The experiment was conducted following the code of ethics for animal experimentation with prior approval by the Ethics Committee on Animal Use of the Suranaree University of Technology, Nakhon Ratchasima, Thailand (document ID - U1-02631-2559).

The slow-growing KR chickens were produced from Leung Hang Khao sires and SUT dams. At hatch, the chickens were sexed by the vent sexing method with 75 male chicks being used for this experiment. The chicks were wing-banded, vaccinated against Marek’s disease then individually weighed. They were brooded in individual cages from 1 day of age and raised under the same environmental and nutritional conditions. All birds were individually fed ad libitum using a starter diet (21% protein), grower diet (19% protein), and finisher diet (17% protein) at 0 to 3, 4 to 6, and 7 to 10 weeks of age, respectively. The individual body weight and feed intake of the KR chickens were recorded weekly from 1 week until 10 weeks of age for RFI calculation as described by N’Dri et al [19]. The formula for RFI calculation is shown below:

where: BWG_i is the body weight gain (g) during week_i and MMW_i the metabolic weight estimated from mean body weight at week_i, i.e., (Initial body weight+Final body weight2)0.75, b₀ is the intercept, and b₁ and b₂ are partial regression coefficients.

The cumulative RFI from hatch to week_j was calculated as:

At 10 weeks of age, the chickens were divided into 2 groups: the 10 chickens with the highest RFI (HRFI) and the 10 chickens with the lowest RFI (LRFI). They were then slaughtered by stunning with electricity, bled, scalded at 60°C, de-feathered and eviscerated. The carcasses were washed then chilled at 4°C. Part of the thigh meat was used to measure meat quality.

The thigh meat samples from the chickens were vacuum-packed individually in plastic bags then placed in a chill room at 4°C for 24 h. The 75 individual thighs were chopped up then spread out in aluminum foil boxes (3×5 cm). The samples were frozen at −80°C for 24 h then freeze dried in a freeze-dryer (Alpha 2–4 LSC plus, Martin Christ Gefriertrocknungsanlagen GmbH, Osterode am Harz, Germany) operating at −80°C and 1.65 Pa for 24 h. The individual samples of freeze-dried thigh meat were ground into powder then stored in a desiccator at 37°C before FTIR spectroscopic analysis.

The nucleotides content of the individual thigh muscles was measured as described by Jayasena et al [21]. A 5-g sample was weighed, mixed with 30 mL of 0.75 M perchloric acid, homogenized for 30 s, centrifuged at 2,000×g (Thermo Fisher Scientific, Germany) at 4°C for 5 min to extract the nucleic acids then filtered through a filter paper (No.1, Whatman International Ltd., UK). The filtrate (5 mL) was analyzed using a high-performance liquid chromatography (HP 1260, Agilent Technologies, Santa Clara, CA, USA). The peaks of the individual nucleotides were identified using the retention times of standards of hypoxanthine, inosine, IMP, and AMP (Sigma-Aldrich Co., St. Louis, MO, USA), and the concentrations were calculated using the area under each peak.

The mean values and standard errors of RFI, feed intake accumulation, body weight, physicochemical properties, and flavor precursors of the thigh meat samples of KR chickens are shown in Table 2. The difference between the mean values of feed intake accumulation for the LRFI and HRFI KR chickens was significant (p = 0.00) while the difference in body weight was not significant (p = 0.15). The differences between the mean values of the physicochemical properties (ultimate pH, WHC, drip loss) of thigh meat from the LRFI and HRFI KR chickens were not significant except for the IMP content with a significant difference at p = 0.03.

These results on feed intake and body weight agreed with those of Liu et al [7], who reported that the feed intake of LRFI broiler chickens was significantly lower than that of HRFI broiler chickens, but the body weights were not significantly different. Wen et al [10] studied slow-growing chickens, reporting a significant positive correlation between RFI and total feed intake (rg = 0.62) and a non-significant correlation with body weight. Both studies suggested that different values of RFI could reduce feed intake with no change in weight gain, because the calculation of RFI takes account of the amount of feed intake for individual chickens thus providing a balance between production and the maintenance of body weight [23].

IMP was the only flavor precursor whose content was significantly different between the LFRI and HRFI chickens. In the present study, the results on the physicochemical properties did not align with those when PCA was applied to analyze the correlations using the loading and score plots. This will be explained and discussed in the section on the correlation loading of FTIR spectra with the physicochemical properties and flavor precursors of the thigh meat of LFRI and HFRI KR chickens.

The average raw spectral features of thigh meat from the LRFI and HRFI KR chickens in the fingerprint region of wave numbers from 3,000 to 900 cm⁻¹ are shown in Figure 1A and 1B. The average spectra from the LRFI and HRFI thigh meat samples revealed distinct differences in the peak heights and peak ratios of biomolecules: in the regions 2,922 and 2,854 cm⁻¹ representing lipid, 1,744 cm⁻¹ representing protein (amide I) and 1,167 cm⁻¹ representing carbohydrate and glycogen. The areas under the peaks in these regions were integrated and the percentages of the integrated area for each biomolecule are shown in Table 3. The secondary structures of protein for the amide I band region from 1,700 to 1,600 cm⁻¹ were examined by curve fitting as presented in Figure 2. The amide I band revealed that beta sheet can be defined by the regions at 1,611, 1,624 and 1,634 cm⁻¹, alpha helix at 1,644, 1,654, and 1,665 cm⁻¹, beta turn at 1,679 cm⁻¹ and antiparallel at 1,692 cm⁻¹. The percentages of each type of secondary structure from curve fitting for LRFI and HRFI chickens are compared in Table 4.

Table 3 shows that the ratios of lipid (shown by FTIR as C-H stretching), lipid ester (>C=O stretching), amide I (C=O stretching) and amide II (C-N stretching + N-H bending coupled out of face) of the thigh meat samples from the LRFI and HRFI KR chickens were significantly different with p-values of 0.00, 0.01, 0.00, and 0.04, respectively. The LRFI chicken samples also exhibited higher ratios of lipids and lipid ester but lower ratios of protein (amide I, amide II) compared with the HRFI chicken samples, while the secondary structure proteins (beta sheet, alpha helix, beta turn and antiparallel) of the HRFI and LRFI chicken samples were not significantly different with p-values of 0.66, 0.76, 0.81, and 0.97, respectively.

A higher fat deposition and lower protein content were detected in the thigh meat samples from LRFI KR chickens, possibly because when an animal receives feed more than its requirements, this source of energy will be converted to fat via lipogenesis then accumulate in the adipose tissues [24]. However, the LRFI chickens consumed feed in significantly lower quantities than the HRFI chickens, thus implying that LRFI chickens utilized energy more efficiently than the HRFI chickens. Therefore, they would need less energy for maintenance than the HRFI chickens, with the excess energy being converted to fat and stored in the body, an interpretation confirmed by reports from Yi et al [25] and Rosa et al [26]. The finding that the thigh meat samples from LRFI KR chickens had a lower protein content than those of the HRFI KR chickens may be explained by reports from Masgrau et al [27] and Yuan et al [28]. These studies reported that the intramuscular deposition of lipids was positively correlated with insulin resistance, so that some intermediates in the insulin signaling pathway were inhibited for the control of protein translation initiation, leading to a lower protein content.

No significant differences were observed in the secondary structure proteins between samples of thigh meat from LRFI and HRFI KR chickens, but these results disagreed with those from the correlation loading. A possible explanation for this will be discussed in the next subtopic.

The present study is the first to combine data from FTIR spectra relating to biomolecules, physicochemical properties and flavor precursors then use PCA for classifying LRFI and HRFI in chickens. The results of a semi-quantitative analysis are shown in Figure 3. The score plots (Figure 3, upper) focusing on PC-1 (horizontal axis) enabled the HRFI and LRFI KR chickens to be differentiated. All variables located in the outer circle region (ultimate pH, WHC, drip loss, IMP, AMP, inosine, lipid, ester bond, amide I, amide II, amide III, beta sheet, alpha helix, and beta turn) showed a significant correlation loading with RFI with a variance greater than 50%. The left-hand side of the loading plot (Figure 3, lower) shows that IMP, amide I (1,700 to 1,600 cm⁻¹), amide II (1,600 to 1,500 cm⁻¹), alpha helix (1,644, 1,654, and 1,665 cm⁻¹), carbohydrate and glycogen (1,250 to 900 cm⁻¹), ultimate pH, WHC were positively correlated with HRFI. Although beta turn (1,679 cm⁻¹) and drip loss were in the outer circle region, their positions were close to the PC-1 axis meaning that they could not be used for differentiating between the thigh meat samples from HRFI and LRFI KR chickens. However, the right-hand side of the loading plot shows that lipid (3,000 to 2,800 cm⁻¹), ester lipid (1,740 cm⁻¹), amide III (1,338 cm⁻¹) and beta sheet (1,611, 1,624, and 1,634 cm⁻¹) were positively correlated with LRFI. The location of AMP and inosine close to the PC-2 axis means that they could not be used for differentiating between the thigh meat samples from HRFI and LRFI KR chickens.

Significant loading correlations were found when PCA was applied, but the results for some variables disagreed with those from the t-test analysis. This is because PCA allows the analysis of datasets containing imprecise measurements by extracting the important information from the data then expressing this information as a set of summarizing indices known as principal components (i.e., PC1, PC2) which can reveal differences in the data matrix [29].

The results from PCA revealed some interesting relationships: the meat samples from HRFI KR chickens were associated with high levels of carbohydrate represented by glycogen. One question is why more glycogen is to be found in the thigh meat of HRFI KR chickens. Glycogen is usually stored in the muscles and liver as an energy reserve [30]. Therefore, the levels of carbohydrate in HRFI KR chicken were significantly higher than in LRFI KR chickens, demonstrating their greater energy reserve. Unfortunately, few studies have explained why HRFI chickens can store more energy than LRFI chicken, except for the studies of Iqbal et al [31] and Van Eerden et al [32], who reported that chickens with a high RFI would be more likely to respond to stress than those with a low RFI. Animals respond to stress by using energy from the breakdown of glycogen. This infers that HRFI KR chickens accumulated more glycogen for use in times of need than LRFI KR chickens.

When animals need energy, glycogen in the muscles is degraded to generate adenosine triphosphate (ATP) [33]. The ATP will then be converted into IMP via adenosine-5-diphosphate, with the degradation of AMP and IMP continuing to produce inosine [34]. This could explain the positive association between HRFI chicken meat and glycogen and IMP, but the other flavor precursors, AMP and guanosine monophosphate (GMP), were shown by PCA to have no association with either LRFI or HRFI. It is possible that storing the meat at 4°C for 24 h before analysis inhibited the processes of AMP or GMP conversion [35].

The association between the secondary proteins, amide I, amide II, and alpha helix, and high protein in HRFI could be explained by insulin resistance control when intramuscular lipids have accumulated as discussed earlier regarding the lower protein content of LRFI KR chickens (Table 3). The high content of alpha helix in the HRFI KR chickens could be explained by the work of da Silva-Buzanello et al [36]. They reported that a high scalding temperature during the chicken slaughtering process could make conformational changes in protein by decreasing the alpha helix and increasing the beta sheet contents and the proportions of beta turn in the breast meat which contains the highest levels of polyunsaturated fatty acids. It could thus be possible that, in the present study, incubating the powdered thigh meat samples at 37°C before the FTIR measurements would not have affected thigh meat samples containing low levels of fat but would have caused lipid oxidation in meat containing high levels of lipid, such as the LRFI thigh meat, leading to conformational changes and a high content of beta sheet.

In the thigh meat from LRFI KR chickens, the association with fat, amide III, led to higher levels of fat deposition because these chickens needed less energy for maintenance leading to an excess of energy, which was then converted into intramuscular fat as mentioned earlier in Table 3. The high collagen levels (amide III) in the thigh meat of LRFI KR chickens could be explained by the reports by Pincu et al [37] and Berria et al [38] who found that high levels of fat deposition in skeletal muscle were strongly associated with insulin resistance. This can result in mitochondrial dysfunction in the composition of the extracellular matrix where changes in the matrix induce greater deposition of collagen in the skeletal muscle.