In order to achieve its milk yield potential the modern dairy cow is dependent on a high intake of dietary nitrogen (Zimmerman et al. Reference Zimmerman, Rakes, Jaquette, Hopkins and Croom1991). Nevertheless such diets, particularly those with high levels of effective rumen degradable protein (ERDP), have also been associated with decreased fertility (Laven & Drew, Reference Laven and Drew1999). However, the evidence for an effect of high intakes of ERDP on fertility is inconclusive, particularly for cattle fed normal commercial diets, with many studies showing little or no effect of increased dietary nitrogen on fertility (Laven et al. Reference Laven, Scaramuzzi, Wathes, Peters and Parkinson2007).
There are two potentially toxic by-products of protein metabolism: urea and ammonia (Visek, Reference Visek1984; Ocon & Hansen, Reference Ocon and Hansen2003). Urea is commonly suggested to be the primary toxic by-product (e.g. Rhoads et al. Reference Rhoads, Rhoads, Gilbert, Toole and Butler2006). However, Laven et al. (Reference Laven, Scaramuzzi, Wathes, Peters and Parkinson2007) concluded that much of the effect of increased intakes of degradable protein was probably mediated by ammonia rather than urea, and that in such cases plasma ammonia concentration was more closely correlated with reduced fertility than plasma urea concentration. There were two main reasons for this conclusion. The first was a comparison of two studies which investigated the effect of intakes of quickly degradable protein (in the form of urea) on the yield and quality of embryos. Dawuda et al. (Reference Dawuda, Scaramuzzi, Leese, Hall, Peters, Drew and Wathes2002) fed a diet that markedly increased plasma urea concentration (to 9 mm) and moderately increased postprandial plasma ammonia concentration (to 70 μm), while Sinclair et al. (Reference Sinclair, Kuran, Gebbie, Webb and McEvoy2000) moderately increased plasma urea concentration (to 7 mm) and greatly increased postprandial plasma ammonia concentration (to >300 μm). Whereas Sinclair et al. (Reference Sinclair, Kuran, Gebbie, Webb and McEvoy2000) reported that feeding their diet for 13 d prior to oocyte collection significantly reduced oocyte cleavage and blastocyst production, Dawuda et al. (Reference Dawuda, Scaramuzzi, Leese, Hall, Peters, Drew and Wathes2002) reported that feeding their diet for 17 d prior to embryo collection had no effect on either the number or the quality of embryos recovered, thus suggesting that plasma ammonia concentration was more important than urea concentration in determining whether a diet had a significant effect on embryo quality. Rhoads et al. (Reference Rhoads, Rhoads, Gilbert, Toole and Butler2006) suggested that Dawuda et al. (Reference Dawuda, Scaramuzzi, Leese, Hall, Peters, Drew and Wathes2002) would have seen an impact of their diet if they had implanted their recovered embryos into recipients. However, this conclusion does not explain why Sinclair et al (Reference Sinclair, Kuran, Gebbie, Webb and McEvoy2000) found a significant impact on embryo quality of a high-protein diet which elevated plasma ammonia concentration without markedly elevating urea concentration. The second rationale stated by Laven et al. (Reference Laven, Scaramuzzi, Wathes, Peters and Parkinson2007) was the lack of evidence, in cows fed diets based on grazed grass, for high dietary nitrogen intakes (particularly in the form of nitrates) having a deleterious effect on fertility (Kenny et al. Reference Kenny, Boland, Diskin and Sreenan2001; Laven et al. Reference Laven, Biggadike and Allison2002; Ordonez et al. Reference Ordonez, Parkinson, Matthew, Holmes, Miller, Lopez-Villalobos, Burke and Brookes2007). Laven et al. (Reference Laven, Scaramuzzi, Wathes, Peters and Parkinson2007) concluded that, although such diets were associated with high urea concentrations in milk and plasma, the relatively constant feed intake associated with grazing resulted in only a limited rise in systemic ammonia concentration. They then suggested that itwas the absence of this ammonia rise that prevented fertility from being affected.
Despite these findings, there has been little use of ammonia as a measure of the likely risk of dietary nitrogen-associated infertility outside of scientific studies because of the inherent difficulty of measurement of blood ammonia (Butler et al. Reference Butler, Calaman and Beam1996) and because of its significant association with time since feeding (Sinclair et al. Reference Sinclair, Kuran, Gebbie, Webb and McEvoy2000a). This has meant that urea, which is easier to measure and less variable during the day (Sinclair et al. Reference Sinclair, Kuran, Gebbie, Webb and McEvoy2000a), has been commonly used as the predictor for dietary nitrogen-associated infertility. This is reflected in the development by Butler et al. (Reference Butler, Calaman and Beam1996) of a blood urea threshold of 6·8 mm above which dietary nitrogen-associated infertility should be suspected.
Clearly, if, as suggested by Laven et al. (Reference Laven, Scaramuzzi, Wathes, Peters and Parkinson2007) and Sinclair et al. (Reference Sinclair, Kuran, Gebbie, Webb and McEvoy2000), plasma ammonia concentration is a better predictor for ERDP-associated infertility than urea concentration, we need to understand better the factors that influence urea and ammonia concentration and the relationship between them in order to identify any potential problems associated with using urea rather than ammonia as a predictor for reduced fertility.
As part of a study which investigated the effect of high quickly degradable nitrogen on follicular development and embryo growth (Laven et al. Reference Laven, Dawuda, Scaramuzzi, Wathes, Biggadike and Peters2004), multiple contemporaneous blood samples were taken and analysed for ammonia and urea. The present study used those data to investigate the factors affecting plasma urea and ammonia and the relationship between these two metabolic by-products.
Materials and Methods
Animals and their allocation
Forty-two Holstein dairy cows were selected from a dairy herd in southern England using the following criteria: (1) third or subsequent lactation, (2) calved in the previous 16 weeks and (3) deemed suitable for re-breeding following veterinary examination. Cows were ranked according to calving date and 9–18-d milk yield (mean 37·2±0·98 kg/d). Sequential pairs of cows within the ranking order were then identified. Within each pair, one cow was allocated randomly to each treatment (control or high-urea diet). Allocation to treatment occurred on two occasions; 22 cows were allocated to treatment on the first occasion and 20 cows 3 weeks later.
Management, feeding and treatments
All cows were housed in the same cubicle building and were bedded daily with access to fresh water at all times. They were milked twice daily and yields recorded automatically on each milking occasion.
The control diet was formulated to meet metabolizable energy and metabolizable protein requirements for early lactation cows according to UK recommendations (AFRC, 1993). The high-urea diet differed from the control diet only in the inclusion of 250 g urea per cow per day (670 g Provitlic120 per cow per day; Dallas Keith Ltd, Witney, Oxon, UK). The diets were fed to appetite once a day as a total mixed ration (TMR). All cows were fed the control diet for an initial acclimatization period of 3 weeks prior to allocation to treatment. The diets fed to the cows allocated at each of the two timepoints were identical. The main constituents of the control diet are shown in Table 1. After allocation to treatment cattle were synchronized using a PRID® (CEVA Animal Health, Chesham, Bucks, UK) for 8 d with an injection of 25 mg of dinoprost (Synthetic PGF2; Enzaprost® CEVA Animal Health), given the day before PRID® removal. Artificial insemination was carried out at 48 h and 72 h after PRID® removal. Subsequent inseminations were carried out as a result of observed oestrus supported by information from ovarian scanning and milk progesterone analysis. Cows were inseminated for a maximum of three cycles. Further details of diets and management are published in Laven et al. (Reference Laven, Dawuda, Scaramuzzi, Wathes, Biggadike and Peters2004).
Table 1. Constituents of the diet fed to the control group
All experimental procedures were carried out in accordance with the UK Animal Scientific Procedures Act (1986).
Measurements
Blood was collected weekly from all cows until 7 weeks post conception into EDTA-impregnated tubes (Greiner Labortechnik Ltd, Gloucestershire, UK) 3–4 h after feeding. The timing of the samples was designed to match the peak ammonia results found in a pre-study pilot. The blood was centrifuged at 1500 g at 4°C for 15 min and plasma recovered and stored at −18°C. Urea concentrations were determined in an auto-analyser using the method of Talke & Schubert (Reference Talke and Schubert1965). Plasma ammonia concentrations were determined using phase 2 of the same method, with decreased absorbance at 340 nm reflecting plasma ammonia concentrations (Mondzac et al. Reference Mondzac, Ehrlich and Seegmiller1965).
Statistical analyses
To compare the change in plasma ammonia concentration over the duration of the study with that of plasma urea concentration, models were created using study data which described how the two parameters changed over time. The data included in the model were individual milk yield, treatment group (control or high urea) and time of study start (whether cows were allocated to study on the first or second occasion). The data structure consisted of repeated measures taken over time on individual cows blocked within treatment groups. The variability of the data structure was preserved by building a mixed effects two level hierarchical model using the SAS PROC MIXED procedure (Littell et al. Reference Littell, Henry and Ammerman1998; Singer, Reference Singer1998) with the following model:
where Yijk is a vector of ammonia or urea responses at time k on the jth cow within treatment i; μ is the intercept, Rj(i), is the random effect of the jth cow nested within the ith treatment (normally distributed with mean zero and variance σd2); αi, is the fixed effect of the ith treatment; β1tjk, β2tjk2, β3tjk3, are the fixed effects of time (β1) with cubic (β2) and quadratic (β3) polynomials; eijk, is the random error of the jth cow at time k on treatment I (normally distributed with mean zero and variance σe2).
In a mixed effects model the fixed effects enter the model through the mean and the random effects enter the model through the variance.
The covariance matrix of the responses was explored using different correlation structures for the repeated measures (Littell et al. Reference Littell, Pendergast and Natarajan2000). Six covariance structures (compound symmetry (CS); first order autoregressive (AR(1)); first order autoregressive with random effects (AR(1)+RE); unstructured (UN); toeplitz (toep); and autoregressive moving average (ARMA(1,1)) were fitted to the data and the best model was selected based on the smallest values of fit statistics for Akaike information criteria (AIC), AIC corrected (AICC), and Bayesian information criteria (BIC). In the event of conflict the more parsimonious model was selected.
The AR(1)+RE covariance structure was proposed by Diggle (Reference Diggle1988) and in contrast to AR(1) the covariance between lags only decreases to a common contribution from the random effect of cow. In data sets, such as this, with a long series of repeated data the correlation from AR(1) can decrease almost to zero, which in many cases can be an inappropriate covariance structure.
The analysis was based on data from the first 102 d after the allocation of the first group of cattle to the high-urea diet. Seventeen repeated measurements were available for urea and 16 for ammonia. The missing data were assumed to be missing at random. When analysing unbalanced data in PROC MIXED, Spilke et al. (Reference Spilke, Piepho and Hu2005) recommended the use of REML to estimate the variance component and the Kenward–Roger (Reference Kenward and Roger1997) procedure for approximating the degrees of freedom. Initial data exploration included fitting separate kernel smoothed regression lines (bandwidth=10·75 d) to scatter plots of the ammonia, and urea responses against time for each of the two treatment groups. Fixed effects were tested using maximum likelihood (ML) and retained in the model at P<0·05 based on the type 3 tests of effects. Pre-planned treatment and start date comparisons were made with the probability of difference (PDIFF) option in the LSMEANS statement and declared significant at P<0·05.
The methods of Hamlett et al. (Reference Hamlett, Ryan, Serrano-Trespalacios and Wolfinger2003) and Roy (Reference Roy2006), which estimate correlation whilst adapting for the presence of repeated measures, were adapted to obtain a correlation coefficient between urea and ammonia.
Results
Urea
The covariance structure which best fitted the data was AR(1)+RE. There was no significant effect of days on trial, milk yield or start date and no significant interactions. The only significant effect found was that of treatment (P<0·001), with cattle on the high-urea diet having a mean plasma urea concentration 1·66 m greater than cattle on the control diet. The intra-class correlation (ICC) was 0·43, indicating that urea was highly correlated within cow.
Model
Ammonia
As with urea an AR(1)+RE model gave the best fit. However, in contrast to urea, factors other than treatment were found to have a significant effect on plasma ammonia concentration, with days on trial and start date included in both models. No significant interactions were found, nor was there a significant effect of milk yield.
Model
High urea
Control
The intra-class correlation (ICC) was 0·28 indicating that the ammonia results were less highly correlated within cow than urea.
Prediction of ammonia from urea results
An AR1+RE correlation structure was used to model the covariance of the repeated measures.
Model
The correlation (r) between urea and ammonia at any one time point measurement was 0·195 (r 2=0·038). This implies that, because of the variation of ammonia with time since the trial start, on its own urea explained only 3·8% of the variability of ammonia.
Discussion
This analysis clearly shows that for cattle fed a consistent diet the relationship between plasma urea and plasma ammonia concentration varied significantly with time. This has been previously shown for different diets (Sinclair et al. Reference Sinclair, Sinclair and Robinson2000a; Laven et al. Reference Laven, Scaramuzzi, Wathes, Peters and Parkinson2007); however, this analysis is the first to show this for cows on a consistent diet. So for cows on the same diet, as well as for cows on different diets, the same plasma urea concentrations can be associated with significantly different plasma ammonia concentrations. The correlation between the two parameters over the 102 d of this study was not significant, and for the cows in this study plasma urea concentration was a poor predictor for plasma ammonia concentration. As the control diet fed in this study was a conventional diet similar to that fed to many housed lactating cows, it is highly likely that these conclusions can be extrapolated to the field situation.
The present study was not designed to investigate how or why the factors that affect plasma ammonia or urea concentration produce their effect. Thus the models used in this analysis were developed to compare how plasma ammonia and urea concentrations changed during this study. The output from these models (Figs 1 and 2) clearly shows that the reason for the poor correlation between plasma urea and plasma ammonia concentration seen in this study was that plasma urea was unaffected by any of the factors included in the model (time after study start, milk yield, time of study start and treatment group) except for treatment group, whereas plasma ammonia concentration was affected by treatment group, time after study start and time of study start.
Fig. 1. Change with time in the concentration of plasma urea, with fitted regression line for each treatment group. Results from cattle given the control diet are represented using open circles, those from cattle given supplementary urea using filled circles.
Fig. 2. Change with time in the concentration of plasma ammonia, with fitted polynomial trend line for each treatment group for each start date. Results from cattle given the control diet are represented using open circles, those from cattle given supplementary urea using filled circles. Separate equations for the treatment and control groups were produced by re-running the final mixed effects model without an intercept.
Clearly the present results cannot identify the reasons for the different behaviour over time of plasma ammonia and plasma urea, but the data strongly suggest that whereas the primary determinant of urea concentration was total crude protein intake (consistent with the findings of Butler et al. Reference Butler, Calaman and Beam1996), the effect of crude protein intake on plasma ammonia was markedly influenced by other factors. The underlying causes of the changes in plasma ammonia seen in this study are likely to be a combination of extrinsic factors (e.g. changes in protein degradability of the diet) and intrinsic factors (e.g. increased rate of ammonia capture by the rumen micro-organisms). Further research is required to establish the most important factors.
Conclusions
If, as Laven et al. (Reference Laven, Scaramuzzi, Wathes, Peters and Parkinson2007) and Sinclair et al. (Reference Sinclair, Kuran, Gebbie, Webb and McEvoy2000) both suggest, increases in plasma ammonia lead to the reduced fertility associated with increased ERDP intake, then plasma urea concentration cannot be used as a proxy for plasma ammonia concentration when investigating fertility problems.
The authors thank the staff at ADAS Bridgets who were involved in the original study, especially Helen Biggadike, and all at RVC Pathology who were involved in the analysis of the samples. This study formed part of LINK project LK0621 which was funded by the Ministry of Agriculture Food and Fisheries (MAFF) and the Milk Development Council (MDC).
