5 0.098), s the 13 insemintions following short cycles resulted in only two emryo survivls. Other previous cycle chrcteristics including P 4 profile ctegory nd stnding ehviour detection did not ffect the model (P. 0.4) nd therefore good oestrus expression ws not ssocited with improved fertility. This result seems logicl with regrd to the previously oserved effects of milk production on oestrus nd of BC on fertility nd suggests tht oestrus cn e compromised without compromising fertility, wheres the converse is lso true. In our previous study (Cutullic et l., 2009), the clerer the ehviourl signs were, the higher ws the reclving rte. However, low-fed cows were not chllenged to the sme intensity s in this study (Dely et l., 2009), nd the oestrus fertility reltionship proly did not rech uncoupling s we hve oserved here. Overll, our results point to possile uncoupling etween cyclicity, oestrus nd fertility nd llow for compenstions leding to similr finl reproductive performnce in mny studies tht compred feeding levels (Horn et l., 2004; Pederner et l., 2008; Pollott nd Coffey, 2008; Dely et l., 2009). Conclusion This study provides new informtion on the respective effects of MYnd of ody lipid reserve level nd chnge on the overll reproductive process. The results indicte tht these effects my e different t different stges, with some reproductive steps relted to ody lipid, wheres others my relte to MY. 485
Cutullic, Dely, Gllrd nd Disenhus Milk yield (kg/dy) 50 40 30 + Wood prmeter c = 0.024 + - + 20 - - NF-EEM 0.42 10 LEM 0.10 Survivl 0.47 0 0 3 6 9 12 Weeks post-prtum Milk yield (kg/dy) 50 40 30 20 10 0 + - Wood prmeter c = 0.095 + - + - NF-EEM 0.28 LEM 0.35 Survivl 0.38 0 3 6 9 12 Weeks post-prtum Figure 6 Exmples of eight Wood lcttion curves (MY 5 3 week 3 exp 2c 3 week ), with the persistency prmeter c equl to either 0.024 (i.e. popultion men 1 s.d.) or 0.095 (i.e. men 11 s.d.) nd the corresponding proilities of non-fertiliztion erly emryo mortlity, lte emryo mortlity nd emryo survivl t first nd second insemintions in Holstein cows (n 5 128). These insemintion result proilities re predicted for clving rtificil insemintion (AI) intervl >50 dys nd Wood curve prmeter c equl to either 0.024 or 0.095, in multinomil logistic model ccounting for clving AI intervl (P 5 0.015) nd Wood curve prmeter c (P 5 0.003; model decrese in residul devince, P 5 2.7E23). For Wood lcttion curves, prmeters, nd c control the milk yield scle, rise nd persistency, respectively. Within ech grph, tht is, for ech fixed persistency prmeter c, the four lcttion curves re generted y fixing the scle prmeter to its men either 11 s.d. (1)or21 s.d. (2) nd y fixing the rise prmeter to its predicted vlue either 11 r.s.e. (1) or 21 r.s.e. (2) in multiple regression model including prmeters nd c. Cyclicity nd erly emryo survivl were ssocited with BCS level nd chnge with deleterious consequences rising from excessively low BCS or high BC chnge. Conversely, oestrous ehviour nd lte emryo mortlity were minly ssocited with MY, level nd chnge, respectively. These results lso point to possile uncoupling etween cyclicity, oestrus nd fertility nd suggest tht there is potentil for compenstion long the reproductive process. In compct clving systems requiring high nd compct pregnncy rtes, high MY strtegies pper to e unsuitle even if BCS is mintined, owing to depressed oestrous ehviour nd proly incresed lte emryo mortlity tht delys rereeding. Similrly, strtegies tht compromise cyclicity nd fertility y excessively low BCS re unsuitle. The results of this study lso indicte tht the thresholds wherein reproductive processes ecome compromised remin to e determined in order to optimize reproductive performnce. Acknowledgements The uthors thnk the stff of the INRA experimentl frm Le Pin-u-Hrs, especilly Yves Cronnier, Loïc Leloup, René Dernucourt, Nicols Desrmé, Guillume Michel, Ségolène Leurent nd Mrion Pvec for the time spent smpling, oserving nimls, recording nd orgnizing dt. Jcques Portnguen nd Clr Lmrd re lso cknowledged for their lrge impliction in P 4 ssys. The uthors lso thnk Brendn Horn for improving the English nd for his scientific suggestions. This work received finncil support from the Bsse-Normndie regionl council. References Brt A, Druet T, Boniti B, Guillume F, Colleu JJ nd Boichrd D 2005. Biln phénotypique de l fertilité à l insémintion rtificielle dns les trois principles rces litières frnçises. Rencontres utour des Recherches sur les Ruminnts 12, 137 140. Bzin S, Augerd P, Crteu M, Chmpion H, Chillird Y, Cuylle G, Disenhus C, Durnd G, Espinsse R, Gscoin A, Godineu M, Jounne D, Ollivier O nd Remond B 1984. Grille de nottion de l étt d engrissement des vches pie-noires. RNED ovin, Pris, Frnce. Boichrd D, Brt A nd Briend M 2002. Evlution génétique des crctères de fertilité femelle chez les ovins litiers. Proceedings of Assocition pour l Etude de l Reproduction Animle (AERA), Reproduction, Génétique et Performnces, 6 décemre 2002, Pris, Frnce, pp. 29 37. Butler WR, Everett RW nd Coppock CE 1981. The reltionships etween energy lnce, milk production nd ovultion in postprtum Holstein cows. Journl of Animl Science 53, 742 748. Chgs LM, Bss JJ, Blche D, Burke CR, Ky JK, Lindsy DR, Lucy MC, Mrtin GB, Meier S, Rhodes FM, Roche JR, Thtcher WW nd We R 2007. Invited review: new perspectives on the roles of nutrition nd metolic priorities in the sufertility of high-producing diry cows. Journl of Diry Science 90, 4022 4032. Cutullic E 2010. Competition etween lcttion nd reproduction in the diry cow. PhD thesis, Cen Bsse-Normndie University, Frnce. Cutullic E, Dely L, Gllrd Y nd Disenhus C 2011. Diry cows reproductive response to feeding level differs ccording to the reproductive stge nd the reed. Animl 5, 731 740. Cutullic E, Dely L, Cuseur D, Michel G nd Disenhus C 2009. Hierrchy of fctors ffecting ehviourl signs used for oestrus detection of Holstein nd Normnde diry cows in sesonl clving system. Animl Reproduction Science 113, 22 37. Cutullic E, Benhim A, Brey S, Mitre H, Crreu S nd Disenhus C 2009. Does low feeding level enhnce estrdiol synthesis in preovultory follicles of Holstein 3 Normnde diry cows? Journl of Diry Science 92 (E-suppl. 1), 73. Dely L, Fverdin P, Michel G, Disenhus C nd Peyrud JL 2009. Effect of different feeding strtegies on lcttion performnce of Holstein nd Normnde diry cows. Animl 3, 891 905. Dillon P, Buckley F, O Connor P, Hegrty D nd Rth M 2003. A comprison of different diry cow reeds on sesonl grss-sed system of milk production: 1. Milk production, live weight, ody condition score nd DM intke. Livestock Production Science 83, 21 33. Disenhus C, Kerrt S nd Philipot JM 2002. L production litière des trois premières semines est négtivement ssociée vec l normlité de l cyclicité chez l vche litière. Rencontres utour des Recherches sur les Ruminnts 9, 147 150. Diskin MG, Murphy JJ nd Sreenn JM 2006. Emryo survivl in diry cows mnged under pstorl conditions. Animl Reproduction Science 96, 297 311. Friggens NC, Disenhus C nd Petit C 2010. Nutritionl su-fertility in the diry cow: towrds improved reproductive mngement through etter iologicl understnding. Animl 4, 1197 1213. 486