Original breeding pairs of Black and Tan Brachyury (BTBR) Pah^enu2 mice were regularly backcrossed with BTBR WT mice. Heterozygous breeding pairs were used to obtain homozygous WT and homozygous Pah^enu2 (PKU) mice. Pups were weaned at postnatal day 28, and the tissue from their ear clips was collected to establish genotype with quantitative PCR (qPCR). The tissue was incubated with a lysis buffer [100 mM Tris-HCL (Sigma-Aldrich Chemie B.V., Zwijndrecht, The Netherlands) (pH 8.5), 5 mM Ethylenediaminetetraacetic acid (EDTA) (Sigma-Aldrich Chemie B.V., Zwijndrecht, The Netherlands) (pH 8.0), 200 mM NaCl (Sigma-Aldrich Chemie B.V., Zwijndrecht, The Netherlands), 2% SDS (Bio-Rad Laboratories B.V., Lunteren, The Netherlands)], and proteinase K (Merck Chemicals B.V., Amsterdam, The Netherlands) [100:1, volume:volume (v:v)] overnight. The samples were then centrifuged (13.000 rpm × 10 min), the supernatant was transferred to clean tubes, and isopropanol (Sigma-Aldrich Chemie B.V., Zwijndrecht, The Netherlands) was added (1:1). The samples were centrifuged a second time (6.000 rpm × 10 min), after which the supernatant was discarded, and the DNA samples were dried before adding 200 μl Tris-HCl-EDTA-buffer (pH 8.0) to each sample. Genetic characterization was performed using qPCR analysis with primers aimed at exon 7 of the PAH gene (forward primer: 5′ CCGTCC TGTTGCTGGCTTAC 3′, reverse primer: 5′ CAGGTGTGTACA TGGGCTTAGATC 3′), and 5 μM probes tagged with a FAM fluorophore (WT probe: CCGAGTCZZLCALTGCA) and a Yakima Yellow fluorophore (PKU probe: CCGAGTCZLLCACTGCA) obtained from Eurogentec (Seraing, Belgium), using SensiMix™ II (GC Biotech B.V., Waddinxveen, the Netherlands) with ROX. The qPCR consisted of a 1 min cycle at 60°C, a 10 min cycle at 95°C, and 40 cycles of 15 s heating at 95°C and 1 min cooling at 60°C.

The animals were housed at 21 ± 1°C on a 12-h light-dark cycle (lights on, 08:00–20:00) with ad libitum access to water and food (Altromin, Lage, Germany 1414 mod.–NL_141004). Before inclusion in the experiment, the animals were group-housed per litter and randomly assigned to one of the four experimental conditions between postnatal days 28 and 35. During the experiment, the animals were individually housed with a paper roll and nesting material and ad libitum access to food and water. The study was approved by the Institutional Animal Care and Use Committee of the University of Groningen (DEC Number: 6926A).

In total, 56 animals were used: 14 WT (7 males and females) and 42 PKU (14 per group, 7 males, and 7 females). The animals received ad libitum access to either a control (the WT and PKU high Phe groups) or (semi) Phe-restricted diet (2 experimental PKU groups) for 10 weeks. Their food intake, water intake, and body weight were monitored daily during the first week and once a week after. The fecal samples for microbiome analysis were individually collected before the start of the dietary treatment (0 weeks), after 6 weeks, and during termination (10 weeks). The behavioral assessment took place during weeks 8 and 9 of the dietary treatment. After 10 weeks, the animals were terminated and plasma samples were collected for amino acid analysis.

The basal diet used for all the interventions was AIN-93M (124 g/kg diet protein), given to the PKU high Phe and WT groups in unadjusted form. For the Phe-restricted groups, casein was reduced by 33% and 75% (Phe-restricted) and compensated by a synthetic amino acid mixture (without Phe). Due to the reduced uptake of synthetic amino acids compared with protein, an assumed protein conversion factor was taken into account, and 20% extra amino acid mixture was added at the expense of cornstarch [For details about the content of the diets, see Supplementary Table 1, which has been adapted from a study by van Vliet et al. (26)]. Diets were prepared by Research Diet Services B.V. (Wijk bij Duurstede, The Netherlands).

Tests were performed between 1 and 5 h after the start of the light phase. The following tests were used, in order, with three resting days between tests: the open field (OF) test to assess exploratory behavior, the novel object recognition (NOR) test, and spatial object recognition (SOR) test for learning and memory assessment (cognition), and the balance beam (BB) to test motor performance. Before each test, the animals were habituated in the experimental room for 2 min. For the OF, NOR, and SOR, the same square arena was used (50 × 50 × 35 cm) with a white Plexiglas floor, gray Plexiglas walls, and a checkerboard cue one of the walls. The trials were video recorded from above and later analyzed with Ethovision tracking software version 11 (Noldus Information Technology B.V., Wageningen, Netherlands) to track the movement and exploration of the different objects. During testing, dim lighting was used (10 lux in the center of the arena). After each trial, the animals were placed in their home cage, and the arena and objects were cleaned with 30% ethanol. For the balance beam, the trials were recorded from behind the beam and later manually analyzed for the time the animals needed to cross the beam and the number of correct steps. A more detailed description of the tests can be found in Bruinenberg et al. (27).

Fecal samples were collected for each mouse individually at 0, 6, and 10 weeks of treatment. At the 0 and 6 weeks of treatment, fecal samples were collected by taking fresh fecal pellets during cleaning of the cages to disturb the animals minimally. At termination (10 weeks), feces were directly collected from the lower gut. The feces were placed in Eppendorf tubes and stored at −20°C until further processing.

The fecal samples were processed for sequencing by isolating the DNA with the Qiagen, Venlo, the Netherlands PowerSoil DNA extraction kit, and DNA concentrations were adjusted to 20 ng/μl. The V4–V6 regions of the 16SrRNA gene were amplified by qPCR (forward primer: 16S-0515F-M 3′ TGYCAGCMGCCGCGGTA 3′, reverse primer: 16S-926R 5′ CCGYCAATTYMTTTRAGTTT 5′) (28) to obtain 50 ng/0.5 μg of the PCR product. The PCR product was purified using the GeneJet purification kit and adjusted to 30 ng/μL. All sequences in each sample were labeled with a unique 10 base pair barcode, after which samples were pooled and sequenced bi-directionally (2 × 300 bp) on the Illumina MiSeq platform at Genewiz (Takeley, the United Kingdom).

Raw sequencing data were demultiplexed and processed using cutadapt and QIIME2/2018.8 to remove primers and barcodes from the sequences, using the high-performance peregrine cluster of the University of Groningen (29, 30). Since the low sequence quality at the 3' end of the forward and reverse reads precluded pairing, the sequences were trimmed at base 196. The resulting forward reads were sufficient in length to analyze the highly variable V4 region but not the V5–V6 regions. The sequencing errors were identified and denoised using Deblur, resulting in amplicon sequence variants (ASV) that are inferred to have a true biological origin (i.e., originate from bacterial species) (31). Taxonomy was assigned using the RDP-Classifier v.2.12 (32) with a confidence cut-off of 0.70. A midpoint-rooted phylogenetic tree was constructed using QIIME2 v.2018.8, after which the data was imported in R version 4.0.4 (33) using qiime2R v.0.99.4 (34) and further analyzed with the Phyloseq package v.1.34.0 (35). Singletons were removed and the archaea, chloroplast, and mitochondrial sequences were filtered from the dataset. All samples were rarefied to 4,498 reads per sample using the Phyloseq package, which led to the loss of five samples divided overall groups and time points. The final set contained 2,220 taxa by seven taxonomic ranks and 148 samples.

The experimental diets were well-tolerated by all mice based on the food intake and growth of the mice. One PKU male on the control diet was euthanized during the experiment, and one female unexpectedly died in the 6th week. Postmortem macroscopic examination did not reveal obvious pathology.

The dietary interventions had the desired effects on the plasma Phe concentrations, in which the unrestricted (high Phe), semi Phe-restricted, and Phe-restricted diet were 2,160, 1,570, and 440% higher, respectively, than the 79 μmol/l in WT mice on AIN-93M diet (Figure 1E) [published in more detail by Van Vliet (26)].

A clear phenotype was observed for exploratory behavior and motor performance, whereas we observed no phenotype on learning and memory outcomes. For the open field test we observed that all PKU animals, regardless of dietary treatment, spent less time in the center compared with the WT animals [Diet; F_(3,39) = 14.93, p < 0.001; all pairwise comparisons with WT, p < 0.001]. No differences between groups were found in learning and memory, although a significant interaction between sex and diet was found for the outcome of the SOR [Diet:sex; F_(3,36) = 3.88, p = 0.017]. A separate analysis of the males and females showed an insignificant trend toward an effect of diet in the male mice [F_(3,19) = 2.85, p = 0.065]. The lack of significant differences between the groups can possibly be explained by the observation that none of the groups performed significantly different from the chance level in both the novel object recognition [WT, t₍₁₂₎ = 0.91, p = 0.190; Phe-restricted, t₍₈₎ = 0.66, p = 0.264; semi Phe-restricted, t₍₁₁₎ = 0.55, p = 0.296; PKU high Phe, t₍₉₎ = 1.16, p = 0.138], and the spatial object recognition [WT, t₍₁₂₎ = 0.37, p = 0.359; Phe-restricted, t₍₈₎ = −0.76, p = 0.766; semi Phe-restricted, t₍₁₀₎ = 0.19, p = 0.427; PKU high Phe, t₍₈₎ = 1.00., p = 0.174], indicating that the task was too hard for the mice to master. For motor performance, we again observed a PKU phenotype for the percentage of good steps on the balance beam, regardless of dietary treatment [Diet; F_(3,34) = 7.77, p < 0.001; all comparisons with WT, p = 0.002]. All behavioral results are visualized in Figures 1A–D.

The ASV richness remained stable over time, with no significant differences for each group across all time points, and no significant main or interaction effects were observed for sex (Figure 2A). At the start of the treatment (weaning), the Shannon Diversity index (Figure 2B) was similar across all groups. However, significant effects of time were observed in the high Phe [F_(2,26)= 6.79, p = 0.004] and Phe-restricted [F_(2,29)= 4.01, p = 0.029] PKU groups, both due to a change in the Shannon diversity after the start of the diet (PKU high Phe, p = 0.009 and p = 0.011; Phe-restricted, p = 0.08 and p = 0.020 for week 6 and 10, respectively). In the latter case, we observed a significant change 6 weeks after initiating the diet, remaining stable afterward. In the high Phe PKU group, we observed a significant effect [H₍₃₎ = 11.38, P = 0.010] after 6 weeks of dietary treatment, revealing an extreme reduction in the Shannon Diversity index compared with all other groups (WT, p = 0.012; Phe-restricted, p = 0.009; semi Phe-restricted, p = 0.020) (Figure 2B). This difference was still present at 10 weeks of treatment [F_(3,43) = 6.26, p = 0.001] (WT, p = 0.004; Phe-restricted, p = 0.012; semi Phe-restricted p = 0.002). Similar results were observed for the Simpson diversity index in which a decrease after the start of diet for the PKU high Phe [F_(2,29) = 4.75, p = 0.017] group (start to 6 weeks, p = 0.012; 10 weeks, p = 0.033), with a significant effect of diet after 6 weeks of treatment [H₍₃₎ = 10.77, p = 0.013]. The Post Hoc analysis showed a lower Simpson diversity index for the high Phe PKU group compared with both degrees of Phe-Restriction (Phe-restricted, p = 0.012; semi Phe-restricted p = 0.009), although we observed no significant effect at 10 weeks of treatment [H₍₃₎ = 6.25, p = 0.1] (Supplementary Figure 1A). For Pielou's evenness index, an effect of diet was observed starting from week 6 of treatment [H₍₃₎ = 15.33, p = 0.002] and still present at week 10 [H₍₃₎ = 13.32, p = 0.004], with a clear time effect in the PKU high Phe group [F_(2,27) = 10.60, p = 0.000] leading to a lower evenness at both time points compared with the start of treatment (6 weeks, p = 0.002; 10 weeks, p = 0.002) (Figure 2C). At both time points, the PKU animals on high Phe showed a lower evenness compared with the WT animals and both forms of Phe-restriction (WT, p = 0.017 and p = 0.009; Phe-restricted, p = 0.002 and p = 0.022; semi Phe-restricted, p = 0.002 and p = 0.004, for week 6 and 10, respectively). No significant differences were observed for Faith's phylogenetic diversity between dietary groups at the different time points. Still, a significant effect of time was observed in the high Phe PKU group [H₍₂₎ = 6.13, p = 0.047], but this did not result in significant post hoc comparisons after p-value correction (Supplementary Figure 1B). No significant effects or interaction effects were observed for sex in any of the alpha-diversity measures.

The PERMANOVA of the Bray-Curtis dissimilarities showed a significant interaction effect in bacterial communities between different time points and diets [F_{(6, 147)} = 1.39, p = 0.015] as well as effects of diet [F_(3,147) = 3.46, p = 0.001], and time point [F_(2,147) = 25.57, p = 0.035] by itself (Figure 3A). At the start of treatments, no differences were found between the different dietary groups. Still, we found that the bacterial communities significantly changed between time points for all dietary treatments (Figures 3B–D). After 6 weeks of treatment, differences were observed between the bacterial communities of the Phe-restricted group and all other dietary treatments (WT, p = 0.009; semi Phe-restriction, p = 0.032; and PKU high Phe, p = 0.014), the PKU high Phe mice compared to the semi Phe-restricted treatment (p = 0.047) (Figure 3C). The difference with the Phe-restricted treatment remained at 10 weeks, although this was no longer significantly different from the semi Phe-restricted animals (WT, p = 0.003; semi Phe-restriction, p = 0.06; and PKU high Phe, p = 0.024). Lastly, we observed a difference between the bacterial communities of the WT animals and bacterial communities of the semi Phe-restriction treated and control PKU mice (Semi Phe-restriction, p = 0.021; and PKU high Phe, p = 0.024), thus showing differences from all PKU groups (Figure 3D).

We examined the multivariate dispersion between groups. We observed no significant effects of dietary treatment [F_(3,144) = 0.635, p = 0.593], nor an interaction with time [F_{(11, 136)} = 1.123, p = 0.348]. We did however observe a significant effect of time [F_(2,145) = 8.0815, p = 0.001]. This effect was due to a difference between the overall dispersion at the last sampling point with the earlier time points (start vs. end, p = 0.001; 6 vs. 10 weeks, p = 0.033; start vs. 6 weeks = n.s.), suggesting that samples were more dispersed in their bacterial communities at the last sampling point (Supplementary Table 2), indicating a divergence in bacterial communities between dietary treatments over time.

Visualization of relative abundance of taxonomic phyla, families, and genera for the different experimental groups for each time point can be found in Supplementary Figure 2. To determine the statistical differences in the observed abundances, pairwise comparisons between dietary groups were made at these taxonomic levels using DESeq2. At the start of the treatment (weaning), we observed one significant difference in abundance at each taxonomic level between the Phe-restricted and semi Phe-restricted animals, which was likely caused due to a difference in the Mucispirillum genus (p = 0.018), the Deferribacteraceae family (p = 0.008) and Deferribacteres phylum (p = 0.004).

After 6 weeks of diet, we observed a reduction in Firmicutes in the protein-restricted animals compared to the PKU animals on high Phe, which was significant for the more severe Phe-restriction (p = 0.012) (Figure 4A–Phylum). At the family level, we observed a decrease in the abundance of Peptostreptococcaceae (PKU high Phe, p = 0.017; Phe-restricted, p < 0.000) and Streptococcaceae (PKU high Phe, p < 0.000; Phe-restricted, p = 0.002) in both the PKU high Phe and Phe-restricted group compared with the WT animals, whereas the abundance of Peptostreptococcaceae seemed unchanged in the semi Phe-restricted group (PKU high Phe, p = 0.042; Phe-restricted, p < 0.000). Compared with the WT, we also found a reduction of Enterococcaceae in the Phe-restricted animals (p = 0.021), which was not seen in the other PKU groups. Additionally, we observed a higher abundance of Erysipelotrichaceae in PKU animals on high Phe compared with both forms of Phe-restriction, although this only reached statistical significance compared with the semi Phe-restricted group (Phe-restricted, p = 0.051; semi Phe-restricted, p = 0.006) (Figure 4A–Family). At the genus level, we observed two significant differences between the WT and the PKU animals on high Phe, with an increase of Alloprevotella (p < 0.001) and a decrease in Romboutsia (p < 0.001) in the PKU animals. The majority of the differences were observed between the high Phe treated animals (both WT and PKU) and the Phe-restricted animals, in which we saw a decrease in Acetatifactor (both genotypes, p = 0.001), Enterococcus (WT, p = 0.016; PKU, p = 0.034), Lactococcus (WT, p = 0.006; PKU, p = 0.025), Olsenella (PKU, p = 0.034), Oscillibacter (WT, p = 0.005), and unclassified_Erysipelotrichaceae (PKU, p = 0.008) in the Phe-restricted animals. Alternatively, increased abundances in this group were found for Clostridium XlVb (WT, p = 0.016; PKU, p = 0.008), Enterorhabdus (WT, p = 0.038), Parabacteroides (both genotypes, p < 0.000), and unclassified_Desulfovibrionaceae (PKU, p = 0.011). In the semi Phe-restricted group, we saw the same reduction in the Olsenella (p = 0.012) and unclassified_Erysipelotricaceae (p = 0.003) families compared to PKU on high Phe. Still, the increase of Parabacteroides that we found in the Phe-restricted group was not present on a semi-restricted diet (p = 0.000), and the increase in unclassified_Desulfovibrionaceae did not reach statistical significance (PKU high Phe, p = 0.053). Similar to the results of Peptostreptococcaceae at the family level, we observed that the decrease of Romboutsia in the PKU groups on high Phe and the Phe-restriction compared to the WT animals was again not present on a milder protein restriction (PKU high Phe, p = 0.036; PKU Phe-restricted, p = 0.002) (Figure 4A–Genus).

After 10 weeks of diet, we no longer observed a statistical difference in Firmicutes between the dietary groups; instead, we observed a large reduction in Tenericutes and a smaller reduction in Actinobacteria in animals that were treated with the semi Phe-restricted diet, which was significant compared to the other PKU animals (High Phe, both p = 0.031; Phe-restriction, both p = 0.027) (Figure 4B–Phylum). At the family level, the differences found at 6 weeks between WT and the PKU animals on high Phe and the Phe-restricted diet were no longer present, mainly due to increased abundances in the PKU groups. However, differences were found in abundance of Enterobacteriaceae (decrease; Phe-restricted, p = 0.046), Erysipelotrichaceae (increase; high Phe, p = 0.003), and Porphyromonadaceae (increase; Phe-restricted, p = 0.003) compared with WT animals. When comparing the PKU animals on high Phe and the Phe-restricted diet, we found differences in Enterobacteriaceae (p = 0.005) and Enterococcaceae (p = 0.009), which seem to be a result of lower abundances in the Phe-restricted group. On the other hand, we observed higher abundances of both Desulfovibrionaceae (p = 0.037) and Porphyromonadaceae (p = 0.000) in the PKU Phe-restricted group, the latter being also significantly different compared to the semi Phe-restricted group (p = 0.019). We also observed an increase in Coriobacteriaceae in the Phe-restricted group, but this only reached statistical significance compared to the semi Phe-restricted animals (p = 0.004) (Figure 4B–Family). Like the family level, we observed some shifts at the genus level, although this still resulted in the major differences found between the PKU animals on the Phe-restriction compared to the WT and PKU animals on high Phe. Similar differences as at 6 weeks of treatment were found for Acetatifactor (WT and PKU on high Phe, p = 0.001), Enterococcus (PKU, p = 0.003), Enterorhabdus (WT, p = 0.000), Lactococcus (WT, p = 0.000; PKU, p = 0.017), Parabacteroides (WT, p < 0.000) and unclassified_Desulfovibrionaceae (PKU, p = 0.002), whereas Clostridium XlVb, Oscillibacter and Romboutsia no longer showed a significant difference in abundance between diets. New differences were found between Alloprevotella (WT, p < 0.000; PKU, p = 0.038), caused by an increase in the Phe-restricted group not present in the semi-restricted group (p = 0.005) and Clostridium XVIII (PKU, p = 0.017) and Intestinimonas (PKU, p = 0.009), which seemed to be due to a decrease of this genus in the animals treated with High Phe. Additionally, we observed a reduction of Escherichia/Shigella in both Phe-restricted groups, which reached statistical significance for the full restriction (WT, p = 0.035; PKU, p = 0.002). Contrary to week 6, we observed that the Olsenella genus was now more abundant in the Phe-restricted group (WT, p = 0.032) and appeared to be low to not abundant anymore in the WT and semi Phe-restricted animals. Lastly, we observed a significant decrease in Rikenella (WT, p = 0.000; PKU, p = 0.014) and an increase in both unclassified_Porphyromonadaceae (WT, p = 0.026). Between the WT animals and PKU animals on high Phe, we only observed three differences, which were an increase in the PKU group for Enterococcus (p = 0.017) and unclassified_Erysipelotrichaceae (p = 0.015), and a decrease in unclassified_Desulfovibrionaceae (p = 0.015). Interestingly, no statistical differences were observed at both family and genera levels between the PKU group on the high Phe and the semi Phe-restricted group. However, differences in average abundance can be seen visually (Figure 4B–Genus). All results are visualized in Figure 4, showing both the abundances and log2-fold changes of comparisons between dietary groups.

To link the bacterial taxonomy to the protein-restricted diet, we correlated the count data of the significant families and genera of the PKU animals with the plasma Phe levels we collected at termination (10 weeks) as an outcome measure of the efficacy of the dietary Phe-restriction (Figure 5). At the family level, we observed a significant negative relationship with Coriobacteriaceae, Desulfovibrionaceae, and Porphyromonadaceae (all comparisons, R = −0.43, p = 0.018, and a positive relationship with Enterococcaceae (R = 0.38, p = 0.027). At the genus level, we again observed mostly negative correlations (3 out of 4), specifically for Enterorhabdus (R = −0.56, p = 0.002), Intestinimonas (R = −0.42, p = 0.037), Parabacteroides (R = −0.57, p = 0.002), and one positive relationship with Enterococcus (R = 0.38, p = 0.049). A trend toward a positive relationship with Acetatifactor was also observed, but this did not reach statistical significance (R = 0.37, p = 0.053). All correlations, estimates, and unadjusted and adjusted p-values are summarized in Supplementary Table 3.

The same analysis was repeated for the behavioral parameters for both the count data at 6 and 10 weeks of treatment (Figure 6) to examine whether performance in the behavioral tasks could be related to the abundance of taxa that were significantly different between groups at their relative time points. Although our animals could not master the learning tasks (NOR and SOR) on a group level, we observed large individual variation between the animals within the same group, allowing us to examine whether specific bacterial taxa were related to better performance on these behaviors. Limited relationships were observed with the microbiome data at 6 weeks but included a negative relationship with the Alloprevotella genus (R = −0.62, p = 0.002). With the inclusion of the WT animals, a negative relationship between the Streptococcaceae (R = 0.43, p = 0.007) family and the time spent in the center of the open field was observed and still led to a significant correlation with the Alloprevotella genus (R = −0.47, p = 0.013) Correlations with microbiome data of week 10 of treatment resulted in several relationships with the performance in the SOR (time spent exploring the moving object). At the family level, we observed positive relationships between the Enterecoccaceae (R = 0.50, p = 0.017) and Erysipelotricaceae (R = 0.46, p = 0.028) families and a negative relationship with the Porphyromonadaceae family (R = −0.52, p = 0.017). The correlations between the PKU animals and taxonomic abundance are shown in Supplementary Table 3; correlations with the inclusion of the WT animals are presented in Supplementary Table 4.

The results of our study showed that untreated PKU greatly impacted the microbiome alpha diversity in terms of the Shannon Diversity Index and that PKU influenced gut bacterial community structure (beta diversity) regardless of treatment in the PKU mouse model (see Figure 7 for a schematic overview). These results are similar to previously reported findings in both humans (21, 24, 25) and mice (22), in which PKU individuals on a Phe-restricted diet have, or tend to have, a less diverse and even bacterial community that differs from either non-PKU or mild hyperphenylalaninemia subjects, although inconsistencies between studies exist. Remarkably, Phe-restriction leads to the largest change in community structure, which seems to be lessened by allowing some intake of natural protein (semi Phe-restriction). It is interesting to note that these differences seem to develop faster depending on the severity of Phe-restriction; after 6 weeks of diet, the PKU Phe-restricted animals already showed a different community structure compared with the other PKU and WT animals, whereas at the same time point, the semi Phe-restricted animals did not yet significantly differ from WT. After 10 weeks, however, all PKU animals had significantly different microbial communities compared with the WT animals. There was a difference between the PKU animals on high Phe compared with both forms of Phe-restriction. This suggests that although diet influences the microbiome composition strongly (i.e., the taxonomic differences between the WT animals and PKU animals on the high Phe diet seem to be less pronounced than those between the WT animals and Phe-restricted animals), the PKU condition still affects the community structure. However, in terms of alpha diversity, the semi Phe-restricted animals seem to be most successful in maintaining a rich and even microbiome in the context of PKU compared to the WT mice. A high(er) diversity leads to a biotic barrier, preventing the establishment and growth of (non-)pathogenic invaders from the environment. These disturbances are common and can lead to significant shifts in the community due to microbial competition if the resident microbial community is not able to outcompete the invader and return to the original (healthy) state (16, 43, 44). Therefore, from an ecological perspective, the semi Phe-restricted diet could potentially lead to an altered but more stable and resilient gut microbiome than the high Phe or Phe-restricted diets.

Of note, although our data suggest that PKU is the largest contributor to changes in alpha diversity, these changes were not evident at the start of treatment, the point up until they had been housed with their heterozygous (HTZ) mother and mixed genotype (WT, HTZ, PKU) siblings. Therefore, our data implies that either the proximity of diverse microbiomes (and potentially maternal nutrition) can prevent the effects of PKU on the microbiome regardless of high Phe levels or that these differences develop at a later time frame. It has been shown that living together has a higher impact than genetics on microbiome composition, even in humans (45). Additionally, Dill-McFarland et al. (45) also showed that people living alone have less diverse microbiomes than married individuals, suggesting that the differences are not only explained by nutritional habits. As mice are coprophagous, these effects of living together could be amplified by the recurrent inoculation of healthy microbes in social situations (46). If this is indeed the case, microbiome transplantations could potentially successfully establish a rich and diverse microbiome in PKU.

Looking at the specific members of the bacterial communities, previous studies have shown shifts in abundance at the phylum level. The most pronounced decrease in Bacteroidetes, Firmicutes, and Verrucomicrobia in children with PKU (±5 years) in the study of De Oliveira et al. (21). Although we did not observe any significant differences for any of these phyla between our Phe-restricted and WT animals, we did observe a tendency for a reduction in Firmicutes in our Phe-restricted animals. While there seems to be a general trend in the studies that show a decrease in Bacteroidetes (21, 24) in PKU subjects, remarkably, we observe higher base mean levels in the PKU Phe-restricted animals compared with our other groups, which however seems to be comparable to the new data provided by Mancilla et al. (25) which shows an increase of Bacteroidetes in the adult PKU microbiome. Similarly, the families that changed in response to treated PKU differ between studies, with substantial decreases in Clostridiaceae, Erysipelotricaceae, and Lachnospiraceae (21), Veillonellaceae and Ruminococcaceae (24), and Lactococcus in our study. Inconsistencies include an increase in Lachnospiraceae observed by Bassanini, but not by De Oliveira et al. (21), suggesting this is Phe related in young children with PKU. Lastly, we observed a decrease in Peptostreptococcaceae, contrary to the findings of De Oliveira et al. (21). Interestingly, we also show a variation between time points: although insignificant, after 6 weeks, lower base mean levels of the Erysipelotrichaceae family and the unclassified_Erysipelotrichaceae genus were observed in PKU Phe-restricted animals compared with WT, whereas after 10 weeks, the opposite was shown. Inconsistencies between these results can have numerous reasons, including the age or species of the subjects, environmental availability of microbes, and methodological differences (47, 48). Indeed, the recently published data from Mancilla et al. (25) show that the adult PKU microbiome differs from that of children, which could explain the differences observed between studies (25). At the genus level, there is considerable variability between studies as to whether particular bacteria are significantly abundant or not in response to PKU and/or the diet. However, these studies do show a consistency in the direction of change (i.e., an increase or decrease in abundance) for the same bacterial genera suggesting that these differences might be specific responses to PKU or the diet, regardless of age or host species. For example, both Mancilla et al. and De Oliveira et al. show an increase in Akkermansia in treated patients with PKU. Although we did not observe any significant changes, we did observe a vast increase at 6 weeks in both treated and untreated PKU animals, indicating that high variability could play a role in discrepancies between studies. Notably, abundances of several genera, including Bacteroides, Blautia, Lactococcus, Oscillibacter, Romboutsia, Ruminococcus, and Veillonella, remain inconsistent between studies and causes other than PKU/diet could contribute to their abundance.

Despite the differences between the human and murine gut-microbiome, our results provide insights into how different Phe contents in the diet, leading to altered blood Phe levels, relate to distinct bacterial taxa present in the microbial community. We observed that PKU itself, on a high Phe diet, is associated with an increase in Firmicutes after 6 weeks of a diet that normalizes after 10 weeks of treatment. At the family and genus level, we observe a reduction in Peptostreptococcaceae, Streptococcaceae, and Romboutsia as opposed to an increase in Alloprevotella at 6 weeks and Erysipelotrichaceae and Enterococcus at 10 weeks. Interestingly, four relationships were found with plasma Phe levels out of the eight families that showed differential abundances between diets. Three of them led to lower abundances under higher Phe levels. Out of these four relationships, we observed that Enterococcaceae and Porphyromonadaceae are related to hippocampal-dependent memory in the spatial location object tests in favor of better performance under high Phe levels. Of note, although we do not observe a (significant) correlation with Erysipelotrichaceae and Phe levels, this family is present in high abundances in both treated and untreated PKU and is positively related to the performance in spatial memory. This relationship is supported by a positive correlation between abundances of the same family with Y-maze performance in a study by Sanguinetti et al. (49). Overall, our observations of relationships between microbiome and behavior could suggest that there might be adaptive mechanisms to prevent brain damage as a result of high Phe in untreated conditions and potentially explain why the behavioral testing in untreated PKU mice shows inconsistencies and a much milder phenotype than expected based the severe symptoms of untreated PKU in humans, as also reported in the current study.

At the genus level, we observed minimal differences in the abundance of individual bacterial taxa between the WT and PKU animals on high Phe, even though these differences were present at the community level (alpha and beta diversities). This could suggest that the changes in bacterial abundance per genera might be more subtle and/or show more variation than the Phe-restriction treated animals. For the selected genera, we established four relationships with Phe levels, of which the majority were negative (Enterorhabdus, Intestinimonas, and Parabacteroides), except for Enterococcus. Contrary to what we observed at the family level, none of these genera were related to behavioral outcomes. However, we observed that the abundance of the Alloprevotella genus at 6 weeks was associated with the novel object recognition test outcome, at the time point at which we see a significant increase in this genus in the PKU animals on the high Phe diet. Although the microbiome is a community of interacting members that affect host physiology and behavior, these correlations could indicate that some members directly affect brain functioning. On the other hand, there is a possibility that these bacteria have the potential of canceling each other's function, thereby contributing to the very heterogenic outcome/phenotype observed in both PKU mice and humans.

While our study was exploratory, our results can be used as the starting point for future research into the development of probiotics to supplement dietary restriction in PKU treatment. Ideally, the presence and abundance of the chosen strains should be little affected by fluctuating Phe levels and should additionally show a positive relationship with cognitive behavior. We identified several bacterial taxa that adhere to this. Based on our results, we consider it would be worthwhile to focus the research on members from the Enterococcaceae and Erysipelotricaceae families, as they were in high abundance under high(er) Phe conditions and additionally showed a positive relationship with hippocampal-dependent memory in mice. In particular, research on the genera Enterococcus, unclassified_Erysipelotricaceae, and Alloprevotella should be conducted to further clarify their association with cognitive functioning and whether there is a potential to use these taxa in probiotic supplementation for PKU. In addition, taxa such as Clostridium XVII, Escherichia/Shigella, and Lactococcus, which were either more or exclusively abundant in the microbiome of high Phe animals relative to those of Phe-restricted animals, show potential for improvements not directly linked to behavior, such as an ability to increase diversity and resilience of the microbiome under high Phe levels.

It should be noted that literature on the effects of specific families or genera is often without consensus on their beneficial or detrimental impact on health. This could have several reasons, including that these effects occur primarily at the species level (i.e., families and genera could include both beneficial and disadvantageous microbial members), but also because the microbiome functions as one ecological community, in which bacteria might play different roles depending on their specific environment. For example, members of the Entereococcaceae family have been associated with disease, but others are already used as probiotics (50). Moreover, increased abundance of Erysipelotricaceae has been reported in patients with autism Specter disorder (51) and inflammation models (52), but otherwise is lowered in depression and Parkinson's disease (53, 54) and to be positively correlated with cognitive function (49). For Erysipelotricaceae, research also shows that it responds differently to diet or host health phenotypes, indicating that alterations in abundances can have a negative or beneficial impact depending on the disease (55). Among our genera of interest, we find genera that are already used in the food industry (Lactococcus) or as probiotics (Enterococcus) (56). Interestingly, the Alloprevotella genus has been associated with the production of short-chain fatty acid and amino acid metabolism in specific pathogen-free mice (57). Short-chain fatty acids (SCFAs) include acetate, propionate, and butyrate and are mainly produced by fermenting dietary fiber but can also be produced from amino acid metabolism. SCFAs are thought to play a crucial role in the functioning of the microbiome-gut-brain axis [for a review on the role of SCFAs in gut-brain communication, see (58)], making these genera interesting candidates to improve cognitive outcomes potentially.

Due to the lack of knowledge about microbial functioning and the inconsistency in the function of specific bacterial genera, the importance of isolating and identifying particular species for the development of microbiome-based therapeutic strategies should be emphasized. We have previously argued that for successful probiotic supplementation, general concepts of community dynamics (i.e., the potential interactions within the gut microbiome) and microbial invasions and colonization should be taken into consideration during the development of microbiome-based therapeutic strategies to increase the treatment effectiveness and safety (44). Isolation of the specific species will allow the determination of their metabolic potential and reinoculation of these species to examine their effects specifically in the context of PKU-related alterations in the microbiome composition. Ideally, these species could include members capable of utilizing and degrading Phe, thereby lowering Phe on the microbiome level if present in large but stable numbers. Using microbial-based therapeutic strategies in addition to the Phe-restricted diet can potentially improve the metabolic and behavioral outcome in treated patients with PKU by increasing the microbiome diversity and resilience while simultaneously allowing a more liberalized Phe-restriction by utilizing microbial metabolism.