Feed Clusters According to In Situ and In Vitro Ruminal Crude Protein Degradation

Simple Summary The objective of the present study was to assess the suitability of an enzymatic laboratory method to estimate ruminal protein degradation. In situ data were used as reference. Appropriate in vitro methods are important to overcome methodological and ethical shortcomings, associated with the use of experimental animals. A cluster analysis was performed on the basis of differences between in vitro and in situ protein degradation. Among the 40 feedstuffs we tested, this difference was lowest in legume grains and highest in cereal by-products and barley. The feedstuffs clustered unspecific, not relatable to nutrient composition, origin or treatment. However, it was often obvious that additional carbohydrate-degrading enzymes should be used to assist the laboratory method, based solely on protease, to make it more conform with the in situ reference data. Abstract Effective degradation (ED) of crude protein (CP) was estimated in vitro at 0.02, 0.05 and 0.08 h−1 assumed ruminal passage rates for a total of 40 feedstuffs, for which in situ ED was available and used as reference degradation values. For this, the Streptomyces griseus protease test was used. The differences between in vitro CP degradation and the in situ CP degradation values were lowest in legume grains and highest in cereal by-products and barley. The differences between in situ and in vitro ED were expressed using a degradation quotient (degQ), where degQ = (EDin vitro − EDin situ)/EDin situ. Among the tested feedstuffs, eight specific clusters were identified according to degQ for the assumed passage rates. The feedstuffs clustered in an unspecific way, i.e., feedstuffs of different nutrient composition, origin or treatment did not necessarily group together. Formaldehyde–treated rapeseed meal, soybean meal, wheat, a treated lupin, sunflower meal and barley could not be assigned to any of the clusters. Groupwise degradation (range of degQ for assumed passage rates are given in brackets) was detected in grass silages (−0.17, −0.11), cereal by-products together with sugar beet pulp (−0.47, −0.35) and partly in legume grains (−0.14, 0.14). The clustering probably based on different specific nutrient composition and matrix effects that influence the solubility of feed protein and limit the performance of the protease. The matrix can be affected by treatment (chemically, thermally or mechanically), changing the chemical and physical structure of the protein within the plant. The S. griseus protease test had reliable sensitivity to reflect differences between native feedstuffs and treatments (thermally or chemically) that were found in situ. The in situ results, however, are mostly underestimated. The clustering results do not allow a clear conclusion on the groupwise or feed-specific use of carbohydrate-degrading enzymes as pre- or co-inoculants as part of the S. griseus protease test and need to be tested for its potential to make this test more conform with in situ data.


Introduction
The estimation of ruminal crude protein (CP) degradation is an essential part of feed protein evaluation, with feed protein being a limited and expensive nutrient source. Therefore, sufficient quantification of ruminal CP degradation is of high interest in order to assess nitrogen utilization efficiency of ruminant livestock to meet precisely the animal's requirement. The degradation of CP which is measured in vivo has been considered as reference. However, this method is laborious, time-consuming and associated with errors due to variation among individual animals and use of markers [1][2][3]. Therefore, in situ determination of ruminal CP degradation is widely used as reference [4]. The use of in situ data is often critically discussed, especially in terms of repeatability of the results. There is, however, a reliable methodological protocol existing that ensures repeatability [5]. This protocol recommends, for example, the correction of in situ degradation data for the microbial nitrogen that is synthesized during the incubation of feeds in the rumen [6]. For future applications, it seems worthwhile to search for methods that do not rely on cannulated animals and are potentially useful for routine analysis [7]. A purely enzymatic in vitro method is the Streptomyces griseus protease test, which was developed by Krishnamoorthy et al. [8]. Protease from the bacterial species S. griseus has a broad activity spectrum and may hydrolyze proteins (i.e., oligopeptides) up to 90% [9,10]. Its high reactivity comes from endo-and exopeptidases, especially metalloendopeptidase activity [10,11]. Licitra et al. [10] indicated the ratio of protease to true protein (TP) concentration has influence on CP degradation and standardized the method to that effect. Moreover, incubation times were referenced to type of feed and feed characteristics [10,12]. Several studies have shown close agreement between CP degradation estimated in situ or using the S. griseus protease test both in concentrates and roughages [7,[13][14][15][16]. Feed-specific degradation kinetics and effective degradation of protein (ED) at different passage rates have been barely described. A large part of feed protein is associated to carbohydrates, i.e., starch and fiber, as a kind of matrix that influences the degradation capability of proteases. Such matrix effects could be responsible for the inability of protease to degrade the entire feed protein [17][18][19].
We have evidence from a pilot study that the reliability of the S. griseus protease test estimating ED by the measure of in situ ED strongly depends on the incubated feedstuff and clusters may be defined that rely on feed or treatment characteristics [20]. From this, we conclude that the efficiency of the S. griseus protease may be influenced by matrix effects, which lead to protein degradation of feedstuffs clusters according to similar nutrient characteristics.
The objective of the present study was to assess the suitability of the S. griseus protease test for estimating ED of CP from 40 feedstuffs using the in situ test as a reference method.
Our hypothesis was that specific characteristics (e.g., nutrient content, treatment) of individual feedstuffs or groups of feedstuffs lead to a differentiation with regard to the susceptibility of the feed protein to protease, and thus, to specific clustering.

Origin of In Situ Data
Animal experiments were not part of this study because all in situ data originated from preliminary studies conducted under approval no. V319/14 TE. Ruminal CP degradation of concentrates was determined at the Institute of Animal Science, University of Hohenheim using a standardized assay [22,[25][26][27][28]. In brief, feedstuffs were incubated in the rumen of rumen-fistulated Jersey cows and three cows were used for each feedstuff. Incubations were made in polyester bags (Ankom Technology, Macedon, New York, USA) with a pore size of 50 µm (30 µm for RSM) and internal dimensions of 5 × 10 cm, 10 × 20 cm and 11 × 22 cm for a time period of up to 72 h and in case of SFM and all pea cultivars for a time period of up to 48 h. A minimum of three bags was used of each point in time and cow and contents after incubation were pooled for chemical analysis. The in situ degradation data of RSMa, native and treated RSMc and RSMd were not published yet. In situ protein degradation was expressed as a percentage of CP at each specific incubation time.

In Vitro Incubations
The S. griseus protease test was performed according to Licitra et al. [10]. The feedstuffs were ground through 1 mm sieve size using a standard laboratory sample mill. Briefly, 0.5 g were weighed in 50 mL centrifuge tubes, filled with 40 mL of borate-phosphate buffer (12.20 g NaH 2 PO 4 × H 2 O + 8.91 g Na 2 B 4 O 7 × 10 H 2 O/L with pH 6.7-6.8) and placed into a drying oven for 1 h at 39 • C as pre-incubation. The protease solution contained 0.58 U of nonspecific type XIV S. griseus protease (Sigma-Aldrich Chemie GmbH, Munich, Germany) per mL and was added after 1 h pre-incubation at a ratio of 24 U/g TP. The concentration of TP was calculated according to the Cornell Net Carbohydrate and Protein System (CNCPS) as CP minus non-protein nitrogen (fraction A) [23]. Samples of incubation time 0 h were taken immediately after pre-incubation without addition of protease solution. Subsequently, the feedstuffs were incubated in duplicate for 2, 4, 6, 8, 16 and 24 h, respectively. Afterwards, sample tubes were filtered through Whatman #41 filter circles and rinsed out with 150 mL distilled water each. The filters were air-dried overnight, and nitrogen was analyzed in the residues and blank filters using a FOSS Kjeltec TM 8400 unit (Foss GmbH, Hamburg, Germany). This procedure was repeated a minimum of four and a maximum of six times to obtain at least four replicates for each feedstuff. Concentrations of rumen undegraded protein (RUP) were determined according to Bachmann et al. [29] as follows (considering a sample weight of 0.5 g): RUP (g/kg DM) = ((N residue × 6.25 × 10)/(0.5 × DM feed )) × 10, where N resdiue is the nitrogen measured in filter residues (mg) corrected by blank filters and DM feed is the DM content of feedstuff (%). Degraded protein (% of CP) was considered the reciprocal of RUP at each specific incubation time.

Effective Protein Degradation
The following calculations were made using SAS 9.4 (SAS Institute Inc., Cary, NC, USA). In a first step, the in situ CP degradation data of the tested substrates were reanalyzed by fitting CP degradation (as % of CP) measured after 0, 2, 4, 6, 8, 16, 24, 48 and, if applicable, for 72 h of incubation to the exponential equation provided by McDonald [30] and Steingass and Südekum [31] using the MODEL procedure of SAS 9.4. To describe CP degradation, the washout protein a, which instantly disappears at time t = 0, b, which is the protein potentially degradable in the rumen, and c, which is the degradation rate of fraction b, were estimated. The possible appearance of a discrete lag phase L, at which no ruminal degradation occurs, was considered using a broken-line approach. As long as t ≥ L, CP degradation was fitted to the regression function, whereas if t < L, CP degradation was considered to be equal to a. The estimates of the lag phase were set to be greater or equal to zero; a + b was restricted to be lower or equal 100%. Note that the slightly different methodological approach led to ED estimates which were somewhat different from those previously published [22,[25][26][27][28] using partly the same feedstuffs as in the current study. In a second step, the in situ CP degradation data were corrected for the amount of microbial nitrogen present in the feed residues at each specific incubation time using the equations of Parand and Spek [6] as recommended by GfE [5]. For this, feedstuffs were grouped by roughages, concentrates and low protein concentrates (CP < 300 g/kg DM) and the specific equations provided by Parand and Spek [6] were applied. The potential contamination with microbial nitrogen over time is summarized in Table S1. In a third step, in situ CP degradation was estimated once with 72 h or 48 h maximal incubation time and once with maximal incubation time reduced to 24 h in accordance with maximal in vitro incubation time, and all estimations were repeated both without and with correction upon microbial nitrogen. The in situ dataset comprised three replicates per feed sample (i.e., three animals). The in vitro CP degradation was analyzed analogously with a maximal incubation time of 24 h. Within the in vitro dataset, outliers were identified using boxplots and eliminated. Outliers were defined as observations more far than three times of interquartile range. The in vitro dataset comprised six replicates (i.e., six runs) and a minimum of three replicates after elimination of outliers. Effective CP degradation, either in situ or in vitro, was calculated on the basis of the estimated parameters a, b, c, and L as described by Wulf and Südekum [32] for assumed ruminal passage rates of 0.02 (ED 2 ), 0.05 (ED 5 ) and 0.08 h −1 (ED 8 ). The whole investigation was conducted under the assumption that the degradation data obtained by the in situ method is the reference to which in vitro ED was compared. The SAS script used for all calculations can be obtained from the authors on request.

Chemical Analyses
Concentrations of DM, crude nutrients and detergent fiber fractions were analyzed according to the Association of German Agricultural Analytic and Research Institutes [33] using methods no. 3.1 (DM), 4.1.1 (CP), 5.1.1 B (AEE), 6.5.1 (aNDFom), 6.5.2 (ADFom) and 8.1 (CA). Neutral detergent fiber was determined after amylase treatment. Neutral detergent fiber and acid detergent fiber were expressed exclusive of residual ash. Starch was determined using the amyloglucosidase method (VDLUFA, 2012; method no. 7.2.5) similar to Grubješić et al. [25] and enzymatically according to Seifried et al. [34]. The CNCPS protein fraction A (non-protein nitrogen) was determined according to Licitra et al. [23] and in grass silages according to Higgs et al. [24]. All measurements of nitrogen were performed using the Kjeldahl method.

Statistical Analysis
Statistical analysis was performed using SAS 9.4. The effects of a correction of in situ CP degradation for percentages of microbial nitrogen contained in the feed residues and a reduction of maximal incubation time (t = 72 h/48 h or t = 24 h) on ED 2 , ED 5 and ED 8 were tested using the NPAR1WAY procedure and Kruskal-Wallis test. Differences between in situ and in vitro estimates of ED 2 , ED 5 or ED 8 for both 24 and 72 h maximal incubation times (in situ) were tested using pooled t-test or the Satterthwaite approximation of the t-test if applicable according to folded F-test. In cases where Gaussian distribution of the studentized residuals was not given, we used the Wilcoxon rank sum test with the NPAR1WAY procedure. Differences between in situ and in vitro estimates of ED 2 , ED 5 or ED 8 in native or treated feed samples were tested by pooled t-test. Homogeneity of sample variances and Gaussian distribution of the studentized residuals were confirmed. For all tests, statistical significance was given with p < 0.05. To compare effective CP degradation between the two different methodologies (in situ vs. in vitro), differences between in situ and in vitro ED 2 , ED 5 and ED 8 were additionally expressed as a degradation quotient (degQ). The degQ was calculated as follows: Clustering was examined by single linkage method separately for every feedstuff and for assumed ruminal passage rates of 0.02 h −1 , 0.05 h −1 and 0.08 h −1 including degQ or including only the concentrations of crude nutrients, detergent fibers and starch. Missing data of starch concentration in feedstuffs were provided by DLG [35]. Grass silages, SBP, native and treated RSMc and RSMd did not contain starch and soybeans contained 58 g starch/kg DM. A dendrogram was created showing the clusters.

Results
The estimated amount of microbial nitrogen present in feed residues is listed in Table S1. Microbial nitrogen was highest in grass silages (55.8-58.5% of total nitrogen) and lowest in the native lupin Boregine (5.1% of total nitrogen) after 72 h incubation time. The effects of a correction for microbial nitrogen contamination and a reduction of maximal incubation time (72 h or 48 h to 24 h) on in situ ED 2 , ED 5 and ED 8 are shown in Table S2 and Figure S1. Irrespective of the incubation time (72 h or 48 h or 24 h), correction for microbial nitrogen elevated in situ ED which reached significance in wheat grain with up to 2% points, in RSMa (expander-treated) and in RSMb (over-toasted) with up to 5% points and in all grass silages with up to 6% points (p < 0.05). The reduction of incubation time to maximal 24 h had mostly no or merely a small effect on in situ ED; only in RSMd (formaldehyde-treated and nitrogen-uncorrected), ED 2 was reduced by a maximal of 11% points. However, reduction of incubation time from 72 h or 48 h to 24 h affected ED of protein and comparison to in vitro ED (Table S4) far less than the correction for microbial nitrogen contamination (Table S2). On that basis, in situ CP degradation over maximal 72 h corrected for microbial nitrogen contaminations was used as reference for comparison of in vitro results reported in the following (Table 3).
Reliable estimation of ED in vitro was not possible in case of faba beans and corn due to implausible estimates of CP degradation parameters (Table S3). Effective CP degradation was mainly underestimated using the S. griseus protease test (by maximal 48% points; p < 0.05; Table 3), which is shown by negative quotients (Table 4). Only in the treated lupins Boregine and Boruta, the pea Navarro and the SFM at a passage rate of 0.08 h −1 and in the native lupins Boregine and Boruta and SBM at 0.05 and 0.08 h −1 passage rates, ED was higher in vitro than in situ ( Table 3). Regardless of ruminal passage rate, the largest differences between in situ and in vitro CP degradation were found in barley grains and industrial by-products (DDGS, CGF, wheat bran and SBP). In these feedstuffs, a, b or c were underestimated up to 53% points by the in vitro method (p < 0.05). Significant differences between in situ and in vitro estimates were also found in oilseeds following fat extraction and other processing processes (p < 0.05). In legume grains (with exception of faba beans), in situ and in vitro estimates agreed well, although some differences were significant (p < 0.05).  The calculated degQ for the assumed passage rates showed that lupins and pea grains had nearly degQ of 0, whereas by-products and barley grain had the lowest degQ of less than −0.50 (Table 4).
In Figure 3, all feedstuffs were arranged in the same order as in Figure 2. A total of eight clusters were identified and were delimited by dashed lines (cluster 1: soybeans and expander treated RSM (RSMc); cluster 2: native variants of RSMc and RSMd; cluster 3: all grass silages and expander-treated RSM (RSMa); cluster 4: pea varieties Navarro and Astronaute; cluster 5: pea variety Hardy and treated variant of lupin Boregine; cluster 6: native variants of lupin Boruta and Boregine; cluster 7: wheat bran and DDGS; cluster 8: SBP and CGF) (Figure 2, Table S5).
Within the in situ dataset, significant differences in ED were found between native and treated feedstuffs in most of the comparisons ( Figure S3). These differences between native and treated feedstuffs were likewise obtained using the S. griseus protease test (Figure 4).   In Figure 3, all feedstuffs were arranged in the same order as in Figure 2. A total of eight clusters were identified and were delimited by dashed lines (cluster 1: soybeans and expander treated RSM (RSMc); cluster 2: native variants of RSMc and RSMd; cluster 3: all grass silages and expander-treated RSM (RSMa); cluster 4: pea varieties Navarro and Astronaute; cluster 5: pea variety Hardy and treated variant of lupin Boregine; cluster 6: native variants of lupin Boruta and Boregine; cluster 7: wheat bran and DDGS; cluster 8: SBP and CGF) (Figure 2, Table S5).  Within the in situ dataset, significant differences in ED were found between native and treated feedstuffs in most of the comparisons ( Figure S3). These differences between native and treated feedstuffs were likewise obtained using the S. griseus protease test (Figure 4).

Discussion
Protease from S. griseus has widely been used for estimation of ruminal CP degradation [7,13,16,36]. As Edmunds et al. [7] described, comparison among studies is difficult, because either enzyme concentration or incubation conditions differ. The standardized protocol of Licitra et al. [10] is, therefore, a basis on which the S. griseus protease test can be performed under defined conditions. In accordance with Cone et al. [36] and Cone et al. [37], in this study, we measured degradation of CP at 0, 2, 4, 6, 8, 16 and 24 h, which allowed displaying specific degradation kinetics and to compare them with in situ results. Although degradation values from in vivo studies are considered to be the best possible reference, they are almost not available. The results of the S. griseus protease test were, therefore, compared to in situ degradation values as the best available reference. It should be noted that the in situ method is associated with relevant uncertainties (i.e., microbial attachment and particle losses), which is why the results are subjected to variability and bias [38]. These limitations highlight the potential of in vitro methods in terms of standardizable, reproducible and inexpensive methods for estimating ruminal protein degradation.
As a first step, we examined the impact of nitrogen from increasing adherence of microbial biomass to the feed residues during in situ incubation [6,39], and secondly of the reduction of the incubation time on in situ predictions of ED2, ED5 and ED8. Microbial nitrogen adhering to in situ feed residues was estimated according to Parand and Spek [6]. As shown previously [6,39], microbial nitrogen contamination of feed residues during ruminal in situ incubation is substantially lower in concentrates than in roughages. We

Discussion
Protease from S. griseus has widely been used for estimation of ruminal CP degradation [7,13,16,36]. As Edmunds et al. [7] described, comparison among studies is difficult, because either enzyme concentration or incubation conditions differ. The standardized protocol of Licitra et al. [10] is, therefore, a basis on which the S. griseus protease test can be performed under defined conditions. In accordance with Cone et al. [36] and Cone et al. [37], in this study, we measured degradation of CP at 0, 2, 4, 6, 8, 16 and 24 h, which allowed displaying specific degradation kinetics and to compare them with in situ results. Although degradation values from in vivo studies are considered to be the best possible reference, they are almost not available. The results of the S. griseus protease test were, therefore, compared to in situ degradation values as the best available reference. It should be noted that the in situ method is associated with relevant uncertainties (i.e., microbial attachment and particle losses), which is why the results are subjected to variability and bias [38]. These limitations highlight the potential of in vitro methods in terms of standardizable, reproducible and inexpensive methods for estimating ruminal protein degradation.
As a first step, we examined the impact of nitrogen from increasing adherence of microbial biomass to the feed residues during in situ incubation [6,39], and secondly of the reduction of the incubation time on in situ predictions of ED 2 , ED 5 and ED 8 . Microbial nitrogen adhering to in situ feed residues was estimated according to Parand and Spek [6]. As shown previously [6,39], microbial nitrogen contamination of feed residues during ruminal in situ incubation is substantially lower in concentrates than in roughages. We calculated a maximal contamination with microbial nitrogen of 58% of total nitrogen in feed residues after 72 h of ruminal in situ incubation with lower values for concentrates (5-45% of total nitrogen) and higher values for grass silages (55%-58% of total nitrogen) as the only forage source in the current study. The correction for microbial nitrogen contamination elevated the estimated ED 2 , ED 5 and ED 8 in all tested feedstuffs. The correction for microbial nitrogen resulted in greater differences between in situ and in vitro estimated ED especially in grass silages and SBP and is, therefore, deemed to be relevant in at least some of the concentrates, and especially in roughages.
For routine applications, a simple and timely affordable in vitro test for the determination of CP degradation of feeds is required. Using S. griseus protease, incubation times of maximal 30 h are thought to be reliable for concentrates, by-products from food processing and forages [10]. Although in our re-calculation of in situ data, the reduction of incubation time in situ from 72 h or 48 h to 24 h had just a small effect on ED 2 , ED 5 and ED 8 (Table S2), we followed the assumption of Steingass and Südekum [31] and recommendations by GfE [5] that the incubation time of at least 48 h is necessary for the reliable estimation of ED.
The rumen is colonized by a diverse commensal microbiota consisting of bacteria, protozoa, and anaerobic fungi [40]. Among bacteria, the most intensively studied group of rumen microbes, Prevotella was the most dominant genus found in ruminal fluid [41,42]. They are closely associated with protein and carbohydrate degradation [42,43] and may also act cellulolytic or synergistically with other cellulolytic microorganisms [42,44]. Rate and extent of CP degradation largely depend on the proteolytic activity of ruminal microflora and feed protein composition [45], but also amylolytic and cellulolytic activities of the microbes support ruminal degradation of proteins [17,18,45,46]. This might cause the gap between in situ estimates and those obtained using S. griseus protease as a sole agent.
Degradation of feed protein is mainly influenced by its solubility. In grass silages and legume grains, soluble protein was highest, whereas it was lowest in RSM and soybeans ( Table 2). The proportion of washout protein (a) (in situ) or the protein soluble in borate-phosphate buffer (in vitro) plays an essential role especially in legume grains (with exception of faba beans). Hedqvist and Uden [47] determined the highest proportion of buffer soluble nitrogen (B1) in pea grains, lupin grains and grass silages. The protein solubility is influenced by the native protein composition, i.e., the distribution of prolamins, glutelins, albumins and globulins [48,49]. Most important, however, is the localization of proteins. The plant protein is structurally enclosed in the matrix composed of cellulose, hemicellulose and pectin or associated to starch and have to be dissolved prior to efficient CP degradation [45]. The cereal by-products (wheat bran, CGF and DDGS) are enriched by plant cell wall constituents (aleurone and pericarp). These parts of the grain comprise cell wall associated proteins [50,51]. More than 50% of total protein of wheat bran, CGF and DDGS is associated to fiber and is, therefore, less accessible to proteases [50,52]. These feedstuffs and SBP had the lowest degQ compared to the oilseed by-products (RSM, SFM and SBM). Pedersen et al. [19] found protein solubilization in DDGS to be increased by up to 31% following the addition of xylanase. The combination of xylanases and protease had the greatest potential to degrade non-starch polysaccharides, such as arabinoxylan, and release nutrients from DDGS [19]. The fiber-protein matrix can be influenced by heating and chemical treatment during food/feed processing. The large number of treatment options, i.e., the combination of time, temperature, use of water and reducing substances, results in a wide range of differently processed feedstuffs [53,54]. As a result of processing, the protein as a component of the matrix is structurally and chemically altered [54][55][56][57]. During heating and chemical treatment, the resistance to proteolysis might increase by denaturation of protein and/or formation of Maillard reaction products [54,[58][59][60][61]. Effective CP degradation of by-products was considerably underestimated in vitro compared to the in situ results. The reduced in vitro ED can be attributed to separation processing, which leads to products enriched in cell wall-associated proteins (wheat bran, DDGS and CGF) or to physical or chemical post production treatments, as in DDGS, RSM, SFM, SBM and SBP.
Proteins localized in grains (cereal grains and legume grains) are associated to starch and this may influence protein solubility by interactions between protein and starch. These proteins surround or encapsulate starch granules and are a physical barrier to starch digestion [62]. In corn, in situ starch degradation was negatively correlated with the CP concentration of the grain [34]. In the endosperm, embedding of starch granules occurs in highly variable spherical structures which depend on source, genotype and environmental conditions. This affects the vulnerability to enzymes [63][64][65]. Assoumani et al. [18] reported increased CP degradation through S. griseus protease with the addition of amylase and ß-glucanase in feeds with more than 23% starch on DM basis (corn, wheat and barley). They attributed the smaller differences between in situ and in vitro CP degradation to improved accessibility or vulnerability of protease to the protein matrix. Literature data revealed that toasting may decrease degradation of protein in lupins [26,58,59]. Bachmann et al. [66] showed, on the basis of scanning electron micrographs, that in pea grains, heat treatment led to an alteration of the protein matrix. Then, heat treatment limits proteolysis of the matrix surrounding starch granules. This is probably an effect of Maillard reactions or the inactivation of trypsin inhibitor activity [67,68].
Full fat soybeans, as another example, are characterized by large differences between in situ and in vitro ED estimates as well. Guillamón et al. [69] reported that soybeans contain high levels of trypsin inhibitors, which can inhibit protease activity. Another reason could be the high concentration of AEE influencing the activity of protease.
In forages, it was found that the S. griseus protease test had good accuracy with estimating ruminal CP degradation probably due to a high protein solubility [7,15,37]. In grass silages, specifically, proteolysis during ensiling increased the proportion of nonprotein nitrogen, whereby protein is released from the fiber matrix [7]. Despite their high solubility, the grass silages in the present study had markedly lower in vitro ED compared to in situ ED. Abdelgadir et al. [17] used fibrolytic enzymes prior to the incubation with protease, which reduced differences between in situ and in vitro degradation of CP from alfalfa and meadow hay. The grass silages mostly had a degQ between −0.1 and −0.2, which was very uniform both among variants and among ED 2 , ED 5 and ED 8 . Prospectively, this makes mathematical correction of in vitro estimates possible and does not necessarily require modification of the test.
Our hypothesis was that specific characteristics (i.e., nutrient content or treatment) of individual feedstuffs or groups of feedstuffs lead to feed clusters with regard to the susceptibility of the CP to protease. On the basis of the selected feeds, clusters could be identified with regards to the degQ at 0.02, 0.05 and 0.08 h −1 assumed ruminal passage rates. The inclusion of all degQ resulted in eight clusters (Figure 3). Other feedstuffs could not be assigned to any cluster. Some clusters were characterized by similar nutrient compositions of included feedstuffs (clusters 2, 4 and 6), others by partial (cluster 1) or very clear differences (clusters 3, 5, 7 and 8). Such differences can be attributed to treatment effects on feed protein. The clustering of degQ clearly shows diffuse assignment of untreated with treated feedstuffs in common clusters (clusters 1, 3 and 5) or individually outside of any cluster (SBM, SFM and treated lupin Boruta) ( Figure 3). Especially, the aggressive treatments of RSM could contribute to separate arrangement of over-toasted RSM (RSMb) and formaldehyde-treated RSM (RSMd). This contrasts with the cereal byproducts and SBP, which were also subjected to heat and pressure treatments, but clustered together regardless of the assumed ruminal passage rate. Cereal grains (barley and wheat) were allocated differently although they have similar nutrient composition ( Figure S2). Matrix effects and treatments of feedstuffs seem to have a decisive influence on in vitro CP degradation as well and determine whether and to what extent protease can work. Thus, matrix effects determine the feed-specific difference between in situ and in vitro CP degradation.
Clustering was tried to identify groups of feedstuffs that should be tested with specific additional enzymes. In general, most of the feedstuffs clustered diffusely of origin and treatment, resulting in clusters that were not characterized by feedstuffs with uniform nutrient composition. The feedstuffs which were arranged outside of any cluster were not characterized by uniform nutrient composition. This makes it difficult to derive clear specific recommendations for type and quantity of the additions of carbohydrate-degrading enzymes to improve the vulnerability of the feed protein in the S. griseus protease test. Groupwise degradation occurred in grass silages, cereal based-byproducts together with SBP and partly in legume grains. The addition of fiber-or cell wall-degrading enzymes seems appropriate for the above-mentioned feedstuffs with exception of legume grains, to minimize differences to in situ CP degradation. The legume grains clustered, but the low degQ showed good agreement between in situ and in vitro ED ( Figure 3). However, in the case of faba beans and corn, the addition of starch-degrading enzymes appears to be necessary to enable the estimation of effective CP degradation.
An objective of the present study was to assess the sensitivity of the S. griseus protease test displaying feed-specific treatment effects, because this is an essential requirement for feed evaluation purposes. As our results have shown, thermic, chemical and expander treatments were well discerned from the untreated materials. The differences of in situ ED ( Figure S3) were in principle reflected by ED estimated in vitro ( Figure 4). Thus, our results clearly confirmed that the sensitivity of the S. griseus protease test is reliable for evaluating specific treatment effects.

Conclusions
Results of the current study revealed that in situ CP degradation was mainly underestimated using the S. griseus protease test, probably due to associations of protein to carbohydrates. Feed characteristics such as nutrient composition or treatment did not fully explain the clustering of feedstuffs we observed with regard to differences between in situ and in vitro CP degradation. The clustering results do not allow a clear conclusion on the groupwise or feed-specific use of carbohydrate-degrading enzymes. The addition of amylolytic and/or fibrolytic enzymes or multi-enzyme mixtures as pre-or coincubation agents in the S. griseus protease test seems to be required in some cases to support starch associated and fiber-bound protein degradation. The S. griseus protease test displays effects of nutrient composition and treatment and could, therefore, become a reliable tool in routine feed evaluation.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ani13020224/s1, Figure S1: Effect of the correction for microbial nitrogen contained in the degradation residues and reduction of maximal incubation time on in situ effective crude protein degradation (ED) at 0.02 h −1 (ED 2 ), 0.05 h −1 (ED 5 ) and 0.08 h −1 (ED 8 ) assumed ruminal passage rate; Figure S2: Cluster analysis including crude nutrients, detergent fibers and starch of used feedstuffs; Figure S3: Effective crude protein degradation (ED) at 0.02 h −1 , 0.05 h −1 and 0.08 h −1 ruminal passage rate estimated in situ in native and treated feedstuffs: lupin Boregine (A), lupin Boruta (B), RSMc (C) and RSMd (D) (72 h incubation time); Table S1: Estimated microbial nitrogen in the residues of in situ crude protein degradation at the various incubation times; Table S2: Estimated parameters of ruminal in situ crude protein degradation at 72 h and 24 h incubation time considering the effect of the correction for microbial nitrogen contained in the degradation residues; Table S3: Estimated parameters of ruminal in vitro crude protein degradation at 24 h incubation time; Table S4: Comparison of in situ and in vitro estimates of effective crude protein degradation (ED) at 0.02 h −1 (ED 2 ), 0.05 h −1 (ED 5 ) and 0.08 h −1 (ED 8 ) assumed ruminal passage rate and 24 h incubation time; Table S5: Parameter describing clusters.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.