Introduction

The main objectives of the nutritional management of heat-stressed dairy cows are to increase energy and nutrient density to counteract the reduction in dry matter intake (DMI) while maintaining milk production (MP), and to reduce the heat increment (HI) of the diet to improve thermoregulation. In general, this can be achieved with low-fiber and high-energy diets containing cereal grains (Baumgard et al. 2014). However, cereal grains differ in their HI due to differences in the rate and extent of rumen fermentation (Herrera-Saldana et al. 1990; Benninghoff et al. 2015). Previous studies using grain-fed sheep have shown that differences in rumen starch fermentability among grains have an impact on the severity of HS responses (Gonzalez-Rivas et al. 2016; Gonzalez-Rivas et al. 2017) and it was observed that treating wheat with 2% of a commercial starch-binding agent (SBA; Bioprotect, RealisticAgri, Ironbridge, UK) reduced the rate of in vitro rumen fermentation of wheat (Dunshea et al. 2012).

It is hypothesized that feeding slowly fermentable grains could reduce the amount of heat released from fermentation and digestion and this will ameliorate the physiological responses to HS and improve MP in dairy cows during summer. The aim of this study was to investigate the effect of feeding a diet based on corn grain or wheat treated with 2% SBA on the severity of HS responses including rectal temperature (RT), respiration rate (RR) panting score (PS), and rumen temperature (RuT), and on productive responses of lactating dairy cows under naturally occurring HS conditions during summer in Queensland, Australia.

Materials and methods

Animals and treatments

This study was undertaken at the University of Queensland, Gatton Campus; 27.4986 °S, 153.0155 °E, 89 m elevation during summer 2015 (February to March) in the sub-tropical region of Queensland, Australia. Twenty-four Holstein-Friesian cows ranging from first to third lactation, 635 ± 78.5 kg body weight (BW), 3.5 body condition score (BCS), 248.5 ± 64.6 days in milk (DIM), and 24.1 ± 5.5 kg milk/day, were randomly divided into three dietary treatment groups of eight cows based on parity, milk production, BW, BCS, and DIM using a randomized block design. Three dietary treatment groups were a control diet; a total mixed ration (TMR) plus crushed wheat (TMRW) and two intervention diets; a TMR plus crushed wheat treated with 2% of SBA (Bioprotect), (TMRB); and a TMR plus crushed corn (TMRC), (Table 1). Corn and wheat grains were crushed through a roller mill to a particle size of 2–5 mm.

Table 1 Composition of experimental diets fed to dairy cows on a DM basis
Full size table

Before the experiment, cows were provided with an introduction period of 7 to allow a gradual access to the experimental diets to avoid acidosis. For the duration of the study (29 days), diets were offered ad libitum once a day (0800 h) with a target DMI of 20.5 kg DM per cow per day (164 kg DM per group per day) and cows had ad libitum access to water. Cows were housed on a North-South-oriented feed pad divided in three pens (40 × 12 m) with one pen per treatment group each equipped with a water trough and a feed pad. Shade was provided in the feed pad and in the middle of the pens by an iron roof. No supplementary cooling was provided to the cows in the feeding and resting area. However, sprinklers and fans were used for 5 min before the afternoon milking.

Ambient temperature and relative humidity were obtained from an on-site weather station. Temperature-humidity index (THI) was calculated using the equation THI = (0.8 × T) + [(RH/100) × (T − 14.30)] + 46.4 where T is the ambient temperature (°C) and RH is the relative humidity (%) (Mader et al. 2006).

Physiological variables and milk sampling

Respiration rate and PS were measured 3 times a day: at 0600 h before morning milking, at 1100 h, and at 1600 h before afternoon milking. Respiration rate was measured by observing flank movements. Panting score was measured using the eight-point score scale described in Gaughan et al. (2008). Individual PS was collated for each treatment group and mean panting score (MPS) for each observation time (h) was then calculated.

Rumen temperature was recorded for 15 days using rumen boluses (RFID transmitter; Smartstock, OK, USA). One bolus per cow was orally inserted and placed in the rumen and RuT was transmitted at 20-min intervals towards two solar-powered stations: one in the feed pad and another in the milking area. Radio transmission data was converted to temperature values using a computer software (TechTrol Inc., OK, USA). Rumen temperature data was expressed as hourly and daily average for each cow. Data from one cow was not included due to zero observations. Rectal temperature was measured once every 4 days in the morning after milking (between 0700 and 1000 h) using a digital thermometer (DT-01, Tollot PTY, Ltd., NSW, Australia; range 32–41.9 ± 0.1 °C).

Cows were milked twice a day (~ 0600 and ~ 1600 h) in a herringbone milking parlor separated ~ 600 m from the feeding and resting area. Milk samples (30 mL) were collected twice every 4 days: on the afternoon before RT measurement and on the morning of RT measurement in milk preservative-containing flasks. Samples were maintained at 4 °C until sent to the laboratory (Australian Herd Recording Services, Nundah, QLD, Australia) for lactose, protein, and fat percentage analysis. Milk production data of day 24/02/2015 and milk quality data of day 9/3/2015 for all cows were removed from analyses because of unusual low values.

Average daily feed intake (ADFI) as DM for each group was recorded daily by weight difference between feed offered and refusals before the morning feeding. Individual BW was obtained weekly using walk over scales to calculate average daily gain (ADG).

Statistical analysis

GenStat V16 (GenStat release 16 VSN International Ltd., Hemel Hempstead, UK) was used for all statistical analysis. A restricted maximum likelihood (REML) analysis was used to test the effect of diet and time of observation on RR and PS. The model included diet (TMRW, TMRB, and TMRC) and time of observation (0600, 1100, and 1600 h) as fixed terms. The random terms were cow ID, lactation (1st, 2nd, and 3rd), pregnancy status (non-pregnant and pregnant), and day. For RT and milk quality, the fixed terms were diet and day of sampling (S1 to S8). The random terms were cow ID, lactation, and pregnancy status. Rumen temperature pooled during 24 h was analyzed as above; the fixed terms were diet, time of observation (from 0000 to 2400 h), and date. The random terms were cow ID, lactation, and pregnancy status. For individual BW change and ADG, the fixed term was diet and the random terms were cow ID, lactation, and pregnancy status. The Wald test and the Tukey pairwise comparison were used to test significance and to determine differences respectively. Fixed factors and interaction were statistically significant at P ≤ 0.05. Results were reported as means and pooled SED and descriptive statistics were used for ADFI.

Correlations were analyzed using Pearson; RR was correlated with daily average RuT. Individual RR was correlated with PS, RuT measured at each observation time (0600, 1100, and 1600 h), and with the THI at each observation time, THI of the day before the observation (1-day lag), THI of 2 days before the observation (2-day lag), and THI of 3 days before the observation (3-day lag). Rectal temperature was correlated with RuT, and THI pooled by hour during the time of RT measurement (between 0700 and 1000 h), 1-, 2-, and 3-day lag, RuT pooled by hour of the day was correlated with THI by hour of the day, and RuT pooled by day was correlated with Av THI, 1-, 2-, and 3-day lag.

Average daily feed intake was correlated with THI pooled by day (24 h THI average), 1-, 2-, and 3-day lag. Individual MP was correlated with Av RuT, daily THI, and 1-, 2-, and 3-day lag. Milk quality and milk solids were correlated with RuT, THI of the milk-sampling day, 1-, 2-, and 3-day lag. Variables were significantly correlated when P ≤ 0.05 and only significant correlations were reported. Rumen temperature during the time of RT measurement was subtracted from RT and then, a t test was used to determine significant differences between RuT and RT.

Results

Thermal conditions

Temperature-humidity index (mean ± SD) during the experiment was 72.4 ± 2.0 with min 65.4 ± 3.5 and max 79.0 ± 1.8; 76% of the experimental days had average THI ≥ 72, and THI < 64 was observed on 24% of the experimental nights at the end of the experiment. THI (mean ± SD) at observation times was 66.3 ± 3.7 at 0600 h, 75.9 ± 2.7 at 1100 h, and 77.3 ± 2.5 at 1600 h. THI (mean ± SD) during the time of RT measurement (between 0700 and 1000 h) was 70.0 ± 3.0, 72.8 ± 2.3, 74.0 ± 2.4, and 74.9 ± 2.4 for 0700, 0800, 0900, and 1000 h, respectively.

Average daily feed intake and average daily gain

Average daily feed intake (mean ± SD) per group was 170.6 ± 17.2 kg DM per day (~ 21.3 kg DM per cow per day) for TMRB, 168.0 ± 18.2 kg DM per day (~ 21.0 kg DM per cow per day) for TMRW, and 166.0 ± 18.9 kg DM per day (~ 20.8 kg DM per cow per day) for TMRC. Average daily feed intake was not correlated with THI (P > 0.05) and there was no effect of diet on ADG and BW change (P > 0.05; data not shown).

Milk production and quality

Cows fed TMRC and TMRB had greater MP than those fed TMRW (P = 0.004; Table 2). There was a significant effect of experimental day on MP (P < 0.001). There was a negative correlation between MP and THI 2-day lag in all diets (r = − 0.2; P < 0.001) and there was a negative correlation between RuT and MP for cows fed TMRB (r = − 0.43; P < 0.001) and TMRC (r = −0.41; P < 0.001).

Table 2 Variation in milk production and composition in Holstein-Friesian dairy cows fed TMRB, TMRC, and TMRW during the experiment. Means are predicted means pooled by diet
Full size table

Cows fed TMRC fed had lower milk fat percentage than cows fed TMRB and TMRW (P = 0.049). There was a significant effect of milk sampling day on milk protein and lactose percentage and milk solids (P < 0.05; Table 3). Cows fed TMRB exhibited a positive correlation between milk fat percentage and RuT (r = 0.48; P = 0.0094) and a negative correlation between lactose percentage and RuT (r = − 0.67; P < 0.001). Cows fed TMRW diet exhibited a negative correlation between milk protein percentage and RuT (r = − 0.5; P = 0.005).

Table 3 Variation in milk production and composition in Holstein-Friesian dairy cows fed TMRB, TMRC, and TMRW during the experiment. Means are predicted means pooled by sample day
Full size table

Mean panting score and respiration rate

There was no significant effect of diet on MPS (P > 0.05). Across diets, there was a significant effect of time of observation on MPS (0.8, 1.0, and 1.2 ± 0.04 for 0600, 1100, and 1600 h respectively; P < 0.001; Fig. 1a). There was no significant effect of diet on RR (P > 0.05). Across diets, there was a significant effect of time of observation on RR (57, 65, 74 ± 0.9 breaths/min for 0600, 1100, and 1600 h respectively; P < 0.001). Cows fed TMRC had higher RR at 1100 h than TMRB or TMRW fed cows (P = 0.013; Fig. 1b). Panting score and RR were positively correlated in cows fed all diets (r = 0.56; P < 0.001). Panting scores and RR were also positively correlated with THI at the time of observation and with 1-, 2-, and 3-day lag in cows fed all diets (P < 0.001 for all).

Fig. 1
figure 1

Relationship between mean panting score (a), respiration rate (b) and observation time in Holstein-Friesian dairy cows fed TMRB, TMRC, and TMRW. Results are means pooled across days and pooled SED for the interaction diet × observation time. P values for the effects of diet, observation time, and diet × observation time were 0.19, < 0.001, and 0.24, respectively for panting score, and 0.27, < 0.001, and 0.013, respectively for respiration rate

Full size image

Rectal temperature

Cows fed TMRC had lower RT than those fed TMRW and TMRB (38.9 vs 39.1 and 39.1 ± 0.06 °C respectively; P < 0.001). Across treatments, there was a significant effect of day of observation on RT (P < 0.001). For cows fed TMRB, there was a positive correlation between RT and THI (r = 0.26; P = 0.04), THI 1-day lag (r = 0.38; P = 0.002) and 2-day lag (r = 0.37; P = 0.003). For cows fed TMRC, there was a positive correlation between RT and THI (r = 0.27; P = 0.03), THI 1-day lag (r = 0.4; P = 0.001) and 2-day lag (r = 0.42; P < 0.001). For cows fed TMRW, there was a positive correlation between RT and THI 1-day lag (r = 0.31; P = 0.011) and 2-day lag (r = 0.40; P < 0.001).

Rumen temperature

There was no significant effect of diet on RuT pooled by hour of the day (P > 0.05). Figure 2 depicts the variation in RuT during the day (P < 0.001). Cows fed TMRC had higher RuT than those fed TMRB and TMRW between 0900 and 1400 h (P < 0.001). There was a negative correlation between RuT and THI by hour for cows fed TMRB (r = − 0.531; P = 0.0076). There was no significant effect of diet on RuT pooled by day (P > 0.05). Figure 3 depicts the variation in RuT during the experimental days (P < 0.001). Cows fed TMRC had higher RuT than those fed TMRB and TMRW during the initial days of the experiment. Then, cows fed TMRW had higher RuT than those fed TMRC and TMRB towards the end of the experiment (P < 0.05).

Fig. 2
figure 2

Relationship between rumen temperature and temperature-humidity index (THI) during the day (Av THI/h) in Holstein-Friesian dairy cows fed TMRB, TMRC, and TMRW. Results are estimated means per hour and pooled SED for the interaction diet × hour. P values for the effects of diet, hour, and diet × hour were 0.61, < 0.001, and < 0.001, respectively

Full size image
Fig. 3
figure 3

Relationship between rumen temperature and temperature-humidity index (THI) during the experimental days (THI/day and critical THI) in Holstein-Friesian dairy cows fed TMRB, TMRC, and TMRW. Results are estimated means per hour and pooled SED for the interaction diet × experimental day. P values for the effects of diet, experimental day, and diet × experimental day were 0.62, < 0.001, and 0.003, respectively

Full size image

There was a positive correlation between RuT and THI of the day (r = 0.91; P < 0.001), 1-day lag (r = 0.70; P = 0.001), 2-day lag (r = 0.56; P = 0.016), and 3-day lag (r = 0.50; P = 0.033) in cows fed TMRB. There was a positive correlation between RuT and the THI of the day (r = 0.92; P < 0.001), 1-day lag (r = 0.81; P < 0.001), 2-day lag (r = 0.75; P < 0.001), and 3-day lag (r = 0.74; P < 0.001) in cows fed TMRC. In cows fed TMRW, there was a positive correlation between RuT and the THI of the day (r = 0.88; P < 0.001), 1-day lag (r = 0.70; P = 0.001), and 2-day lag (r = 0.75; P = 0.038). Rumen temperature was positively correlated with RR in cows fed all diets (r = 0.53; P < 0.001).

There was a positive correlation between RT and RuT in all diets: TMRB (r = 0.613, P = 0.001), TMRC (r = 0.428, P = 0.046), and TMRW (r = 0.552, P = 0.0026). The differences (± SED) between RuT and RT were 0.26 °C ± 0.081; P = 0.001. For TMRW- and TMRB-fed cows, were no differences between RT and RuT (P > 0.05) while for TMRC fed cows, there were significant differences between RT and RuT (38.7 vs 39.3 ± 0.14 °C respectively: t (6) = − 8.18; P < 0.001).

Discussion

Cows in this experiment were exposed to summer conditions conducive to HS characterized by THI > 72 that exceeded the limit of mild heat stress conditions (Silanikove 2000). However, the observed night-time recovery (THI < 64) might explain the amelioration in the HS responses towards the end of the experiment (Gaughan et al. 2008). It cannot be negated that this experiment was carried out late in summer, possibly, after the onset of HS, and therefore, the reduction of MP might already have occurred.

The primary mechanism of reduction in MP during HS is via enhanced extra-mammary glucose utilization (Wheelock et al. 2010). Therefore, provision of glucose precursors to dairy cows is crucial during HS (Baumgard and Rhoads 2012). This was demonstrated with the higher MP observed in cows fed TMRC and TMRB, because slowly fermentable grain feeding increases the amount of glucose being absorbed from the small intestine to produce lactose, the main regulator of milk volume (Liu et al. 2013; Moharrery et al. 2014). Variations in milk quality in this experiment appeared to be a consequence of HS and diet. It was observed that Holstein cows fed forage plus concentrate diets as TMR had reduced milk production, fat, and milk protein during HS (Smith et al. 2013; Bertocchi et al. 2014). In addition, results obtained in our experiment agreed with Reynolds et al. (2001) where dairy cows fed slowly fermentable grains had reduced milk fat percentage and higher milk production because high-energy diets are associated with reduction in fat percentage due to a lower ratio acetate to propionate in the rumen (Bauman and Griinari 2001).

Respiration rate and PS were elevated during periods of elevated THI exceeding the critical level of 80 breaths/min and MPS > 0.8 indicating critical hyperthermia (Gaughan et al. 2002; Gaughan et al. 2008; Davison et al. 2016). The positive correlation between RR, PS, and RuT and between PS, RR, and THI demonstrates the negative effect of high thermal load on dairy cow’s physiological variables in agreement with Sullivan et al. (2014a). Kabuga (1992) and Sullivan et al. (2014a) observed a positive correlation between RT, MPS, and the average ambient temperature and THI on the day of observation and 1-day lag. Similarly, Sullivan et al. (2014b) observed a negative correlation between MP and the average THI on the day of observation and 1-, 2-, and 3-day lag. These observations agreed with the data obtained in the current experiment; this delay is explained by complex physiological responses including a lengthening in nutrient digestion and utilization during HS (West 2003).

Differences on RT between cows fed TMRC and TMRW or TMRB are concurrent with the results shown previously in sheep fed slowly fermentable grain-based diets (Gonzalez-Rivas et al. 2016; Gonzalez-Rivas et al. 2017). These findings support the idea that slowly fermentable grain feeding can reduce HI and metabolic heat production (Reynolds 2006; Russell 2007). This may indicate an improved heat-tolerance in cows fed TMRC characterized by a low RT and high RR and PS in the morning. This phenomenon is viewed as a pre-compensatory mechanism in ruminants that increases the temperature gradient between the body and the environment favoring heat dissipation (Sullivan et al. 2014a). The temperature gradient between the rumen and the rest of the body is also necessary to drive heat exchange (Beatty et al. 2008). It was confirmed that RT was lower than RuT and both were highly correlated in all diets in this experiment as reported by Beatty et al. (2008), Bewley et al. (2008), and Lees et al. (2014). Significant differences between RuT and RT in TMRC indicate a large thermal gradient between the rumen and the body.

The observed variation in RuT during the day is a consequence of the increase in core body temperature in response to feed intake, high ambient temperatures, and HS. Mohammed et al. (2014) observed a daily increase in RuT with a maximum 12.7 h post-feeding and a subsequent reduction until the next meal. Gaughan et al. (2002) described that core body temperature may peak up to 5 h after the hottest part of the day, normally between 2000 and 2200 h. These observations agree with our experiment where lower RuT was observed in the morning before feeding, then RuT had a peak around 2100 h, approximately 5 h after the peak THI (1600 h).

In summary, the results of this experiment demonstrated that feeding a corn-based TMR ameliorated HS responses and improved MP in dairy cows. Milk production was also improved with SBA treatment to wheat, and determination of RuT allowed the prediction of physiological and productive responses to HS in lactating dairy cows.