Porcine Processed Animal Protein can be used as an alternative protein in modern broiler diets without any apparent detrimental effects on the bird’s microbiome, productivity, health and welfare.

C Hughes 1

K Lawther 1

Dimonaco NJ 1

Richmond AS 2

U Lavery 2

GM Pangga 4

N Corcionivoschi 3

Huws SA 1

Professor

Sharon Huws 1✉ EmailS.Huws@qub.ac.uk

Institute for Global Food Security, School of Biological Sciences Queen’s University Belfast BT9 5DL United Kingdom

Pilgrims Europe 39 Seagoe Industrial Estate, Craigavon, Co. Armagh BT63 5QE

3 Agri-Food and Biosciences Institute 18a Newforge Lane Belfast United Kingdom

4 Faculty of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine London UK

Hughes C ^123*, Lawther K^1*, Dimonaco NJ¹, Richmond AS², Lavery U², Pangga, GM⁴, Corcionivoschi N³, Huws SA¹.

¹ Institute for Global Food Security, School of Biological Sciences, Queen’s University Belfast, BT9 5DL, United Kingdom

² Pilgrims Europe, 39 Seagoe Industrial Estate, Craigavon, Co. Armagh, BT63 5QE

³ Agri-Food and Biosciences Institute, 18a Newforge Lane, Belfast, United Kingdom

⁴ Faculty of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London, UK

C Hughes and K Lawther contributed equally to this work.

Running title

PPAP as an alternative protein for poultry

Keywords:

Protein

broilers

poultry

porcine processed animal protein

PPAP

microbiome

Correspondence: Professor Sharon Huws, Institute for Global Food Security, School of Biological Sciences, Queen’s University Belfast, BT9 5LN, United Kingdom. Email: S.Huws@qub.ac.uk

Abstract

Background

Soybean production is under heavy scrutiny from consumers and other stakeholders due to deforestation associated with soybean cultivation. Broiler chickens depend heavily on soybeans in their diet due to their high protein requirements and soybeans' high bioavailability. Alternative protein sources to soybeans are available; however, the required scale and volume of these alternatives vastly exceed their availability. Porcine Processed Animal Protein (PPAP) has the potential to bridge this gap, but little is known regarding its suitability for broiler nutrition, which is the focus of this study.

Methods

A multifactorial trial using 840 Ross 308 broilers (as hatched) was conducted, birds were split over six treatments as follows; T1 (Control) a typical Soybean Meal (SBM) based diet, T2 a copy of T1 but with reduced levels of SBM due to 5% PPAP inclusion, T3 a copy of T1 with further reduced levels of SBM and an inclusion of 10% PPAP. T4 was a zero-SBM diet, T5 was a copy of T4 but with the inclusion of 5% PPAP, and T6 was a copy of T4 with the inclusion of 10% PPAP. We monitored bird productivity, health and the caecal microbiome to evaluate the viability of feeding PPAP as an alternative.

Results

Diets containing SBM performed better than zero-SBM diets in terms of Feed Conversion Ratio (FCR), with trends toward enhanced weight gain, regardless of whether PPAP was used. The use of soybean meal negatively impacted hockburn, however, this is primarily explained by enhanced weight gain.

Microbial diversity increased with age across all treatments but did not differ significantly between dietary groups, indicating that PPAP inclusion does not negatively affect gut microbiome development. Functional analysis revealed metabolic adaptations in PPAP-fed birds, including upregulation of fatty acid degradation pathways, suggesting the microbiome maintains metabolic flexibility in response to dietary reformulation.

Conclusion

To conclude, this work demonstrates that PPAP can be effectively incorporated into broiler diets at 5% or 10% inclusion levels, displacing soybean meal, without causing detrimental effects on broiler health, welfare, or performance. However, the presence or absence of soybean meal is shown to have a greater impact on these parameters, highlighting that the industry should focus on partially replacing soybean rather than completely removing it to improve sustainability.

Introduction

Globally, there are increasing environmental, biodiversity, and welfare concerns associated with conventional poultry meat production. This is primarily driven by the high use of soybeans in poultry diets, which is often grown on land which has undergone unethical land use change, highlighting the need for novel and alternative feedstuffs (1, 2). This tension is expected to increase due to rapid population increase, whilst the Western human consumption patterns will continue to move from red to white meat due to its lower cost, adding to current pressures (3). Currently, the primary protein source in broiler feeds is soybean meal (SBM). The United States, Brazil and Argentina produce 80% of the world’s supply (4), which leaves the rest of the world vulnerable due to its dependency on imports, potentially exposing them to increased price volatility and trade disruptions. Concurrently, there is a carbon cost to importing SBM, which, alongside land-use change challenges, also poses a challenge for the industry to achieve net zero for societal benefit. These issues create food security concerns for both consumers and the industry, as well as concerns for planetary health. In addition to the economic impact of the global reliance on imported SBM, several reports have highlighted concerns regarding SBM's sustainability (5, 6). To meet monogastric animal feed production requirements, soybean production would need to double by 2050(7). Predictions suggest a 2.4% increase in crop production growth per year; however, currently, predictions are only averaging between 0.9–1.6% per year, meaning that production will fall short compared to demand (7, 8). This highlights the need to identify alternative, novel proteins to help feed the population rather than relying entirely on SBM.

SBM is widely used due to its optimal amino acid content and high digestibility, achieved through the action of the gastrointestinal tract (GIT) microbiome, particularly in the caeca (9). The caecum is the primary site of fermentation and is dominated by bacteria, with phage also present; protozoa are typically absent, though they may be present in disease states, such as coccidiosis(10). The microbiome is a crucial factor to consider when evaluating alternative protein sources, as it is often directly influenced by the diet fed, which in turn affects broiler productivity through changes in nutrient availability and digestibility. Protein fermentation (PF) can also be elevated by high consumption of indigestible proteins and increased endogenous protein losses, which can lead to gut health issues and thus affect productivity. When excess protein reaches the caeca in the absence of sufficient fermentable carbohydrates, the gut microbiome shifts from fermenting carbohydrates – the usual energy source and the primary source of volatile fatty acids (VFAs)- to utilising protein instead (9). This is particularly important to consider when investigating different proteins, because many protein sources are less digestible than soy and therefore require higher feeding levels (9). For example, research shows that broilers fed a diet containing sunflower meal, brewers' dried grain, and wheat middlings as a partial replacement for soybean meal had higher levels of Colidextribacter and Oscillibacter compared to a control of a soybean-based diet (11). Additionally, using Black Soldier Fly (BSF) to replace soybean meal further elevated levels of Rikenella and Colidextribacter spp. relative to the control, while Desulfovibrio and Lachnoclostridium were more abundant in diets containing sunflower meal, brewers' dried grain, and wheat middlings compared to those without BSF (12). Colidextribacter is often associated with fibre fermentation and, consequently, the production of short-chain fatty acids, especially butyrate, which typically enhances gut health and nutrient absorption. In addition, Colidextribacter can promote inosine production, which helps regulate inflammation and maintain gut lining integrity (11).

As a replacement for soybeans, one potential alternative protein is porcine-derived Processed Animal Proteins (PAP) due to its desirable protein content and commercial availability. Since the Bovine Spongiform Encephalopathy (BSE) outbreak in 1990, the use of PAP has been illegal in the UK (Regulation (EC) No. 999/2001) (13). However, the European Union (EU) has recently renewed its laws to allow PAP as a feed ingredient; this is only viable for monogastrics - Porcine PAP (PPAP) can be fed to poultry and vice versa, with no species-to-species consumption permitted (14). PPAP is also used globally outside the EU, especially in the United States. Nonetheless, the effects of feeding PPAP on the microbiome, bird productivity and health and welfare are largely unknown, especially under UK and EU management systems. This study aimed to investigate the effects of partially replacing SBM with PPAP in broiler diets on the GIT microbiome, productivity, health, and welfare. This research is crucial for increasing the ethical production of broilers, while ensuring countries that cannot grow soy become less vulnerable and more food secure, and simultaneously achieving net zero. Incorporating alternative, locally sourced proteins will, in turn, contribute to the circular economy and significantly reduce the reliance on soybeans in broiler diets, which have a high carbon footprint. To the authors' knowledge, no other research has investigated the influence of PPAPs on the microbiome using shotgun metagenomic sequencing.

Materials and Methods

Experimental design and Animal Housing

The Animal Welfare Ethical Review Body approved the study at the Agri-Food and Biosciences Institute (AFBI) and was conducted under the Animals Scientific Act 1986. The study was conducted in the poultry research facilities at the Veterinary Science Division at AFBI, Belfast, UK. The project consisted of 840 Ross 308 broilers (as hatched), supplied by Moy Park (39 Seagoe Industrial Estate, Portadown, Craigavon, Co. Armagh, BT63 5QE, UK). Incubation eggs were provided from a single-sourced parent flock and incubated in a commercial hatchery at Moy Park. The 840 as hatched (AH) Ross 308 broiler birds were randomly assigned to 24 pens (35 broilers/pen; 140 broilers/treatment, 6 treatments). All 24 pens shared the same airspace and were reared from 0 to 32 days of age. House temperatures were 32°C at placement (day 0) and decreased by 0.5°C per day until 22°C was reached. Perches were available for birds at all times. All birds were stocked to a maximum density of 33kg/m². At placement, the chicks were assigned wing tags for individual bird identification, enabling monitoring of weight gain throughout the trial and allowing for the random selection of sample birds. Birds were slaughtered in line with the Home Office licence.

Roslin Nutrition Ltd, based in Scotland, produced the feed. The diets were all formulated to meet the Ross 308 Nutrition Specifications (2021). The diets were fed in pellet form. Water and feed were available ad libitum. There were three stages of feed: Starter (0–10 days), Grower (11–21 days) and Finisher (21 days onwards). The trial design is multifactorial (Fig. 1), comprising T1, an industry standard formulation including SBM; T2, a copy of T1 with reduced levels of SBM and a 5% inclusion of PPAP; and T3, a copy of T1 with further reduced levels of SBM and a 10% inclusion of PPAP. T4 was an all-vegetable diet with zero-SBM present, T5 was a copy of T4 with the inclusion of 5% PPAP, and T6 was a copy of T4 with the inclusion of 10% PAP.

Fig. 1

Experimental treatments. CON1: Control diet containing Soybean Meal; PPAP5: Processed Porcine Animal Protein at 5% inclusion; PPAP10: Processed Porcine Animal Protein at 10% inclusion; CON2: Second control diet with no soybean meal.

To substitute for SBM in T4-T6, we used a combination of maize, ExtruPro™ (a blend of full-fat rapeseed and field beans produced by extrusion), sunflower, and rapemeal to meet the recommended nutrient requirements.

Health and Welfare Assessments

To determine whether incorporating PPAP into modern broiler diets has positive or negative implications on bird health, we monitored all incidences of hockburn and pododermatitis. All remaining birds were assessed for pododermatitis at day 21 and again at day 32, with hockburn evaluated at both time points. The hockburn and pododermatitis were scored using the 5-point Welfare Quality recommended scale (Welfare Quality™). Assessment of animal welfare measures for layers and broilers, 2009), where 0 = no lesion and 4 = very severe lesion. For each measure, the bird was assigned the highest score apparent on either leg or foot.

Performance Parameters

All living birds from each pen were weighed at days 0 (to balance for weight in treatment allocation), 10, 21 and 32 from the first day of placement to monitor their average daily gain (ADG). Other key performance indicators (KPIs) that were measured include feed conversion ratio (FCR) and body weight (BW). Mortality was recorded daily and was reported as a percentage of total birds placed on days 0–3, 4–10, 11–21, 22–32 and total. Mortalities were not recorded cumulatively except for the total. Feed intake was recorded on days 0–10, 10–21, 21–32, this was then used to calculate FCR which was corrected into a 2kg FCR using Eq. 1, accounting for weight influences on the overall (day 32) FCR.

Equation 1–2kg FCR conversion

Microbiome Sampling, Sequencing and Analysis

The caecal microbiome of broilers was studied to assess whether PPAP used in diets has any detrimental effects on the microbiome and, subsequently, on performance, health, and welfare. Caecal microbiome sampling from slaughtered birds took place on days 10, 21 and 32 (representing starter, grower and finisher feeding periods, respectively) from 72 birds in total (24/timepoint and 4/treatment/timepoint). Birds were randomly selected through the Research Randomiser (Research Randomizer) for sampling prior to the trial beginning by preselecting wing tag numbers, to avoid any bias when selecting birds to be sampled. After the birds were weighed, the predetermined sample birds were removed from the pen and euthanised. Specific segments of the gastrointestinal tract were identified; ileum and jejunum (small intestine), caeca and colon (large intestine). The caeca were then cut off at the ileocecal junction, and all samples were placed into pre-labelled tubes. These samples were then stored in a -80 freezer for further processing.

QIAamp® PowerFecal Pro DNA Kit (Qiagen, Germany) was used to extract DNA from caecal chyme, following the manufacturer’s instructions. A total of 75 µL of ZymoBIOMICS™ Microbial Community Standard (D6300) was used as a positive control. For each batch of extractions, a blank kit reagent was used as a negative control to monitor any potential kitome contamination. Total DNA was eluted in 50 µl of the elution buffer and stored at -80°C. The initial DNA concentration was measured using a Nanodrop ND-1000 (NanoDrop Technologies, Inc., Wilmington, US). Extracted DNA was sent for shotgun metagenomic sequencing at the Queen’s University Belfast Genomics Core Technology Unit (GCTU). The KAPA DNA HyperPlus kit was used for library preparation, the prepared libraries underwent paired-end sequencing on the Illumina NovaSeq 6000 (cart S2 300), with a read length of 150bp. Positive controls were compared at genus level to the expected theoretical ZymoBIOMICS Microbial Community Standard composition, and negative controls investigated for potential contamination.

A total of 72 DNA samples, comprising 4 samples per treatment/timepoint (at 10, 21, and 32 days), were sequenced, yielding an average of ~ 43.19 million paired-end reads per sample. FastQC reports were generated to report overall sequence quality and distribution, which reported an average Phred score of ~ Q36. Trimmomatic (v0.39) (15) was then used with the following parameters: ILLUMINACLIP: TruSeq3-PE-2:2:30:10 LEADING:3 TRAILING:3 SLIDINGWINDOW:4:20 MINLEN:50 to trim low-quality bases and adapters and identify reads with corresponding pairs (removing reads if one pair was missing). The reads were then trimmed at any site of four sequential bases within SLIDINGWINDOW if the average quality dropped below Q20. Reads reporting a quality score of < Q20 and short reads less than 50bp (after trimming) were discarded. Following this, the `cleaned’ reads from each of the samples were then mapped to the human (GCF_000001405.40_GRCh38.p14) and chicken (GCA_000002315.5_GRCg6a) reference genomes using Bowtie2 (v2.5.1) (16) with default parameters to identify the level of contamination and remove human contaminants, and host-associated DNA. Reads that aligned concordantly with references were considered contaminants and discarded. This resulted in an average of ~ 40.07m paired-end reads per sample which were then metagenomically assembled into an average of ~ 19,500 contigs per sample (counting those greater or equal to 2,500bp in length) with metaSPAdes (spades v3.15.5) (17). These contigs were then taxonomically classified using Kraken2 (v2.1.3) (18)against the `PlusPFP’ (Standard plus Refeq protozoa, fungi & plant) precomputed database (https://genome-idx.s3.amazonaws.com/kraken/k2_pluspfp_20230605.tar.gz) with default parameters. Pyrodigal (v3.0.1) (19) was used to identify and extract the protein sequences from the contigs for functional analysis with eggNOG-mapper (v2.1.12) (20) using the eggNOG 5.0 database (21). Finally, Bowtie2 was then used to map the cleaned reads back to the assembled contigs to count the proportion of reads assigned to each taxon that had been identified by Kraken2. bedtools intersect (22) (v2.31.1) was then used to report the reads that were mapped specifically to regions of the contigs identified previously by Pyrodigal as containing protein coding genes. Outputs from these tools were then combined into tab-separated files containing the complete taxonomic and functional classifications along with read mapping numbers for each sample using the MetaPont python package (https://pypi.org/project/MetaPont/). See https://github.com/TheHuwsLab/Metagenomic_Workflow for the assembly workflow.

For all samples unknown taxa names were concatenated with its last known classification e.g. g_unknown_f_Gallionellaceae and g_unknown_o_Micrococcales. Genus-level results were then normalised using the trimmed mean of M-values (TMM) (23, 24). Normalisation was performed using the limma (3.62.1) package and calcNormFactors function in the edgeR package (v4.4.0) and in RStudio (v2024.9.1.394, R v4.4.2) enabling comparison of samples considering different sequencing depths and other variations including library preparation biases. Once normalised at genus level, the family and phyla levels were summarised and the bacteria and archaea portions of the data were selected and relative abundances calculated for each of the three taxonomic levels. Functional data was also normalised using the TMM method, functional categories included Carbohydrate-Active enZYmes (CAZy) families, Kyoto Encyclopaedia of Genes and Genomes (KEGG) Enzyme numbers (EC) and KEGG pathways (25, 26). Stacked bar charts were then prepared and plotted using ggplot2 (v3.5.2). Principal component analysis was completed using the prcomp function and diversity indices were calculated using the vegan package (v2.6.8).

Statistical Analysis of Performance, Health and Welfare Parameters

All data on performance, health and welfare parameters was analysed using JMP®, Version 17, SAS Institute Inc., Cary, NC, 1989–2007. Production data (FCR, weight, and mortality) were analysed using ANOVA. Normality and homogeneity of variance assumptions were analysed, and it was found that mortality on days 3, 10, 21, and 32, Weight at Day 32, and FCR at day 10 had unequal variances. Therefore, an ANOVA with Welch’s correction was used to assess treatment effects. Fisher's LSD was used as a posthoc test to assess individual treatments. As the hockburn and pododermatitis scores are ordinal data, treatment effects on these parameters were evaluated using Kruskal-Wallis tests with a Wilcoxon test for each pair used as a post-hoc test.

All microbiome statistical analyses were conducted in RStudio (R v4.5.1, RStudio v2025.5.1.513). Alpha diversity metrics (Inverse Simpson, Shannon, and Chao1) were compared across experimental factors (timepoint, treatment, and medication status) using pairwise Wilcoxon rank-sum tests with Benjamini-Hochberg (BH) correction for multiple comparisons. Beta diversity was assessed using Bray-Curtis dissimilarity calculated on log-transformed abundance data.

Permutational multivariate analysis of variance (PERMANOVA) was performed using the adonis function in the vegan package (v2.6.8) with 999 permutations to test the effects of treatment and time. Pairwise PERMANOVA comparisons were conducted using the pairwiseAdonis package (v0.4.1), and p-values were adjusted using the BH method. Analyses were repeated for bacterial genera, archaeal genera, CAZy families, EC numbers, and KEGG pathways.

Spearman’s rank correlation was used to assess associations between microbial genera or enzyme abundances and sample weight. Microbial genus-level abundances and enzyme profiles were filtered to remove low-abundance features (genus threshold: 0.0001; enzyme threshold: 100 normalized reads), and p-values adjusted for multiple testing using the BH false discovery rate (FDR) method.

Both Linear discriminant analysis Effect Size (LEfSe) analysis was completed and LDA were completed using the lefse bioconda (v25.1.0) package ([v1.1.2], Segata et al., 2011). Default parameters were utilised and the option -o 1000000 included to “obtain more meaningful values” for the LDA score. This package includes an internal Wilcoxon test and outputs discriminative features with abs LDA score > 2.0 (27).

Results

Welfare Assessments

Pododermatitis at day 21 was significantly affected by treatment (H (5) = 26.341, p = < .0001), T2 (CON1 + PPAP5) significantly better than T5 (CON2 + PPAP5), T6 (CON2 + PPAP510) and T4 (CON2). T1 (CON1) did differ to T4 (CON2) and T6 (CON2 + PPAP10) (Supplementary excel; Tab labelled Table S1_welfare). However, T4 was significantly worse than T5. Mean ranks in order of best to worse were: T2, (285.327) T1, (297.613) T3, (319.777) T5, (340.482) T6, (359.018) T4, (400.131) (Supplementary excel; Tab labelled Table S1_welfare). Pododermatitis scores for Day 32 were not significant. Hockburn scores were significantly different (H(5) = 12.255, p = 0.0315) at Day 32 with T5 being significantly better than T2, T4, T3 and T1, Mean rank from best to worse: T5, (249.928) T6, (284.843) T2, (296.905) T4, (304.750) T3, (313.480) T1, (313.918) (Supplementary excel; Tab labelled Table S1_welfare). Table 2 shows no significant (P = < 0.05) differences between treatments for mortality at any age.

Performance Parameters

Weight

Bird weight results presented in Table 1 show there were no treatment effect at day old, showing consistency at the start of the experiment. At Day 10 T1 (CON1) was significantly heavier compared to T3 (CON1 + PPAP10), T4 (CON2) and T6 (CON2 + PPAP10), T3 (CON1 + PPAP10) was also significantly lighter than T2 (CON1 + PAP5) and T5 (CON2 + PPAP5). No other differences were seen between treatments. Results for Day 21 showed that T1 (CON1), T2 (CON1 + PPAP5), T5 (CON2 + PPAP5) was significantly heavier than T4 (CON2) and T6 (CON2 + PPAP10), T3 (CON1 + PPAP10) was significantly heavier than T6 (CON2 + PPAP10). Importantly, there were no significant differences between treatments at Day 32, which was slaughter age.

Feed Conversion Ratio

As shown in Table 1 Day 10 FCR results show there is treatment differences, T6 (CON2 + PPAP10) and T5 (CON2 + PPAP5) were significantly worse than T1-T3 (CON1, CON1 + PPAP5 and CON1 + PPAP10 respectively). T4 (CON2) was worse than T3 (CON1 + PPAP10) and T1 (CON1). T3 (CON1 + PPAP10) and T1 (CON1) had the best FCR. At Day 21 there were also significant treatment effects, T4-T6 (CON2, CON2 + PPAP5 and CON2 + PPAP1, respectively) were significantly worse than T1-T3 (CON1, CON1 + PPAP5 and CON1 + PPAP10, respectively). T2 (CON1 + PPAP5) was significantly different to all other treatments (worse than T1 (CON1) and T3, (CON1 + PPAP10) but better than T4-T6 (CON2, CON2 + PPAP5 and CON2 + PPAP1 respectively). Therefore, T3 (CON1 + PPAP10) and T1 (CON1) were again significantly better than all other treatments. The overall crop (32-day slaughter timepoint) FCR T4-T6 (CON2, CON2 + PPAP5 and CON2 + PPAP1 respectively) were significantly worse than T1-T3 (CON1, CON1 + PPAP5 and CON1 + PPAP10 respectively). This is consistent for the calculated 2kg FCR (Table 1).

Table 1

Bird performance parameters over time periods: day old, day 10, day 21 and day 32 (overall crop). Treatment: T1 (CON1); industry standard formulation including SBM, T2 (CON1 + PPAP5); a copy of T1 with reduced levels of SBM and a 5% inclusion of PPAP, T3 (CON1 + PPAP10); a copy of T1 with further reduced levels of SBM and a 10% inclusion of PPAP. T4 (CON2 + PAP5); all-vegetable diet with zero-SBM present, T5 (CON2 + PPAP5); copy of T4 with the inclusion of 5% PPAP and T6 (CON2 + PPAP10) a copy of T4 with the inclusion of 10% PAP.
Treatment	T1 – CON1	T2 – CON1 + PPAP5	T3 – CON1 + PPAP10	T4 – CON2	T5 – CON2 + PPAP5	T6 – CON2 + PPAP10	S.E.M.	P-value	Sigⁿ
Mortality (%)
Day 0–3	0.000	2.140	0.000	0.000	0.000	0.710	0.630	0.401	NS
Day 4–10	1.430	0.000	2.140	1.430	0.710	3.570	1.357	0.834	NS
Day 11–21	0.000	0.710	0.000	0.710	2.860	0.000	0.630	0.348	NS
Day 22–32	0.710	0.000	0.710	5.000	1.430	0.000	1.024	0.408	NS
Total	2.140	2.860	2.860	7.140	5.000	4.290	1.632	0.317	NS
Weight (g)
Day old	43.970	43.030	42.910	44.130	44.230	43.720	0.345	0.052	NS
Day 10	324.790^a	319.150^ab	289.660^c	300.940^bc	312.660^ab	305.010^bc	6.200	0.010	**
Day 21	1076.060^a	1047.270^a	1027.450^ab	993.300^bc	1052.480^a	955.740^c	17.688	0.002	**
Day 32	2128.390	2124.850	2281.480	2052.620	2061.310	1903.270	79.034	0.193	NS
Weight (Day 32–Day 10)	1803.600	1805.700	1991.820	1751.680	1748.650	1598.260	76.417	0.0503	NS
FCR
Day 10	1.005^c	1.028^bc	1.012^c	1.067^ab	1.071^a	1.072^a	0.013	0.002	**
Day 21	1.265^c	1.323^b	1.268^c	1.424^a	1.422^a	1.461^a	0.013	< 0.0001	***
Overall crop	1.497^b	1.501^b	1.466^b	1.703^a	1.659^a	1.680^a	0.043	0.002	**
2 kg corrected	1.474^b	1.478^b	1.415^b	1.694^a	1.647^a	1.697^a	0.049	0.001	**

NS – not significant; P < 0.05 – significant; P < 0.01 – highly significant; P < 0.0001 – very highly significant Means within rows which do not share a common superscript are significantly (P < 0.05) different.

Microbiome Analysis - Comparing treatments and timepoints

In total 72 poultry samples were sequenced (4 birds caecal samples from starter, grower and finisher periods/treatment), along with a positive and two negative kitome controls. The two kitome controls resulted in 2 and 46 paired-end reads each, 100% of which were unclassified. The Zymo mock community that served as a positive control contained 58,253,902 paired-end reads and aligned with the expected taxonomy as detailed in Supplementary Excel; Tab labelled Table S2. A single sample was removed at this stage due to the fact that it had disproportionate numbers of reads (11-fold increase) assigned to the domain Eukaryota (7,152,507) in comparison to all other samples (PN0536_0017_S10 D10 T3; representing a caecal sample from a bird on treatment 3 taken on day 10). At a domain level, all remaining samples were dominated by Bacteria with a normalised average relative abundance of 95.48%, followed by Eukaryota (2.28%), unknown (2.21%), Archaea (0.02%) and lastly Viruses (0.01%). The bacterial and archaea fractions were explored further, as well as functional annotations at the KEGG Pathway, CAZy and EC levels (Supplementary excel).

Genus-level differences were further examined using Principal Component (PC) Analysis (PCA) (Figs. 2A and 2B). The analysis revealed that samples clustered primarily by period (Fig. 2A), with D21 and D32 overlapping, whereas no apparent clustering was observed by treatment (Fig. 2B). In terms of PC contributions, Romboutsia, Gemella, and Mammaliicoccus were the strongest contributors to PC1, indicating their significant role in driving the main variation in microbiome composition. Conversely, Eikenella, Lacibacter, and Trueperella were the largest contributors to PC2, representing the secondary axis of variation. A similar pattern, where time was the primary driver of variation along PC1 and PC2, was also observed in the functional analysis (Supplementary excel; Tab labelled Figure S1). Similarly, when considering beta diversity (assessed using bray-curtis dissimilarity measure), a timepoint-driven pattern of variation was observed, with all pairwise comparisons between timepoints showing significant differences (p.adj ≤ 0.05; Supplementary excel; labelled Table S3_diversitystasts). In contrast, there were no significant differences in Shannon diversity (which accounts for both richness and evenness) or inverse Simpson diversity (which measures evenness) across timepoints (p.adj > 0.05; Supplementary Table S3_diversitystats).

Fig. 2

Principal component analysis showing effect of time (A) and effect of treatment (B) on the chicken caecal microbiome at the genus level, T1 – CON1, T2 – CON1 + PPAP5, T3 – CON1 + PPAP10, T4 – CON2, T5 CON2 + PPAP5 and T6 – CON2 + PPAP10. D10 – Day 10, D21 – Day 21, D32 – Day 32

At a phyla level Bacillota were the most abundant phyla, with an average of 91.0% relative abundance across samples, followed by Actinomycetota (4.40%) and Pseudomonadtoa (2.63%) (Supplementary excel; Tab labelled Table S4_phyla). At the genus level, Blautia and Faecalibacterium were the most abundant genera, with average relative abundances of 15.9% and 12.3%, respectively (Fig. 3, Supplementary excel; Tab labelled Table S5_genus), followed by Lachnoclostridium (5.5%) and Mediterraneibacter (5.3%) (Fig. 3). At the first time point (D10), Blautia was the most abundant genus across all treatments. Over time (by D32), treatments T1, T3, T5, and T6 showed a shift from a Blautia-dominated microbiome to one where Faecalibacterium became the most dominant genus.

Fig. 3

Stacked bar graph showing effects of treatment on bacterial and archaea at Genus level, technical replicates were averaged and to further ease visualisation the top 20 genera are shown with all other genera grouped as Other. T1 – CON1, T2 – CON1 + PPAP5, T3 – CON1 + PPAP10, T4 – CON2, T5 CON2 + PPAP5 and T6 – CON2 + PPAP10. D10 – Day 10, D21 – Day 21, D32 – Day 32.

PERMANOVA analysis further supported the finding that microbiome composition was primarily shaped by time (Table 2). Pairwise comparisons revealed significant differences (p.aj < 0.05) between D10 vs. D21, D10 vs. D32, and D21 vs. D32 in the bacterial and archaeal microbiome at genus level, as well as in KEGG pathways and EC numbers. However, no significant differences were detected in CAZy enzyme profiles (p.adj > = 0.05) (Table 2). In contrast, treatment alone did not significantly affect the microbiome at the genus level (p.adj > = 0.05) (Supplementary excel; Tab labelled Table S6_fullstats). Likewise, combined comparisons of timepoint and treatment groups also showed no significant differences (Supplementary excel; Tab labelled Table S6_fullstats).

Table 2

PERMANOVA results testing for differences in microbiome composition across timepoints and treatments. Results are shown for the genus level, KEGG pathways, CAZy enzyme families, and EC numbers. T1 – CON1, T2 – CON1 + PPAP5, T3 – CON1 + PPAP10, T4 – CON2, T5 CON2 + PPAP5 and T6 – CON2 + PPAP10. D10 – Day 10, D21 – Day 21, D32 – Day 32. Significant results where p.adj < 0.05 are underlined.
Pairs	Genus	Pathway	CAZy	EC
D10 vs D21	0.0010	0.0015	0.9540	0.0010
D10 vs D32	0.0010	0.0015	0.7230	0.0010
D21 vs D32	0.0010	0.0050	0.5760	0.0010
T1 vs T2	0.3865	0.3950	0.7230	0.1841
T1 vs T3	0.4832	0.2833	0.0094	0.0281
T1 vs T4	0.2509	0.0600	0.0100	0.0236
T1 vs T5	0.1733	0.0300	0.0038	0.0236
T1 vs T6	0.1733	0.0150	0.0038	0.0150
T2 vs T3	0.8610	0.3950	0.0086	0.6310
T2 vs T4	0.1733	0.0150	0.0086	0.0236
T2 vs T5	0.1350	0.1564	0.0038	0.0150
T2 vs T6	0.2509	0.0375	0.0038	0.1841
T3 vs T4	0.1733	0.1564	0.1746	0.0236
T3 vs T5	0.1733	0.2833	0.0105	0.0150
T3 vs T6	0.3763	0.6150	0.0086	0.2169
T4 vs T5	0.1733	0.3660	0.0825	0.2169
T4 vs T6	0.1733	0.4119	0.1638	0.1841
T5 vs T6	0.1733	0.7130	0.0314	0.2657

By day 32 (D32), LEfSe analysis revealed differences in the number of genera associated with each treatment. Treatments T2 and T3 each had only one significantly associated genus, Shewanella and Sellimonas respectively, representing the lowest number of associations among treatments (Table 3). In contrast, T4 and T5 showed the highest numbers of associated genera, with 7 and 4, respectively. In T5, three of the four genera were unique to D32 and not observed at earlier timepoints. For T4, four genera were specific to D32, while the remaining three were consistently associated at both D21 and D32.

Table 3

LEfSe results at the genus level showing taxa significantly associated with individual treatments at D32, and whether the genus was also associated at additional timepoints.
Genus	Treatment		p.adj	Additional Timepoint
g__Marvinbryantia	T1	3.61	0.016
g__Paracoccus	T1	2.13	0.045
g__Ruminococcus	T1	3.66	0.015
g__Shewanella	T2	2.04	0.005
g__Sellimonas	T3	4.00	0.007
g__Aeromonas	T4	2.01	0.033
g__Arabiibacter	T4	3.32	0.004
g__Denitrobacterium	T4	2.34	0.009
g__Eggerthella	T4	3.74	0.003
g__Gordonibacter	T4	3.43	0.002	D21
g__Mediterraneibacter	T4	4.20	0.014
g__unknown_f__Eggerthellaceae	T4	2.15	0.012	D21
g__Adlercreutzia	T5	2.61	0.046	D21
g__Anaerobutyricum	T5	3.82	0.041
g__Enterocloster	T5	3.82	0.035
g__Raoultibacter	T5	2.27	0.005
g__Mycobacterium	T6	2.26	0.032	D21
g__Romboutsia	T6	3.44	0.006

Considering the functional repertoire of the microbiomes, by the end of the trial (D32), treatments T2 and T5 had no significantly associated microbial-associated KEGG pathways (MAPs), whereas T3 and T6 each had one (fatty acid degradation and benzoate degradation, respectively; Supplementary excel; Tab labelled S7_LEfSe_function). T4 exhibited the greatest MAP diversity, with three pathways: ubiquinone and other terpenoid-quinone biosynthesis, sulfur metabolism, and a two-component signalling system. In addition, T4 was the only treatment with a significantly associated CAZy enzyme family, Glycosyl Transferase Family 66 (GT66). At the enzyme classification level, hydrolases were the most commonly associated class across treatments, with seven enzymes in total, including Exodeoxyribonuclease III and Beta-ureidopropionase. Most of these (five) were linked to T1, which also had an additional significantly associated transferase, Thymidylate synthase (FAD). Ligases were the least represented class, with only a single enzyme, DNA ligase (NAD⁺), identified in T6, which was also the sole enzyme associated with that treatment. T4, similar to T1, had six associated enzymes, the majority (four) belonging to oxidoreductases, alongside two transferases. T2 had four associated enzymes, comprising two transferases (Uroporphyrinogen-III C-methyltransferase and Biotin synthase) and two lyases (2-iminoacetate synthase and Uroporphyrinogen-III synthase). T3 displayed a broader functional spectrum, with enzymes spanning oxidoreductases, transferases, hydrolases, and lyases. In contrast, T5 had no significantly associated enzymes (Supplementary excel; Tab labelled Table S7_LEfSe_function).

Microbiome Analysis – Comparing the impact of soy content

To further investigate the impact of soy versus non-soy diets on the microbiome, Treatments 1, 2, and 3 were grouped as Soy, and Treatments 4, 5, and 6 as NoSoy. A significant difference in microbiome composition was observed over time, and treatment also had a significant effect (p.adj < 0.05; Supplementary excel; Tab labelled Table S6_fullstats). This pattern was similarly observed for pathway level and EC data, whereas no significant differences were detected in CAZy profiles when comparing the non-soy diets over time (Table 4).

Table 4

PERMANOVA results testing for differences in microbiome composition across timepoints and treatments. Results are shown for the genus level, KEGG pathways, CAZy enzyme families, and EC numbers, where treatments are grouped based on soy content (T1,T2,T3 Soy, T4,T5,T6 NoSoy). Significant results where p.adj < 0.05 are underlined.
Pairs	Genus	Pathway	CAZy	EC
Soy_D10 vs Soy_D21	0.0015	0.0014	0.9970	0.0013
Soy_D10 vs Soy_D32	0.0015	0.0014	0.4268	0.0013
Soy_D10 vs NoSoy_D10	0.0069	0.0014	0.0090	0.0139
Soy_D21 vs Soy_D32	0.0380	0.0118	0.6589	0.0013
Soy_D21 vs NoSoy_D21	0.0204	0.0014	0.0090	0.0013
Soy_D32 vs NoSoy_D32	0.0041	0.0350	0.0150	0.0330
NoSoy_D10 vs NoSoy_D21	0.0015	0.0014	0.4375	0.0013
NoSoy_D10 vs NoSoy_D32	0.0015	0.0014	0.6589	0.0013
NoSoy_D21 vs NoSoy_D32	0.0069	0.0014	0.3180	0.0046

In the LEfSe analysis, we again focused on genera showing significant differences with a log fold change of at least 2 that were maintained at D32 (Table 5). Fourteen genera met these criteria, with two, Pseudomonas and Ruminococcus, associated with soy-based diets, and the remaining twelve linked to non-soy diets. Within the soy-based diets, Ruminococcus was associated with both D21 and D32, whereas Pseudomonas was significantly associated at D32 only. Of the twelve genera associated with non-soy diets, the majority (seven) belonged to the family Eggerthellaceae. Several genera, including Geobacillus, Lachnoclostridium, Ligilactobacillus, Raoultibacter, and Romboutsia, were significantly associated with D32. In contrast, Adlercreutzia, Arabiibacter, Denitrobacterium, Eggerthella, and Mediterraneibacter were associated with both D21 and D32, while Gordonibacter was consistently associated with non-soy diets across all timepoints.

Table 5

LEfSe results at the genus level showing taxa significantly associated with Soy and Non soy diets at D32, and whether the genus was also associated at additional timepoints.
Genus	Treatment	LDA score	p.adj	Additional Timepoint
g__Pseudomonas	Soy	2.68	0.0282
g__Ruminococcus	Soy	3.57	0.0002	D21
g__Adlercreutzia	NoSoy	2.48	0.0015	D21
g__Arabiibacter	NoSoy	3.16	0.0001	D21
g__Denitrobacterium	NoSoy	2.16	0.0002	D21
g__Eggerthella	NoSoy	3.57	0.0001	D21
g__Geobacillus	NoSoy	2.54	0.0094
g__Gordonibacter	NoSoy	3.30	0.0000	D21 D10
g__Lachnoclostridium	NoSoy	3.71	0.0377
g__Ligilactobacillus	NoSoy	3.30	0.0153
g__Mediterraneibacter	NoSoy	3.98	0.0005	D21
g__Raoultibacter	NoSoy	2.15	0.0001
g__Romboutsia	NoSoy	3.32	0.0003
g__Simiaoa	NoSoy	2.74	0.0377	D10

Regarding functional analysis at the pathway level, fatty acid metabolism and fatty acid biosynthesis (map01212 and map00061) were significantly associated with soy-based diets. In contrast, non-soy diets were linked to sulfur metabolism, oxidative phosphorylation, and the pentose phosphate pathway (Supplementary excel; Tab labelled Table S8_LEfSe). LEfSe at D32 indicated that enzymes significantly associated with non-soy diets were dominated by class 1 oxidoreductases (6/9), with two kinases (2.7.11.1 and 2.7.13.3), whereas only one of the eight soy-associated enzymes was an oxidoreductase (formate dehydrogenase [NADP⁺], 1.17.1.10); the remainder comprised hydrolases, lyases, and ligases.

After considering all functional categories, we next focused specifically on microbial enzymatic genes related to protein and energy metabolism, identifying their diet associations, taxonomic contributors, and correlations with host traits (Fig. 4). Body weight and FCR were generally improved with soy inclusion in the diet as compared with no-soy regardless of PPAP inclusion. Soybean inclusion promoted fatty acid, nitrogen, and carbohydrate metabolism. Fatty acid biosynthesis was supported by long-chain-fatty-acid–CoA ligase (6.2.1.3) and acetyl-CoA carboxylase (6.4.1.2), with the former positively correlated with feed intake (Spearman ρ = 0.856; Supplementary Table S9), indicating enhanced lipid processing. Additional enzymes linked to energy and carbohydrate metabolism, formate dehydrogenase (NADP⁺, 1.17.1.10) and β-glucuronidase (3.2.1.31), were also enriched, with formate dehydrogenase positively correlated with body weight (Spearman ρ = 0.676). Within soy treatments, T3 was distinguished by enrichment of glycine reductase (1.21.4.2) and cysteine-S-conjugate β-lyase (4.4.1.13), reflecting enhanced amino acid and sulfur-linked metabolism (Fig. 4). Together these results underpin the mechanism of action of enhanced body weight and FCR following soy inclusion in the broiler diet.

Conversely, only two protein- and energy-related enzymes were significantly enriched in non-soy diets: indolepyruvate ferredoxin oxidoreductase (1.2.7.8), which links aromatic amino acid catabolism to energy metabolism and was negatively correlated with body weight in T4 (Spearman ρ = −0.808; Supplementary Table S9), and shikimate kinase (2.7.1.71), involved in aromatic amino acid biosynthesis and associated with T4 (the non-soy diet lacking PPAP). Together, these results indicate that non-soy diets primarily influenced aromatic amino acid metabolism with few enzymatic genes related to energy provision found, likely underpinning the reduced broiler growth and efficiency.

Fig. 4

Box plots of microbial enzymes significantly associated with dietary treatments. Enzymes were identified using LEfSe and are shown for each treatment (T1–T6) in D32. Plots display normalized read counts for each enzyme, highlighting differences in abundance across diets.

Finally, we explored which bacteria harboured genes encoding enzymes of interest, and how these altered over both time and treatment. In the non-soy T4 diet, shikimate kinase showed a distinct taxonomic profile, with decreased contributions from Blautia and increased contributions from Faecalibacterium, Lachnoclostridium, Mediterraneibacter, and Wansuia (Fig. 5). Under soy diets, fatty acid biosynthesis enzymatic genes were primarily carried by Clostridium, while Subdoligranulum contributed less than in non-soy treatments. Similarly, formate dehydrogenase and β-glucuronidase exhibited diet-dependent taxonomic shifts, with Faecalibacterium increasing and Subdoligranulum decreasing under soy. Treatment-specific soy effects further highlighted this functional reallocation: T1 showed enrichment of beta-ureidopropionase linked to higher Blautia and lower Faecalibacterium, whereas T2 was characterised by a Megamonas-associated signature for 2-iminoacetate synthase. Collectively, these results demonstrate that diet-driven changes in protein source and energy metabolism could be linked to shifts in which taxa carry key functions (Fig. 5).

Fig. 5

Stacked bar chart showing the average taxonomic contributions to protein- and energy-related enzymatic genes at P3. Bars represent the proportion of reads assigned to contigs from each taxon carrying the indicated functional (EC) ID within each sample. Only functional IDs significantly associated with dietary treatments are shown. For clarity, taxa contributing less than 10% of the total have been grouped as “Other.” Asterisks represent a significance association with a treatment by D32, as determined by lefse.

Discussion

The purpose of this study was to investigate how PPAP in modern broiler diets could be incorporated and to evaluate whether there are any detrimental effects on broiler health, welfare or performance, whilst investigating how the microbiome alters when broilers are fed an animal soured protein. Our main findings indicate that PPAP is a viable source of protein that can partially replace soybean meal without compromising performance parameters, health, or welfare.

Welfare Assessments

Pododermatitis is one of the biggest welfare concerns in the UK broiler industry. The lesions seen on the footpad of birds are usually superficial; however, they can cause pain and discomfort and develop into deeper acute lesions (28). Many factors can influence the incidence of Pododermatitis, include types of nutrient deficiencies such as AA – Methionine, Pantothenic Acid, Biotin, Riboflavin and Zinc, protein levels, litter moisture and type, stocking density and environmental conditions (28–31).

Pododermatitis at day 21 showed that soybeans with 5% PPAP had lower scores, indicating better outcomes, compared to treatments without soybeans. Treatments containing soybeans had numerically lower pododermatitis severity than the zero-SBM treatments. This is contrary to findings from other studies, which suggest that soybean inclusion may increase pododermatitis incidence by increasing water consumption. This is explained by the high fibre content and fermentable sugars in soybeans, which cause water retention (32). Soybeans also affect Dietary Electrolyte Balance (DEB), which is linearly related to water intake (33). However, the alternative raw materials present in the zero-SBM diets could also have a similar impact on litter moisture. Dried Distillers Grains and Solubles (DDGS) are known to contain higher levels of sodium, which can affect water intake and, in turn, litter moisture. DDGS is one of the alternative raw materials used in this research. Factors such as water consumption, bedding quantity, and litter conditions need to be evaluated to better understand these differences.

Hockburn showed a different trend compared to pododermatitis, where diets containing zero -SBM and PPAP gave better scores. One explanation for this is the weight differences; typically, heavier birds are more susceptible to hockburn (34–36). Although there was no significant difference in weight at Day 32, numerical differences and trends were observed: all zero-SBM diets were lighter than those containing soybeans. However, this does not explain why T5 (CON2 + PPAP5) had a lower, and therefore better, score than T4 (CON2), despite having the same weights. Haslam et al (2003) suggests that deficiencies in certain micronutrients (biotin, zinc, copper, molybdenum and sulphur-containing amino acids) could increase the incidence of hockburn, which may explain why T5 (CON2 + PPAP5) was better than T4 (CON2) (37).

Other studies have looked at levels of Ammoniacal Nitrogen (NH₃ -N) present in litter samples from broilers fed either all-vegetable diets or vegetable and animal protein diets, and have found that diets including animal protein showed significantly lower levels of NH_{3 –}N and higher levels of N in the all-vegetable diets (29). This is highly beneficial, as ammonia emissions are a key environmental factor in the broiler industry. However, although ammonia is not a Greenhouse Gas (GHG); it has environmentally damaging properties and is the most significant emission from poultry production(38).

Performance Parameters

Mortality was not affected by treatments at any age; this is also consistent with findings from Nagaraj et al. (2007) (29), who reported no significant difference in mortality when feeding different diets, including those with animal protein inclusions. This suggests that including PPAP in broiler diets has no impact on animal health. There was also no difference in weight at day old; this was to be expected, as the birds were from the same hatchery, setter, hatcher, and farm shed.

The first sample time point (Day 10) showed that soybean-containing diets were heavier; this is particularly interesting because feed intake was lower in these treatments, resulting in better FCRs. This could partly be due to soybeans' highly digestible, bioavailable amino acid profile and high protein content, which are well recognised in broiler nutrition (39, 40).

At subsequent sample time points, the broilers responded better to higher levels of PPAP, indicating that PPAP, when combined with soybean, becomes more beneficial later in life. However, this was not the case for birds fed zero-SBM diets or diets containing zero-SBM with high levels (10%) of PPAP; this could be a zero-soy effect, highlighting the need to reduce soybean meal rather than replace it. However, interestingly, birds fed a diet with zero-SBM and 5% PPAP grew exceptionally well, raising the question of whether PPAP could be used in zero-SBM diets at lower levels. In terms of diet formulations, when the nutritionist removes soybeans, overall protein levels generally increase. This is mainly because soybeans are naturally rich in protein with highly digestible amino acids, especially lysine. Therefore, more lysine per gram of protein is found in soybeans compared to many other plant proteins(41). When soybean is not used, alternative protein sources are required to meet nutritional requirements, especially lysine. However, these alternative raw materials are needed at higher levels to meet the equivalent digestible lysine requirements that soybean typically meets with ease, due to their lower lysine-to-protein ratio(41). Inadvertently, this increases the overall protein mass. This excess protein is generally highly undigestible and accumulates in the hindgut, causing issues with nutrient absorption as the hindgut is overpopulated with undigestible fibre fractions. Protein fermentation is typically caused by excess protein; this process leads to the production of various protein-derived metabolites, including ammonia, biogenic amines, indoles, phenols, and gases such as methane and carbon dioxide, as well as secondary compounds like lactate and succinate (42). An increase in these metabolites and gases is typically associated with gizzard erosions and poor gut health, as demonstrated in research by Qaisrani et al, (43). Although gut morphology and gizzard analysis was not investigated in this research, it could explain some of the results this work has shown.

Our data is consistent with work completed in 2004 (44), in which three levels of Meat and Bone Meal inclusion (1, 3.5, and 5%) were incorporated into diets. Diets with the highest PPAP level (5% inclusion) performed significantly better, highlighting the need for further research at higher levels of incorporation, which was a focus of our experiment (44). However, other literature comparing vegetable diets with animal protein diets found no benefit from animal protein inclusion; in fact, body weight and FCR were negatively affected by its inclusion. Conversely, this research used Poultry by-product meal as the source of animal protein, whereas our research used Porcine PAP (29). It is also essential to consider that the research above was completed in 2007; since then, there has been rapid development in broiler digestibility, genetics, and increasingly sophisticated formulations.

At Day 32, although there were numerical differences in weight with T3 (CON1 + PPAP10) being the heaviest, these results were not significant. However, when the weight increments between day 10 and day 32 were analysed, they approached significance (p = 0.0503), showing that birds fed T3 (CON1 + PPAP10) were the heaviest. FCR data showed variable results between treatments from day 10 to day 21; however, a clear trend emerged by day 32, with soybean-containing diets having lower FCR than those without soybeans, regardless of PPAP inclusion.

As mentioned earlier, the difference observed may be attributed to the exceptionally well-balanced amino acid (AA) profile of soybeans, which allows them to meet the birds’ nutritional requirements more efficiently than alternative protein sources such as sunflower meal, DDGS, or oilseed rape(31). Wu and Wang (2022) report that PPAPs are high energy content, elevated crude protein levels, and a well-balanced AA profile (45), which is similar to soybean. This suggests that PPAP could be a promising candidate for partially replacing soybean in broiler diets. However, determining its suitability for higher inclusion rates in zero-SBM diets would require further animal studies.

Overall microbiome analysis

The gut microbiome of broilers plays a critical role in performance outcomes, particularly body weight, and is therefore a key consideration when formulating diets with alternative protein sources (46, 47). Across all treatment groups, microbial diversity increased with age, a trend consistent with established findings that the broiler microbiome matures over time, becoming more complex and functionally diverse (48, 49). However, the diversity did not differ significantly across treatments, suggesting that PPAP in broiler diets does not negatively affect the gut microbiome, highlighting its potential as an alternative to soy.

By the end of the trial, T3 (CON1 + PPAP10) showed a notable increase in Sellimonas, a Gram-positive, anaerobic, non-motile genus commonly associated with high-performance broilers(50). This may be linked to its role in glucose regulation and lipid metabolism (50, 51), which aligns with the numerically higher body weights observed in T3 (CON1 + PPAP10) birds.

Treatment 4 (CON2) exhibited seven genus-level associations, four of which were unique to this group. While this does not necessarily indicate higher overall diversity, it suggests a distinct microbial signature potentially driven by the unique composition of the T4 (CON2) diet. The exclusion of soybeans in T4 (CON2) necessitated a broader range of alternative protein sources, which may have inadvertently increased the diet's cellulose and polyphenol content. Although cellulose is relatively indigestible, types of cellulose (Amorphous) have been shown to alter the caecal microbiome composition, particularly by increasing the genus Alistipes, which is known for its role in SCFA production (52, 53). This increase in diet diversity may have promoted the growth of short-chain fatty acid (SCFA)-producing bacteria and taxa involved in aromatic compound metabolism. The enrichment of genera such as Eggerthella, Gordonibacter, and Mediterraneibacter under the T4 (CON2) diet supports a microbial shift toward aromatic amino acid degradation and compensatory energy metabolism. Gordonibacter and Eggerthella are particularly notable for their roles in tryptophan and polyphenol metabolism (54), which align with the observed increase in enzymes such as indolepyruvate ferredoxin oxidoreductase. These taxa are also associated with the transformation of dietary polyphenols into bioactive metabolites, which may influence host physiology (55). Additionally, the presence of unclassified members of the Eggerthellaceae family in T4 (CON2) suggests the involvement of novel or under-characterised taxa in these metabolic processes. The increase in Denitrobacterium may reflect shifts in nitrogen metabolism (56). At the same time, Aeromonas, although often considered opportunistic, could indicate changes in gut redox balance or microbial competition (57).

Functional analysis revealed distinct metabolic adaptations in response to dietary inclusion PPAP. In T3 (CON1 + PPAP10), which included 10% PPAP, there was a notable upregulation of fatty acid (FA) degradation pathways. This is consistent with the presence of lipid-rich substrates in the diet, as PPAP contains high levels of saturated animal fats such as palmitic and stearic acids. These fats are less digestible than plant-derived oils and may have driven the selection of microbial taxa capable of metabolising saturated lipids. Consequently, the microbiome likely upregulated FA degradation pathways to compensate for the reduced availability of more readily digestible unsaturated fats, such as those found in soybean oil (41). Similarly, T6 (CON2 + PPAP10), which also contained 10% PPAP, showed enrichment in benzoate degradation pathways. This is likely a response to the higher inclusion of plant-based raw materials such as sunflower meal, rapeseed meal, and wheat, which are rich in phenolic compounds, including benzoic acid derivatives. The combination of PPAP and a chemically diverse plant matrix likely created a gut environment that favoured microbial taxa with the enzymatic capacity to degrade aromatic compounds. These findings are supported by Segura-Wang et al (58) who demonstrated that dietary diversity, particularly from plant sources, modulates microbial metabolism of aromatic compounds, including benzoate degradation. T4 (CON2) also exhibited the greatest microbial functional diversity, with enrichment in pathways such as ubiquinone and terpenoid-quinone biosynthesis, sulfur metabolism, and two-component signalling systems. These shifts likely reflect the chemically rich gut environment created by the inclusion of diverse raw materials, such as sunflower meal, rapeseed meal, and ExtruPro™. The significant association of GT66, a glycosyltransferase involved in carbohydrate modification, further supports the idea that the T4 microbiome adapted to the presence of complex plant polysaccharides, enhancing its metabolic versatility (59). These findings align with recent metagenomic studies (e.g., (60)), which demonstrates that dietary diversity and composition are major drivers of microbial community structure and function in poultry. Importantly, our results suggest that dietary reformulation, such as removing soy, can be achieved without negatively impacting microbiome development, as age remains the dominant factor shaping microbial diversity. Together, these results suggest that PPAP inclusion, especially when combined with diverse plant ingredients, can drive functional shifts in the microbiome without negatively impacting overall microbiome development. Instead, the microbiome appears to adapt by enhancing its metabolic flexibility in response to the chemical complexity of the diet.

Soy vs. No Soy microbiome analysis

To better understand the overarching trends in microbial responses, we compared soy-based diets (T1–T3 (CON1, CON1 + PPAP5 and CON1 + PPAP10, respectively)) with non-soy diets (T4–T6 (CON2, CON2 + PPAP5 and CON2 + PPAP10, respectively)). This comparison revealed the most pronounced differences in both microbial composition and functional potential. LEfSe analysis identified only two genera, Pseudomonas and Ruminococcus, as significantly associated with soy-based diets, whereas 12 genera were enriched in non-soy treatments, indicating a broader microbial response to soy exclusion. This pattern mirrors the trend observed in the overall microbiome analysis, where non-soy diets supported greater microbial diversity. A likely explanation is the increased variety of raw materials used to replace soy, which introduced a wider range of substrates and chemical compounds into the gut environment. These substrates may have supported a more functionally diverse and metabolically adaptable microbial community.

Interestingly, Gordonibacter was consistently and significantly associated with soy-fed birds across all time points. This genus is recognised for its significant role in polyphenol metabolism, particularly in the transformation of ellagic acid and related compounds into urolithins, which are bioactive metabolites with anti-inflammatory and antioxidant properties in humans (61–63). Although research on Gordonibacter in poultry is limited, its consistent presence in soy-fed birds suggests it occupies a functional niche driven by substrate availability. Soybeans are rich in isoflavones (e.g., genistein, daidzein) and galacto-oligosaccharides (GOS) such as raffinose and stachyose, which act as prebiotic substrates (64). This supports the hypothesis of substrate-dependent colonisation, in which specific dietary components promote the growth and activity of specialised microbial taxa. As such, Gordonibacter may serve as a biomarker for polyphenol metabolism capacity, prebiotic responsiveness, and urolithin production potential in poultry. These findings underscore the importance of diet composition in shaping not only microbial diversity but also functional specialisation and demonstrate that soy exclusion can be achieved without compromising microbiome development, as age remains the primary driver of microbial maturation.

Conclusion

Our study highlights the potential for PPAP to serve as a partial protein replacement in SBM diets, helping the broiler industry achieve its net-zero ambitions. Production parameters were largely unaffected, and no differences in mortality were seen, suggesting no health implications. Weights were not different between treatments, although SBM provided a benefit in FCR over zero-SBM diets, highlighting the potential application of PPAP to be used in conjunction with SBM. Regarding welfare, differences were observed between treatments at Day 21, with SBM and low PPAP inclusion levels resulting in lower pododermatitis than high PPAP inclusion levels and zero-SBM diets. However, this effect wasn’t seen later on in life. Hockburn was affected by treatments; however, this is primarily explained by weight. Microbiome analysis demonstrated that dietary inclusion of PPAP, even at 10% levels, did not compromise microbial diversity or community maturation, with age remaining the dominant factor shaping microbiome development. The microbial community exhibited functional plasticity, adapting to PPAP inclusion through upregulation of fatty acid degradation and aromatic compound metabolism pathways. These findings support the use of PPAP as a viable alternative protein source in broiler diets, particularly when combined with diverse plant-based ingredients, without negatively impacting gut microbiome health or functionality

Data Availability

The sequencing data supporting this study are available in the European Nucleotide Archive under accession number PRJEB102289. All code used for metagenomic analysis is available at: [https://github.com/TheHuwsLab/Metagenomic\_Workflow](https:/github.com/TheHuwsLab/Metagenomic_Workflow)

Electronic Supplementary Material

Below is the link to the electronic supplementary material

Supplementary Material 1

References

Leinonen I, Kyriazakis I. How can we improve the environmental sustainability of poultry production? Proceedings of the Nutrition Society. 2016;75(3):265 – 73.

CIEL. Net Zero - How Farmers Can Reduce Emissions. (Centre for Innovation Excellence in Livestock); 2022 April 2022.

Vauterin A, Steiner B, Sillman J, Kahiluoto H. The potential of insect protein to reduce food-based carbon footprints in Europe: The case of broiler meat production. Journal of Cleaner Production. 2021;320:128799.

Iji P, Toghyani M, Ahiwe E, Omede AA. Alternative sources of protein for poultry nutrition. 2017. p. 237–69.

Jia F, Peng S, Green J, Koh L, Chen X. Soybean supply chain management and sustainability: A systematic literature review. Journal of Cleaner Production. 2020;255:120254.

He R, Zhu D, Chen X, Cao Y, Chen Y, Wang X. How the trade barrier changes environmental costs of agricultural production: An implication derived from China's demand for soybean caused by the US-China trade war. Journal of Cleaner Production. 2019;227:578–88.

Ray DK, Mueller ND, West PC, Foley JA. Yield Trends Are Insufficient to Double Global Crop Production by 2050. PLoS ONE. 2013;8(6):e66428.

Md Nasir NAN, Kamaruddin SA, Zakarya IA, Aminul Islam AKM. Sustainable alternative animal feeds: Recent advances and future perspective of using azolla as animal feed in livestock, poultry and fish nutrition. Sustainable Chemistry and Pharmacy. 2022;25:100581.

Elling-Staats ML, Gilbert MS, Smidt H, Kwakkel RP. Caecal protein fermentation in broilers: a review. World's Poultry Science Journal. 2022;78(1):103–23.

10.

López-Osorio S, Chaparro-Gutiérrez JJ, Gómez-Osorio LM. Overview of Poultry Eimeria Life Cycle and Host-Parasite Interactions. Frontiers in Veterinary Science. 2020;Volume 7–2020.

11.

Özkan S, Bay V, Cömert Acar M, Yalcın S. Partial replacement of soybean with local alternative sources: effects on behavior, cecal microbiota, and intestinal histomorphometry of local chickens. Frontiers in Veterinary Science. 2024;Volume 11–2024.

12.

Benzertiha A, Kieronczyk B, Kolodziejski P, Pruszynska-Oszmalek E, Rawski M, Jozefiak D, et al. Tenebrio molitor and Zophobas morio full-fat meals as functional feed additives affect broiler chickens' growth performance and immune system traits. Poult Sci. 2020;99(1):196–206.

13.

Department for Environment FRA. Controls on animal by-products. London, United Kingdom: Department for Environment, Food & Rural Affairs (DEFRA); 2011.

14.

Meijer N, Van Raamsdonk LW, Gerrits EW, Appel MJ. The use of animal by-products in a circular bioeconomy: Time for a TSE road map 3? Heliyon. 2023.

15.

Bolger AM, Lohse M, Usadel B. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 2014;30(15):2114–20.

16.

Langmead B, Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nature Methods. 2012;9(4):357–9.

17.

Nurk S, Meleshko D, Korobeynikov A, Pevzner PA. metaSPAdes: a new versatile metagenomic assembler. Genome Res. 2017;27(5):824–34.

18.

Wood DE, Lu J, Langmead B. Improved metagenomic analysis with Kraken 2. Genome Biol. 2019;20(1):257.

19.

Larralde M. Pyrodigal: Python bindings and interface to Prodigal, an efficient method for gene prediction in prokaryotes. Journal of Open Source Software. 2022;7:4296.

20.

Cantalapiedra CP, Hernández-Plaza A, Letunic I, Bork P, Huerta-Cepas J. eggNOG-mapper v2: Functional Annotation, Orthology Assignments, and Domain Prediction at the Metagenomic Scale. Molecular Biology and Evolution. 2021;38(12):5825–9.

21.

Huerta-Cepas J, Szklarczyk D, Heller D, Hernández-Plaza A, Forslund SK, Cook H, et al. eggNOG 5.0: a hierarchical, functionally and phylogenetically annotated orthology resource based on 5090 organisms and 2502 viruses. Nucleic Acids Research. 2018;47(D1):D309-D14.

22.

Quinlan AR, Hall IM. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics. 2010;26(6):841–2.

23.

Pereira MB, Wallroth M, Jonsson V, Kristiansson E. Comparison of normalization methods for the analysis of metagenomic gene abundance data. BMC Genomics. 2018;19(1):274.

24.

Tapio M, Fischer D, Mäntysaari P, Tapio I. Rumen Microbiota Predicts Feed Efficiency of Primiparous Nordic Red Dairy Cows. Microorganisms. 2023;11(5).

25.

Kanehisa M, Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000;28(1):27–30.

26.

Cantarel BL, Coutinho PM, Rancurel C, Bernard T, Lombard V, Henrissat B. The Carbohydrate-Active EnZymes database (CAZy): an expert resource for Glycogenomics. Nucleic Acids Res. 2009;37(Database issue):D233-8.

27.

Segata N, Izard J, Waldron L, Gevers D, Miropolsky L, Garrett WS, et al. Metagenomic biomarker discovery and explanation. Genome Biol. 2011;12(6):R60.

28.

Bilgili SF, Hess JB, Blake JP, Macklin KS, Saenmahayak B, Sibley JL. Influence of bedding material on footpad dermatitis in broiler chickens. Journal of Applied Poultry Research. 2009;18(3):583–9.

29.

Nagaraj M, Wilson CAP, Hess JB, Bilgili SF. Effect of High-Protein and All-Vegetable Diets on the Incidence and Severity of Pododermatitis in Broiler Chickens. Journal of Applied Poultry Research. 2007;16(3):304–12.

30.

Mayne R. A review of the aetiology and possible causative factors of foot pad dermatitis in growing turkeys and broilers. World's Poultry Science Journal. 2005;61(2):256–67.

31.

Teague KD, Tellez-Isaias G, Chai J, Petrone-Garcia V, Vuong CN, Blanch A, et al. Dietary soy galactooligosaccharides affect the performance, intestinal function, and gut microbiota composition of growing chicks. Poultry Science. 2023;102(4):102440.

32.

Li C, Lesuisse J, Schallier S, Clímaco W, Wang Y, Bautil A, et al. The effects of a reduced balanced protein diet on litter moisture, pododermatitis and feather condition of female broiler breeders over three generations. Animal. 2018;12(7):1493–500.

33.

Borges S, Da Silva AF, Ariki J, Hooge D, Cummings K. Dietary electrolyte balance for broiler chickens under moderately high ambient temperatures and relative humidities. Poultry Science. 2003;82(2):301–8.

34.

Dawkins MS, Roberts SJ, Cain RJ, Nickson T, Donnelly CA. Early warning of footpad dermatitis and hockburn in broiler chicken flocks using optical flow, bodyweight and water consumption. Veterinary Record. 2017;180(20):499-.

35.

Hepworth PJ, Nefedov AV, Muchnik IB, Morgan KL. Early warning indicators for hock burn in broiler flocks. Avian Pathology. 2010;39(5):405–9.

36.

Haslam SM, Knowles TG, Brown SN, Wilkins LJ, Kestin SC, Warriss PD, et al. Factors affecting the prevalence of foot pad dermatitis, hock burn and breast burn in broiler chicken. British Poultry Science. 2007;48(3):264–75.

37.

Haslam SM. The development of a unitary index for the assessment of welfare in broiler chickens: University of Bristol; 2003.

38.

Vilela MdO, Gates RS, Souza CF, Teles Junior CG, Sousa FC. Nitrogen transformation stages into ammonia in broiler production: sources, deposition, transformation, and emission into the environment. Dyna. 2020;87(214):221–8.

39.

Xavier Junior MdL, Ferreira RdS, Teixeira LdV, Valentim JK, Gomes KM, Bernandes RD, et al. Metabolizable Energy and Amino Acid Digestibility of Soybean Meal from Different Sources for Broiler Chickens Supplemented with Protease. Animals. 2024;14(5):782.

40.

Ravindran V, Hew LI, Ravindran G, Bryden WL. Apparent ileal digestibility of amino acids in dietary ingredients for broiler chickens. Animal Science. 2007;81(1):85–97.

41.

INRAE FEED TABLES [Internet]. 2002. Available from: https://feedtables.com/.

42.

Qaisrani S, Van Krimpen M, Kwakkel R, Verstegen M, Hendriks W. Dietary factors affecting hindgut protein fermentation in broilers: a review. World's Poultry Science Journal. 2015;71(1):139–60.

43.

Qaisrani SN, Van Krimpen MM, Verstegen MWA, Hendriks WH, Kwakkel RP. Effects of three major protein sources on performance, gut morphology and fermentation characteristics in broilers. British Poultry Science. 2020;61(1):43–50.

44.

Bozkurt M, Basmacioglu H, Ergul M. Effect of dietary concentration meat and bone meal on broiler chickens performance. International Journal of Poultry Science. 2004;3(11):719–23.

45.

Wu N, Wang X, Mao Z, Liang J, Liu X, Xu X. Bioconversion of chicken meat and bone meal by black soldier fly larvae: Effects of straw addition on the quality and microbial profile of larval frass. Journal of Environmental Management. 2022;307:114579.

46.

Ngunjiri JM, Taylor KJM, Abundo MC, Jang H, Elaish M, KC M, et al. Farm Stage, Bird Age, and Body Site Dominantly Affect the Quantity, Taxonomic Composition, and Dynamics of Respiratory and Gut Microbiota of Commercial Layer Chickens. Applied and Environmental Microbiology. 2019;85(9):e03137-18.

47.

Lee K-C, Kil DY, Sul WJ. Cecal microbiome divergence of broiler chickens by sex and body weight. Journal of Microbiology. 2017;55(12):939–45.

48.

Ameer A, Cheng Y, Saleem F, Uzma, McKenna A, Richmond A, et al. Temporal stability and community assembly mechanisms in healthy broiler cecum. Front Microbiol. 2023;14:1197838.

49.

Yin Z, Ji S, Yang J, Guo W, Li Y, Ren Z, et al. Cecal Microbial Succession and Its Apparent Association with Nutrient Metabolism in Broiler Chickens. mSphere. 2023;8(3):e00614-22.

50.

Wang F, Feng J, Yao M, Dou L, Nan S, Pang X, et al. Dietary succinate reduces fat deposition through gut microbiota and lipid metabolism in broilers. Poultry Science. 2024;103(8):103954.

51.

Zhang Y, Liu Y, Li J, Xing T, Jiang Y, Zhang L, et al. Dietary resistant starch modifies the composition and function of caecal microbiota of broilers. Journal of the Science of Food and Agriculture. 2020;100(3):1274–84.

52.

De Maesschalck C, Eeckhaut V, Maertens L, Lange L, Marchal L, Daube G, et al. Amorphous cellulose feed supplement alters the broiler caecal microbiome. Poultry Science. 2019;98.

53.

de Sousa LS, da Silva DHL, Cardoso AR, Moreira LG, Rios DL, Ecco R, et al. Cecal microbial composition and serum concentration of short-chain fatty acids in laying hens fed different fiber sources. Brazilian Journal of Microbiology. 2025;56(1):709–22.

54.

Romo-Vaquero M, García-Villalba R, González-Sarrías A, Beltrán D, Tomás-Barberán FA, Espín JC, et al. Interindividual variability in the human metabolism of ellagic acid: Contribution of Gordonibacter to urolithin production. Journal of Functional Foods. 2015;17:785–91.

55.

García-Villalba R, Beltrán D, Frutos MD, Selma MV, Espín JC, Tomás-Barberán FA. Metabolism of different dietary phenolic compounds by the urolithin-producing human-gut bacteria Gordonibacter urolithinfaciens and Ellagibacter isourolithinifaciens. Food & Function. 2020;11(8):7012–22.

56.

Ruiz-Barrera O, Anderson RC, Hume ME, Corrales-Millan J, Castillo-Castillo Y, Corral-Luna A, et al. Short chain nitrocompounds as a treatment of layer hen manure and litter; effects on in vitro survivability of Salmonella, generic E. coli and nitrogen metabolism. Journal of Environmental Science and Health, Part B. 2017;52(1):23–9.

57.

Wang S, Liu S, Wang C, Ye B, Lv L, Ye Q, et al. Dietary Antimicrobial Peptides Improve Intestinal Function, Microbial Composition and Oxidative Stress Induced by Aeromonas hydrophila in Pengze Crucian Carp (Carassius auratus var. Pengze). Antioxidants. 2022;11(9):1756.

58.

Segura-Wang M, Grabner N, Koestelbauer A, Klose V, Ghanbari M. Genome-Resolved Metagenomics of the Chicken Gut Microbiome. Frontiers in Microbiology. 2021;Volume 12–2021.

59.

McArthur John B, Chen X. Glycosyltransferase engineering for carbohydrate synthesis. Biochemical Society Transactions. 2016;44(1):129–42.

60.

Baker M, Zhang X, Maciel-Guerra A, Dong Y, Wang W, Hu Y, et al. Machine learning and metagenomics reveal shared antimicrobial resistance profiles across multiple chicken farms and abattoirs in China. Nature Food. 2023;4(8):707–20.

61.

Rodríguez-Daza MC, Pulido-Mateos EC, Lupien-Meilleur J, Guyonnet D, Desjardins Y, Roy D. Polyphenol-Mediated Gut Microbiota Modulation: Toward Prebiotics and Further. Frontiers in Nutrition. 2021;Volume 8–2021.

62.

Barnes S, Prasain J, D'Alessandro T, Arabshahi A, Botting N, Lila MA, et al. The metabolism and analysis of isoflavones and other dietary polyphenols in foods and biological systems. Food Funct. 2011;2(5):235–44.

63.

Aglin A. Pharmacokinetics of Dietary Isoflavones. Journal of Steroids & Hormonal Science. 2013;s12.

64.

McGivern BB, Wrighton KC, Hagerman A, Borch T, Prenni J, Wilkins MJ. Putting microbial polyphenol metabolism on the map: using microbiome science to revise soil chemical paradigms. 2021.

Acknowledgements

The authors gratefully acknowledge partial financial support from the World Poultry Science Association (UK Branch) via a research award.

Funding

Author information

Authors and Affiliations

Institute for Global Food Security, School of Biological Sciences, Queen’s University Belfast, BT9 5DL, United Kingdom

Sharon A Huws, Callie Hughes, Katie Lawther, Nicholas J Dimonaco

Pilgrims Europe, 39 Seagoe Industrial Estate, Craigavon, Co. Armagh, BT63 5QE

Callie Hughes, Anne S Richmond

Agri-Food and Biosciences Institute, 18a Newforge Lane, Belfast, United Kingdom

Nicolae Corcionivoschi, Callie Hughes

Faculty of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, London, UK

Gladys M Pangga

Author Contribution

CH designed the research, conducted the animal experiment, and wrote the manuscript. KL performed the microbiome analysis, wrote the microbiome results and methods, and reviewed and edited the manuscript. NJD carried out metagenomic sequence analysis. CH and GMP performed DNA extractions and collected samples. SAH reviewed and edited the manuscript. All authors read and approved the final manuscript.

Corresponding authors

Professor Sharon Ann Huws, Institute for Global Food Security, School of Biological Sciences, Queen’s University Belfast, BT9 5LN, United Kingdom. Email: S.Huws@qub.ac.uk

Ethics Declaration

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing Interests

The authors declare no competing interests.

Supplementary Information

Supplementary Table 1: TableS1_welfare

Supplementary Table 2: TableS2_poscontrol

Supplementary Table 3: TableS3_diversitystats

Supplementary Table 4: TableS4_phyla

Supplementary Table 5: TableS5_genus

Supplementary Table 6: TableS6_fullstats

Supplementary Table 7: TableS7_LEfSe_function

Supplementary Table 8: TableS8_LEfSe

Supplementary Table 9: TableS9_correlation

Supplementary Fig. 1: FigureS1_pcafunction

Rights and permissions

Yes

Abstract

Background Soybean production is under heavy scrutiny from consumers and other stakeholders due to deforestation associated with soybean cultivation. Broiler chickens depend heavily on soybeans in their diet due to their high protein requirements and soybeans' high bioavailability. Alternative protein sources to soybeans are available; however, the required scale and volume of these alternatives vastly exceed their availability. Porcine Processed Animal Protein (PPAP) has the potential to bridge this gap, but little is known regarding its suitability for broiler nutrition, which is the focus of this study. Methods A multifactorial trial using 840 Ross 308 broilers (as hatched) was conducted, birds were split over six treatments as follows; T1 (Control) a typical Soybean Meal (SBM) based diet, T2 a copy of T1 but with reduced levels of SBM due to 5% PPAP inclusion, T3 a copy of T1 with further reduced levels of SBM and an inclusion of 10% PPAP. T4 was a zero-SBM diet, T5 was a copy of T4 but with the inclusion of 5% PPAP, and T6 was a copy of T4 with the inclusion of 10% PPAP. We monitored bird productivity, health and the caecal microbiome to evaluate the viability of feeding PPAP as an alternative. Results Diets containing SBM performed better than zero-SBM diets in terms of Feed Conversion Ratio (FCR), with trends toward enhanced weight gain, regardless of whether PPAP was used. The use of soybean meal negatively impacted hockburn, however, this is primarily explained by enhanced weight gain. Microbial diversity increased with age across all treatments but did not differ significantly between dietary groups, indicating that PPAP inclusion does not negatively affect gut microbiome development. Functional analysis revealed metabolic adaptations in PPAP-fed birds, including upregulation of fatty acid degradation pathways, suggesting the microbiome maintains metabolic flexibility in response to dietary reformulation. Conclusion To conclude, this work demonstrates that PPAP can be effectively incorporated into broiler diets at 5% or 10% inclusion levels, displacing soybean meal, without causing detrimental effects on broiler health, welfare, or performance. However, the presence or absence of soybean meal is shown to have a greater impact on these parameters, highlighting that the industry should focus on partially replacing soybean rather than completely removing it to improve sustainability.