A noncoding RNA modulator potentiates phenylalanine metabolism in mice

Abstract

The functional role of long noncoding RNAs (lncRNAs) in inherited metabolic disorders, including phenylketonuria (PKU), is unknown. We demonstrated that the mouse lncRNA Pair and human HULC associate with phenylalanine hydroxylase (PAH). Pair-knockout mice exhibited excessive blood phenylalanine, musty odor, hypopigmentation, growth retardation, and progressive neurological symptoms including seizures, which faithfully models human PKU. HULC depletion led to reduced PAH enzymatic activities in human induced pluripotent stem cell (hiPSC)-differentiated hepatocytes. Mechanistically, HULC modulated the enzymatic activities of PAH by facilitating PAH-substrate and PAH-cofactor interactions. To develop a therapeutic strategy for restoring liver lncRNAs, we designed GalNAc-tagged lncRNA mimics that exhibit liver enrichment. Treatment with GalNAc-HULC mimics reduced excessive phenylalanine in Pair^−/− and Pah^R408W/R408W mice and improved the phenylalanine tolerance of these mice.

One Sentence Summary:

A noncoding RNA chaperone of phenylalanine hydroxylase alleviated phenylketonuria in a mouse model.

The predominant view of human genetic diseases revolves around the identification of mutations/ dysfunctions of coding genes. However, about 98% of human genomic mutations occur within non-coding regions, and an understanding of the functional roles of noncoding RNAs in human genetic diseases is lacking. Phenylketonuria (PKU, OMIM 261600) and its milder variant hyperphenylalaninemia (HPA) are genetic disorders caused by a deficiency in the hydrolysis of L-phenylalanine (Phe) to L-Tyrosine (Tyr) (1). Over 1,000 PAH variants (PAHvdb database; http://www.biopku.org/home/pah.asp) have been identified and associated with PAH deficiency. Most of the variants are missense, usually resulting in protein misfolding and/or impairment of catalytic functions (2). Roughly 1 in every 10,000 infants is affected by this disease (3). Based on blood Phe concentration during diagnosis or screening in the neonatal period, PKU can be categorized as PKU (Phe >900 μmol/L), mild PKU (Phe <900 but > 360 μmol/L), or mild HPA (blood Phe is higher than the normal limit, but <360 μmol/L) (4, 5). Excessive Phe concentrations in patients with untreated PKU cause brain damage and associated mental retardation (6). Although PKU is largely considered a monogenic disorder, clinical evidence has indicated that PAH variants are incompletely correlated with the metabolic phenotypes of patients with PKU (7-9). In some patients diagnosed with PKU, no mutations could be identified in the PAH gene, suggesting that unknown factors may contribute to PKU (10). Consistent with this notion, the recent discovery of biallelic mutations in the DNAJC12 gene in patients exhibiting high blood Phe concentrations without mutations in PAH or genes involved in BH₄ (tetrahydrobiopterin, a PAH cofactor) metabolism (11) suggested the possibility that non-PAH genes may affect PAH function and subsequently increase blood Phe concentrations.

Newborns diagnosed with PKU are typically treated with a Phe-restricted diet and/or BH₄ supplementation. Patients with untreated PKU exhibit intellectual disability, behavioral issues, seizures, and psychiatric disorders (12). For patients with certain genotypes of mild PKU, supplementation with BH₄ has been suggested as an enzyme enhancement therapy (13). However, patients with a severe form of PKU (Phe concentrations above 900 μmol/L and usually above 1200 μmol/L) require additional treatment considerations. These patients also frequently respond poorly to BH₄ treatment (14). Given the potential immunogenic nature of Phenylalanine ammonia lyase (PAL) enzyme supplementation or substitution therapy (15, 16) and the uncertainty of gene therapy (17), additional therapeutic strategies are urgently needed.

RESULTS

Depletion of lncRNAs drives PKU

Long intergenic noncoding RNAs (lincRNAs) and long non-coding RNAs (lncRNAs) are transcripts with low coding potential. Aiming to investigate the biological importance of lncRNAs, we determined the lncRNA profile of mouse E18.5 embryos and the livers of 2-month-old adult mice (Fig. 1A and fig. S1A to C), finding that 2210408F21Rik (NR_040259) (renamed Pair: PAH-activating lincRNA) is one of the most upregulated lncRNAs in the liver of adults compared to the E18.5 embryos and exhibits low coding potential (CNIT score (−0.3544) (18) (fig. S1D to H, and Table S1). To deplete this mouse lncRNA, we introduced a two-nucleotide mutation using CRISPR/Cas9 to interrupt the splicing site after the first exon (Fig. 1B). Northern blotting indicated that Pair exhibits two major isoforms with molecular weights 730 bp and 1.5 kb, which were both depleted upon the introduction of the splicing-site mutation (Fig. 1C). Compared to wild-type and heterozygous littermates, Pair^−/− mice exhibited similar expression of neighboring genes with no detectable alterations in major organ development (fig. S2A to D).

Fig. 1. Pair^−/− mouse mimics human PKU disease.

(A) lncRNA profiling of livers from E18.5 embryos and from 2-month-old adult mice (n = 3). 2210408F21Rik is highlighted in red. (B) Schematic of using CRISPR-Cas9 to generate a Pair^−/− mouse model. (C) Northern blot detection of the expression of Pair in the indicated mouse livers. β-actin was used as a loading control. (D) Representative images of 12-month-old Pair mice. (E and F) Comparison of body weights of the indicated female (E) and male (F) mice at the age of 3 to 12 weeks. Data are shown as mean ± SD, one-way ANOVA. (G) Blood Phe concentrations in the indicated mice was tested every 2 weeks starting from 4 weeks of age (n = 5 Pair^+/+, n = 5 Pair^−/−, n = 7 Pair^+/− animals). Data are shown as mean ± SD, one-way ANOVA. (H) Cumulative survival curve of cohorts of indicated littermates (log-rank test). (I) Cumulative seizure-free survival curve of cohorts of indicated littermates (log-rank test). (J) Representative images of Pair^−/− mice experiencing seizures. (K) Left: representative images of brains from indicated mice at the age of 12 months (n = 5). Right: scatter plots represent brain weight quantification for both female and male indicated littermates at the age of 12 months (n = 5). Mean ± SD, one-way ANOVA. (L) Coronal sections of adult mouse brains subjected to immunohistochemical staining and quantitative data for TH⁺ neurons in the substantia nigra compact/ventral tegmental area of the indicated mice at the age of 12 months (n = 5). Mean ± SD, one-way ANOVA. Top row, 10× magnification; bottom row, 200× magnification. n.s., not significant at P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Both male and female Pair^−/− mice exhibited hypopigmentation (Fig. 1D), growth retardation and elevated serum Phe concentrations (Fig. 1E-G) reminiscent of human PKU. Pair^−/− livers showed no detectable changes in the expression of BH₄ biogenesis genes or PAH protein abundance (fig. S2E-G). Pair^−/− livers exhibited enzymatic deficiency in converting Phe to Tyr, and Pair^+/− livers showed impaired PAH enzymatic activity (fig. S2H-I). Pair^+/− mice exhibited blood Phe concentrations within the normal range; however, these animals showed elevated blood Phe concentrations upon Phe challenge compared to Pair^+/+ mice (fig. S2J). These results suggested that Pair^+/− livers exhibit partial PAH deficiency. Sanger sequencing of the Pah gene suggested that Pah^enu2 mice harbor a T788C mutation, while Pair^−/− mice harbor a wild-type Pah gene (Table S2 and fig. S2K and L).

The median life-span of Pair^−/− mice is 15.2 months (Fig. 1H), and more than 70% of the mice exhibited seizures starting at the ages of 8-10 months (Fig. 1I-J and Video S1-3). Following cardiopulmonary resuscitation (CPR), Pair KO mice experiencing a seizure can be rescued (Video S4). Pair^−/− brains were smaller and exhibited reduced tyrosine hydroxylase (TH)-positive neurons compared to wild-type and heterozygous littermates (Fig. 1K-L). Pair^−/− livers and serum exhibited diminished tyrosine concentrations. Tyrosine is catalytically produced by PAH in both serum and liver tissue (fig. S3A-E), confirming the presence of PAH deficiency. The role of Pair in seizures, in addition to regulating PAH enzymatic activities, could not be ruled out. Our data suggested that Pair^−/− mice model human PKU.

Pair and HULC associate with PAH

To understand the molecular mechanism of Pair, we performed Pair pulldown using biotinylated sense or anti-sense Pair in mouse livers (Fig. 2A and Table S3). Sense, but not antisense Pair, associated with mouse PAH (Fig. 2A). Beads-only and polyA were used as negative controls, while the association of ELAV1 with AR_3’ UTR served as a positive control (19) (Fig. 2A). We then performed CLIP (cross-linking immuno-precipitation) assay using mouse Pair^+/+ and Pair^−/− livers or human liver tissues from two healthy donors (fig. S4A and B, Table S4). The presence of PAH-RNA complexes was diminished upon Pair knockout (fig. S4A). PAH-lncRNA(s) complexes were detected in the two human liver donors (fig. S4B). The PAH-RNA complexes (fig. S4A-B, blue boxes 1-3) were subjected to reverse transcription and Sanger sequencing (Table S5). As expected, mouse PAH associated with mouse Pair (nt. 460-496) in Pair^+/+ livers (Fig. 2B, bottom row). Human PAH associated with one human lncRNA gene, HULC (nt. 183-216) (Fig. 2B, top row). Although HULC has been suggested to be upregulated in liver cancer (20), northern blotting indicated that HULC, similarly to Pair, is specifically expressed in normal liver tissues (fig. S4C to E), suggesting the biological relevance of HULC in liver homeostasis and function.

Fig. 2. Pair and HULC associate with PAH.

(A) Protein candidates interacting with Pair were revealed by LC-MS. The x-axis indicates different experimental groups. (B) PAH-binding sites along HULC and Pair identified by CLIP assay. The figure represents the read coverage along HULC and Pair transcripts obtained by mapping Sanger sequencing reads to the representative transcript. cDNA counts are shown. (C) Representative images of immunolabeling with fluorescent in situ hybridization to simultaneously detect colocalization of the indicated RNA and protein molecules in human primary hepatocytes. TUG1 and Hu were used as RNA and protein controls, respectively. Scale bars, 50 μm. (D) Resolution of SHAPE reactivity by capillary electrophoresis. Four-color electropherograms for the products of one multicolor run by 6-FAM-labeled NMIA modified RNA (+SHAPE), VIC-labeled control RNA (−SHAPE) of HULC nucleotides 178 to 202, NED-labeled ddA, and a PET-labeled ddT are shown. Bottom panel: normalized SHAPE reactivity of HULC nucleotides 178 to 202. x-axis indicates the sizes of capillary electrophoresis. Data are shown as mean ± SD of n = 15 independent experiments. (E and F) EMSA using recombinant His-tagged PAH and [γ-32P]–labeled Pair nucleotides 467 to 488 (E) or HULC nucleotides 180 to 202 (F) WT or mutant oligonucleotides. Unlabeled Pair or HULC WT or mutant RNA oligonucleotides were included as competitors. (G) MS2-TRAP assay using Pair^−/− hepatocytes expressing indicated plasmids performed by immunoblotting using the indicated antibodies. (H) Immunoblotting (bottom) or autoradiography (top) of CLIP assay using the indicated hepatocytes expressing the indicated plasmids.

RNA immunoprecipitation (RIP) assay confirmed that Pair and HULC associate with PAH protein in mouse and human livers, respectively (fig. S5A to C). We demonstrated that the regulatory (aa. 1-142) and catalytic (aa. 142-411) domains of PAH are required for PAH-Pair interactions (fig. S5D-G). Next, using primary cultured mouse or human hepatocytes, we demonstrated that PAH protein faithfully co-localizes with Pair or HULC in the cytosol, but not with another cytosolic lncRNA, Tug1 (Fig. 2C and fig. S5H to I).

We substituted each nucleotide between Pair (nt. 460-496) and HULC (nt. 183-216) using the most common transition and transversion types (21) and expressed wild-type or mutant Pair or HULC in Pair^−/− or HULC-deficient hepatocytes (fig. S6A and B). RIP assay suggested that Pair (nt. 470-488) and HULC (nt. 183-200) are required for PAH-Pair and PAH-HULC interactions (fig. S6A and B). Furthermore, Pair 479A>G and HULC 191A>G abolished PAH-Pair and PAH-HULC interactions, respectively (fig. S6A and B).

SHAPE assays (22, 23) indicated that HULC U190 and A191 (capillary electrophoresis size #290 and 289) exhibited chemical labeling, suggesting the presence of a loop structure flanked by low chemical probing (Fig. 2D and fig. S6C). Furthermore, HULC harbors 3 additional stem-loop structures, as revealed by SHAPE assays (fig. S7A-D). These 3 stem-loop structures exhibited distinct structures compared to the stem-loop at nt. 184-216 (fig. S7E-H). The chemical labeling of HULC^A191 was verified by in vivo SHAPE (fig. S8A-B). We further looked at the secondary structure of Pair, finding that Pair nt. 467-484 also exhibits a stem-loop structure, with A479 showing robust chemical probing (fig. S9A-C). Pair nt. 467-484 exhibited a similar 3-dimensional structure to HULC nt. 184-216 but not the above 3 stem-loops of HULC (fig. S9D).

Wild-type Pair and wild-type HULC, but not the Pair 479A>G or HULC 191A>G mutants (referred to as Pair mut or HULC mut, respectively) associated with recombinant PAH, as revealed by RNA electrophoretic mobility shift assay (EMSA) assay (Fig. 2E, F). Expression of MS2-tagged wild-type Pair/HULC, but not the mutants, rescued the association with PAH protein in Pair^−/− hepatocytes (Fig. 2G and fig. S10A). We then performed a rescue CLIP assay by expressing exogenous wild-type Pair/HULC or mutants in mouse Pair^−/− hepatocytes (Fig. 2H and fig. S10B). Exogenous Pair or HULC WT, but not mutants, associated with PAH similarly to endogenous Pair or HULC (Fig. 2H). Similar amounts of PAH protein were immuno-precipitated using anti-PAH antibody (Fig. 2H, bottom). These findings suggested that the human lncRNA HULC and mouse Pair both associate with PAH.

HULC/Pair modulate the enzymatic activity of PAH

To demonstrate the underlying molecular mechanisms of HULC/Pair in the enzymatic activity of PAH, we first determined that there are roughly 800 HULC RNA, 700 Pair RNA, and 4,000 PAH protein molecules per human or mouse hepatocyte (fig. S10C-E). In Pair^−/− or HULC-deficient hepatocytes, exogenous expression of Pair/HULC in a dose-dependent manner led to the reduction of cellular Phe concentrations (fig. S10F and G). The concentrations of amino acids other than Phe and Tyr were minimally affected upon Pair depletion (fig. S10H).

The enzymatic activity of PAH requires the presence of a cofactor, BH₄ (24). Mutations of PAH affecting PAH-Phe or PAH-BH₄ interactions impair the catalytic activity of PAH (25). Previous research indicated that the Phe molecule associates with the regulatory domain of PAH as an allosteric activator (26-28). Using the crystal structure information of PAH (PDB 6HYC), protein structural bioinformatics analysis suggested that HULC nt. 184-216 associates with the regulatory domain of PAH and allosteric Phe (Fig. 3A). HULC A¹⁹¹ forms hydrogen bonds with both Thr63 and His64 (Fig. 3A). The mutation of A¹⁹¹ to G¹⁹¹ causes a change at position 6 from an amino group (6-NH₂) to a carbonyl group (6-CO). This change leads to the loss of the hydrogen bond between A¹⁹¹ and His64, because carbonyl groups are hydrogen bond acceptors while amino groups are often hydrogen bond donors. The amino group (2-NH₂) of G¹⁹¹ may also cause steric hindrance with Thr63. Phe forms hydrogen bonds with PAH via Asn61 and Leu62, and it also forms stacking interactions with A¹⁹¹ (Fig. 3B). These interactions stabilize the whole structural complex in a conformation that makes the active site fully accessible to Phe as a substrate and BH₄ as a cofactor. Our findings suggested that HULC serves as an important factor that stabilizes the interaction between allosteric Phe and PAH, as previous hypothesized (26). Aside from A¹⁹¹, HULC interacts with PAH via several other residues to achieve the binding specificity of the PAH-HULC interaction: A¹⁹⁵ forms a hydrogen bond with Tyr166; A²¹⁴ forms a hydrogen bond with Arg157; and G²⁰² also interacts with Tyr154 via a hydrogen bond (Fig. 3C). Pair adapts a binding mode to PAH that is similar to HULC: both Pair and HULC have two nucleotides that stick out to form a T-shape and exhibit stacking interactions with His64 of PAH, and these two nucleotides (A and U) stabilize each other through stacking interactions (Fig. 3D). The only difference between Pair and HULC with regard to this behavior is that the order of the sequences of these two nucleotides in Pair is UA; in HULC, the order is AU (Fig. 3D).

Fig. 3. HULC and Pair modulate the enzymatic activities of PAH.

(A) HULC binds to PAH to stabilize the allosteric Phe-induced open conformation of PAH. (B) Magnified view of HULC’s interaction with the regulatory domain of PAH. HULC A¹⁹¹ forms hydrogen bonds with His⁶⁴ and polar interaction with Thr⁶³. Phe forms stacking interaction with HULC A¹⁹¹. Color representation in both (A) and (B): green indicates PAH; orange and magenta indicate HULC; yellow indicates allosteric Phe; and cyan indicates Thr⁶³ and His⁶⁴. (C) HULC A¹⁹⁵ forms a hydrogen bond with Tyr¹⁶⁶; HULC A²¹⁴ forms a hydrogen bond with Arg¹⁵⁷; HULC G²⁰² forms a hydrogen bond with Tyr¹⁵⁴. Green indicates PAH; orange and magenta indicate HULC; and Cyan indicates Tyr¹⁵⁴, Arg¹⁵⁷, and Tyr¹⁶⁶. (D) Superimposition of the three-dimensional structures of HULC nucleotides 184 to 216 (cyan) and Pair nucleotides 467 to 484 (magenta) after docking to PAH (yellow). Pair A⁴⁷⁹, U⁴⁸⁰ and HULC U¹⁹⁰, A¹⁹¹ are indicated by arrows. (E) Streptavidin pull-down using Bio-Phe/Bio-BH₄ followed by immunoblotting detection using anti-PAH antibody in the indicated hepatocytes. (F) Fold change of peptides recovered from LiP-MS. Top panel: comparison of PAH WT only, PAH incubated with HULC, and PAH incubated with a HULC mutant sequence. Bottom panel: comparison of PAH incubated with HULC, PAH incubated with HULC and Phe, and PAH incubated with HULC and BH₄. The x-axis indicates the amino acid position of full-length PAH; the y-axis indicates the fold change of peptide recovery number. (G and H) ELISA measurement of the percentage of PAH-associated BH₄ (G) or PAH-associated Phe (H) in the indicated hepatocytes expressing the indicated mimics. Data are shown as mean ± SEM of n = 5 independent experiments, one-way ANOVA. n.s., not significant at P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001.

We reasoned that the HULC-PAH interaction may facilitate the binding of PAH to Phe or BH₄. To address this hypothesis, we synthesized biotinylated-Phe and -BH₄ (referred to as Bio-Phe and Bio-BH₄) (fig. S11A). Compared to Pair^+/+ livers, Pair^−/− livers exhibited a similar abundance of PAH protein; however, the PAH-Phe and PAH-BH₄ interactions were impaired following Pair depletion (Fig. 3E). We then determined that digoxin (DIG)-tagged HULC showed detectable interactions with bacterially-expressed PAH proteins but undetectable associations with Bio-Phe or Bio-BH₄ (fig. S11B). Wild-type PAH was included as a positive control (fig. S11B). PAH G46S, F55L, and P281L have been suggested to impair interactions between Bio-Phe and/or Bio-BH₄ (29, 30), which was confirmed (fig. S11B). LncRNA LINK-A and the interaction between LINK-A and PIP₃ (31) were included as negative controls (fig. S11B).

The N-terminal regulatory domain of PAH can undergo a conformational change to switch between “open” and “closed” states (30). We hypothesized that the binding of HULC may stabilize the PAH-Phe-HULC complex and stabilize the PAH protein in the “open” state (fig. S11C). We applied limited proteolysis (LiP) followed by liquid chromatography–mass spectrometry (LC–MS) analysis (LiP-LC–MS) (32) to address this hypothesis (fig. S11C). Open loop regions (gray) and Lip-resistant regions (dark blue) were determined (Fig. 3F and Table S6). Region aa 57-66 showed resistance to LiP in the presence of HULC but not the HULC mut, suggesting that this region associates with HULC (Fig. 3F, top panel-magenta). Notably, aa 57-66 of PAH exhibited recovery in the presence of Phe (Fig 3F, bottom panel-magenta), which was consistent with the structural modeling, demonstrating that Thr63 and His64 associate with HULC and Phe as a complex. Furthermore, LiP-MS suggested a few regions that were dynamically regulated upon HULC binding (Fig.3F, green).

Hence, our data suggested that HULC associates with PAH in vitro and facilitates the potential conformational change of PAH. Consistent with this notion, expression of wild-type Pair or HULC rescued PAH-Phe and PAH-BH₄ interactions and reversed cellular Phe accumulation, while the Pair and HULC mutants failed to do so (Fig. 3G-H and fig. S11D-E).

To determine the functional role of the HULC-PAH interaction in vivo, HULC-deficient hiPSC were further differentiated into hepatocytes expressing wild-type HULC or the A191G mutant (fig. S12A). Expression of wild-type HULC or the A191G mutant showed no detectable effects on the abundance of PAH protein (fig. S12B). Upon depletion of HULC, the conversion of ¹⁴C-Phe to ¹⁴C-Tyr was reduced with concurrent elevation of cellular Phe concentrations (fig. S12C and D). Expression of WT HULC, but not the A191G mutant, restored the enzymatic activity of PAH and reversed the cellular accumulation of Phe (fig. S12C and D). Similarly, wild-type PAH, but not the TH63-64PN mutant, expressed in PAH-deficient hiPSC-hepatocytes rescued the enzymatic deficiency of PAH (fig. S12E-H). Taken together, our findings suggested that Pair-PAH and HULC-PAH interactions facilitate the PAH-driven catalysis of Phe to Tyr.

HULC mimics restore PAH enzymatic activity

We reasoned that supplementing lncRNAs might improve the catalytic activities of the PAH mutants, leading to reduced serum Phe concentrations and improved symptoms in patients with PKU. We designed Scramble (Scr) and HULC mimics representing wild-type HULC nt. 181-201 and HULC A191G mutated sequences for the following studies. We first selected 17 PAH mutants that were identified from patients with PKU and are known to affect the enzymatic activity of PAH (33, 34), and we collected the bacterially-expressed wild-type and mutant PAH (Fig. 4A and fig. S13A). EMSA assay indicated that 13 of the 17 PAH mutants, as well as wild-type PAH, associated with the HULC mimics; on the contrary, PAH TH63-64PN, R157N, N207S, and S349L failed to associate with the HULC mimics (fig. S13B). The denatured wild-type PAH proteins (WT denat.) were included as a negative control (fig. S13B). We further quantified the interactions between PAH proteins (WT/mutants) and HULC mimics (WT), finding that PAH protein interacted with the HULC mimics with a K_d value of 131.6 nM (fig. S13C). PAH TH63-64PN, R157N, N207S, and S349L, but not the other PAH mutants, exhibited decreased binding affinities (fig. S13C).

Fig. 4. HULC mimics facilitate PAH-Phe and PAH-BH₄ interactions.

(A and B) Log₂ of relative fold change of His-PAH WT/mutants and Biotin-Phe (A) or His-PAH WT/mutants and biotin-BH₄ (B) binding affinity in the presence of the indicated lncRNA mimics. The fold change was normalized using His-PAH WT in the presence of the LINK-A mimic. (C) Phe and Tyr concentrations in neonatal blood spots in the presence of His-tagged PAH WT or indicated mutants and the indicated lncRNA mimics. Data are shown as mean ± SD of n = 3 independent experiments, Student’s t test. The lncRNA mimic representing LINK-A nucleotides 1100 to 1117 was included as a negative control. (D) Determination of k_cat of recombinant PAH WT or indicated mutant proteins in the presence of the indicated mimics. Data are shown as the mean ± SD of n = 3 independent experiments, one-way ANOVA. n.s., not significant at P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001.

We measured the binding affinities between PAH (WT/mutants) and Bio-Phe or Bio-BH₄ in the presence of HULC mimics (fig. S14A-O and S15A-O) and summarized the change in binding affinities compared with PAH WT in Figs. 4A-B. LncRNA mimics representing LINK-A 1100-1117 (31) were included as a negative control (Figs. 4A-B). For wild-type PAH protein, the presence of HULC mimics enhanced PAH-Phe (Fig. 4A) and PAH-BH₄ interactions (Fig. 4B), while HULC mut mimics failed to do so (Fig. 4A-B and S14A-O and S15A-O). The PAH mutants we tested all exhibited impaired binding affinities toward Phe and/or BH₄ compared to wild-type PAH protein (Fig. 4A-B and S14A-O and S15A-O). The presence of HULC mimics, but not HULC mut, enhanced the affinity of Phe and/or BH₄ to the PAH mutants. A PAH catalytic pocket deletion mutant (Δ245-379) was included as a negative control (fig. S14O and S15O).

We further demonstrated that wild-type PAH effectively catalyzed Phe to Tyr, which was abolished in all 17 PAH mutants we tested (Fig. 4C, left section, bar #2). In the presence of HULC mimics, but not LINK-A mimics or HULC mut, wild-type PAH showed enhanced enzymatic activity in converting Phe to Tyr (Fig. 4C, middle section, bar #2). Furthermore, 11 of the 17 PAH mutants also showed improved enzymatic activities in converting Phe to Tyr (Fig. 4C, middle section, as indicated in red). The enzymatic activities of the rest of the PAH mutants were not significantly affected by HULC mimics (Fig. 4C, middle section, light blue and brown).

We then determined the k_cat of bacterially-expressed human wild-type and mutant PAH in the presence of Scramble (Scr), wild-type HULC, or mutant HULC mimics (Fig. 4D and fig. S16A-S). Recombinant PAH proteins with deleted catalytic domains (Δcat) were included as a negative control (fig. S16A). In the presence of the wild-type HULC mimics, the k_cat of wild-type PAH was increased (Fig. 4D and fig. S16B). All PAH mutants we tested exhibited impaired k_cat under identical conditions (Fig. 4D and fig. S16C-S). The presence of wild-type HULC mimics, but not mutant HULC mimics, enhanced the k_cat of PAH F39L, A47V, F55L, I65S, P275L, P281L, I283N, F299C, A300S, I318T, and R408W, but not the enzymatic activities of PAH TH63-64PN, R157N, N207S, and S349L (Fig. 4D and fig. S16C-S). Therefore, our data suggested that the presence of HULC facilitates the enzymatic activity of wild-type PAH and a cohort of mutants observed in patients with PKU.

HULC/Pair enhances the enzymatic activity of the PAH R408W mutant

19.2-73% of patients with PKU harbor a R408W mutant (35, 36), and patients with this mutation respond poorly to currently available treatment options. We hypothesized that overexpression of HULC/Pair might improve the enzymatic activity of PAH R408W. To address this, we first generated a Pah^R408W/R408W mouse strain using CRISPR/Cas9 (fig. S17A-B). Similar to Pair^−/− mice, Pah^R408W/R408W mice exhibited hyperpigmentation, growth retardation, seizures, and elevated blood Phe concentrations, with no detectable alterations in the protein stability of PAH or the expression of Pair (fig. S17C-H). These observations provided genetic evidence confirming that Pair/HULC and PAH act in a linear pathway. In primary cultured mouse hepatocytes isolated from Pair^−/− or Pah^R408W/R408W mice, we determined cellular Phe, Tyr, and tryptophan (Trp) concentrations, finding that expression of wild-type Pair/HULC, but not the mutant (fig. S17I), restored PAH-Phe and PAH-BH₄ interactions and cellular Phe and Tyr concentrations, with no effect on Trp status (fig. S17I-L).

We collected skin fibroblasts from healthy donors who harbored wild-type PAH genes and from a patient with PKU (ID: NA02406) who harbored F299C and R408W mutations. We reprogrammed these fibroblasts into hiPSCs. These hiPSCs were further induced into hepatocytes (fig. S18A-B). Two hiPSC clones were used, referred to as PKU #7 and PKU #17 (fig. S18A-C). These hiPSC-derived hepatocytes exhibited similar expression of HULC, PAH, and hepatic markers (fig. S18B-C). The hiPSC-derived hepatocytes derived from both PKU #7 and PKU #17 clones exhibited increased Phe concentrations compared to the control (healthy donor), confirming PAH enzymatic activity deficiency (fig. S18D). Expression of full-length HULC in these hiPSC-derived hepatocytes reduced Phe concentrations in PKU #7 and PKU #17 hepatocytes (fig. S18D-E). This suggested that a supply of HULC might enhance the enzymatic activity of PAH in F299C and R408W mutants.

GalNAc-HULC mimics improve phenylalanine metabolism in mice

To design lncRNA mimics that could provide therapeutic value in vivo, we synthesized scramble (Scr), wild-type, and mutant HULC mimics using 2'-Fluoro (2'-F) modified RNA monomers to provide nuclease resistance in vivo. The potential secondary structure of wild-type or mutant HULC mimics remained identical (fig. S19A). Both full-length HULC and the mimics exhibited similar binding affinities for PAH (fig. S19B). Furthermore, the HULC A191G mutant and HULC mut mimics both abolished these interactions (fig. S19B).

To facilitate the liver-enrichment of HULC mimics, we applied three types of HULC mimics: HULC mimics alone, peptides representing ApoE1 (Apo)-tagged HULC mimics, and N-Acetylgalactosamine (GalNAc)-tagged HULC mimics via intravenous (i.v.) or subcutaneous (SubQ) injection. GalNac-conjugated oligonucleotides have recently been suggested to assist with liver-targeted siRNA delivery (37). The 3'-triantennary GalNAc was conjugated to the HULC mimic.

We considered that a three-day treatment trial (Fig. 5A) would serve as the most convenient method to determine the efficacy of these mimics. Since PAH R408W-harboring patients respond to current BH₄ supplementation treatments poorly, we treated female and male Pah^R408W/R408W mice with the indicated mimics (fig. S19C to D). GalNac-HULC administered via i.v. injection exhibited the highest efficacy in reducing excessive Phe concentrations in both female and male Pah^R408W/R408W mice (fig. S19C to D).

Fig. 5. GalNAc-HULC lncRNA mimics alleviate PKU symptoms.

(A) Graphic illustration of GalNAc-tagged HULC mimics (left) and the treatment schedules (right). (B) Blood Phe concentrations were monitored every day for short-term treatment in female and male Pair^−/− mice with the indicated mimics. Data were analyzed with Student’s t test. (C) Blood Phe concentrations were monitored every other day for medium-term treatment in female and male Pair^−/− mice treated with the indicated mimics. Data were analyzed with Student’s t test. (D) Blood Phe concentrations were monitored every day for short-term treatment in female or male Pah^R408W/R408W mice treated with the indicated mimics. Data were analyzed with Student’s t test. (E) Blood Phe concentrations were monitored every other day for medium-term treatment in female or male Pah^R408W/R408W mice treated with the indicated mimics. Data were analyzed with Student’s t test. (F) Measurement of blood tyrosine concentrations for medium-term treatment in female and male Pah^R408W/R408W mice treated with the indicated mimics, Data were analyzed with Student’s t test. (G and H) Phe clearance test (G) or area under the curve (AUC) (H) of female or male Pah^R408W/R408W mice subjected to a Phe-free diet and pretreatment with the indicated mimics. Data are shown as mean ± SD, Student’s t test. (I) Phe tolerance test of female or male Pah^R408W/R408W mice subjected to a Phe-free diet and pretreatment with indicated mimics for 3 days, followed by water containing 0, 0.75, 1.5, 3.0, or 6.0 mg/ml of Phe (with the dose increasing every 2 days). Data were analyzed with Student’s t test. (J) Blood Phe concentrations were monitored every day for short-term treatment in Pah^R408W/R408W mice subjected to sapropterin alone or in combination with the indicated mimics. Data are shown as mean ± SD, one-way ANOVA. n.s., not significant at P > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

We then determined the concentrations of biotin-labeled HULC or GalNAc-HULC mimics in the major organs, finding that a substantial portion of biotin-labeled GalNAc-tagged HULC mimics was detected in the liver between 3 and 72 hours after dosing, while the lungs and spleen exhibited no detectable accumulation of GalNAc-HULC (fig. S19E to G). Biotin was undetectable in the liver 12 hours post-injection (fig. S19H).

As a proof of concept, we applied GalNAc-tagged Scr or HULC mimics using short-term (3-day) and medium-term (12-day) treatment regimens (Fig. 5A). To rule out the indirect effects of the GalNAc tag, GalNAc-HULC mut mimics were included (Fig. 5B). Both male and female Pair^−/− mice exhibited reduced serum Phe concentrations 24 hours after injection of GalNAc-HULC mimics, but not the Scr or HULC mut mimics (Fig. 5B). During the medium-term treatment, administration of GalNAc-HULC mimics reduced blood Phe concentrations in Pair^−/− mice throughout the treatment term (Fig. 5C). In the Pah^R408W/R408W mice, administration of GalNAc-HULC mimics similarly reduced blood Phe concentrations during the short-term and medium-term treatments (Fig. 5D to E). The Pah^R408W/R408W mice subjected to GalNAc-HULC mimics showed increased blood Tyr concentrations compared to animals given scramble mimics (Fig. 5F). To evaluate Phe clearance capacity, Pah^R408W/R408W mice were subjected to pre-treatment with Scr or HULC mimics followed by a Phe challenge (Fig. 5G). The area under the curve (AUC) showed a reduction in Phe concentrations following GalNAc-HULC mimic treatment compared to GalNAc-Scr (Fig. 5H).

To determine whether the administration of GalNAc-HULC facilitates higher tolerance to dietary Phe intake in PKU animals, Pah^R408W/R408W were fed a Phe-free diet for three days followed by water containing increasing doses of Phe. Administration of GalNAc-HULC mimics allowed Pah^R408W/R408W animals to maintain relatively low blood Phe concentrations (<600 μM) upon Phe challenge up to 3.0 mg/ml (Fig. 5I), suggesting that GalNAc-HULC mimics may also be able to improve tolerance to dietary Phe.

Sapropterin, a synthetic formulation of BH₄, has been used as enzymatic enhancer for patients with HPA (38). Patients with PKU harboring R408W are resistant to sapropterin treatment (14). We tested the efficacy of sapropterin in Pair^−/− and Pah^R408W/R408W mice, finding that supplementation of sapropterin showed no effect on the blood Phe of Pair^−/− and Pah^R408W/R408W animals (fig. S20A, B). A combinational treatment of GalNAc-HULC and sapropterin showed a cooperative effect in reducing the blood Phe concentrations of Pah^R408W/R408W animals (Fig. 5J). These findings suggested that a supply of GalNAc-HULC mimics may enhance the association between PAH and BH₄, thereby improving the therapeutic effect of sapropterin. The HULC mimics showed no detectable effects on the body weight, liver function, and kidney function of Pair^−/− and Pah^R408W/R408W mice (fig. S20C-G). Taken together, our data suggested that application of HULC mimics could enhance the enzymatic activities of certain mutated PAH proteins, offering a potential intervention for patients with PKU.

DISCUSSION

PKU has long been considered a single-gene disease with an autosomal recessive inheritance pattern, in which mutations at the PAH locus lead to impaired enzymatic function and contribute to a hyperphenylalaninemia metabolic phenotype, subsequently resulting in cognitive phenotypes such as mental retardation. Recent genotype-phenotype correlation analyses of PAH mutations suggested a substantial discrepancy between genotypes and their predicted metabolic or cognitive phenotypes (8, 9, 39, 40). Variations in genomic, epigenomic, transcriptomic, proteomic, and metabolic systems could all contribute to the PAH deficiency of patients with PKU (41). It has been shown that patients, even siblings sharing identical mutant PAH genotypes, can have greatly differing cognitive and metabolic phenotypes (10). This highly suggests that a PAH genotype cannot consistently and reliably predict the monogenic phenotype (10). Consistent with this notion, the recent discovery of biallelic mutations in the DNAJC12 gene in patients with high Phe concentrations suggested the possibility that non-PAH genes contribute to this disease (42). Hence, factors modulating PAH expression, PAH protein stability, BH₄ biogenesis, and PAH enzymatic activity may play important roles in maintaining PAH enzymatic proficiency. More than 95% of genetic mutations occur at the non-coding regions of the human genome, yet the functional importance of lncRNAs in human genetic diseases remains elusive. We demonstrated that depletion of the lncRNA Pair leads to phenotypes that model human PKU, including hypopigmentation, growth retardation, seizures, and neuronal loss. Whether the presence of the HULC genomic variants contributes to the severity of PKU requires further investigation, given the heterogeneity of PAH genotypes and the lack of knowledge of HULC genomic variant statuses in current data collections.

The nature of genetic diseases has resulted in the development of compensatory lifestyle considerations that have marked, potentially negative impacts on the overall quality of life of patients. BH₄ and its synthetic analog, sapropterin, have been used as enzyme cofactors for a subset of patients with PKU, particularly those with mild PKU (43-45). However, patients with PKU and Phe >1200 μmol/L are less likely to respond to BH₄ supplementation (46). PKU could potentially be genetically corrected via adenoviral or related vectors (47); nevertheless, their specificity, stability, and the adenoviral genes responsible for immune response require further research studies (48). Enzyme replacement or substitution therapy, in which PAH or PAL fusion proteins (Palynziq) are intravenously injected, leads to decreased plasma Phe (15, 16, 49, 50) but also potentially concerning immune response-related symptoms experienced by a percentage of patients with PKU (51, 52). Hence, development of additional therapeutic strategies is needed, such as an approach using liver-enriched lncRNA mimics that specifically enhance PAH-substrate and PAH-cofactor binding affinities and enhance enzymatic activity in vitro and in vivo.

The HULC mimics we developed target lncRNAs as a therapeutic strategy against a human inherited metabolic disorder. Considering lncRNA mimics as a potential therapeutic option has the following advantages: 1) Flexibility in synthesizing lncRNA mimics according to the target sequence and tagging with organ-targeting peptides for tissue-specific distribution. The target sequence could be designed to improve the enzymatic activity of key enzyme(s) in other metabolic disorders. 2) Profound stability in vivo due to the modified covalent bonds between nucleotides, which are resistant to DNase and RNase. 3) Liver-enriched tissue distribution assisted by the GalNAc-tag. 4) Low potential for organ toxicity, as supported by our observation that administration of lncRNA mimics resulted in no detectable effects on liver or kidney function. It is likely that combining HULC mimics with current dietary restrictions, sapropterin supplements, and enzyme replacement or substitution therapies would further improve patient outcomes, particularly for patients with severe PKU.

Low conservation between mouse and human lncRNAs has greatly hindered the discovery of disease-driving lncRNAs and their roles in human diseases; however, it is possible that RNA “domains” might be conserved between human and mouse lncRNAs. HULC and Pair associate with PAH protein at the N-terminal regulatory domain. The presence of HULC could rescue PAH enzymatic activity in Pair-deficient cells and vice versa, shedding light on the therapeutic potential of the vast, currently-unconsidered, and poorly conserved lncRNAs.

PAH is self-regulated by the N-terminal regulatory domain. The crystal structure of PAH suggests that the C-terminal helix is required for the tetramerization of PAH. HULC/Pair is unlikely to affect the oligomerization of PAH, as supported by the finding that HULC mimics showed no notable effect on the enzymatic activities of PAH Y414C and Y417H. The N-terminal regulatory domain can undergo a conformational change to switch between “open” and “closed” states (30). We reason that the binding of HULC to the N-terminal regulatory domain of PAH may stabilize the PAH-Phe-HULC complex, leading the PAH protein to operate in the “open” state and allowing Phe and BH₄ to access the enzyme. Our finding is consistent with the previous hypothesis that an “unknown” allosteric factor stabilizes the interaction between PAH and Phe, serving as a therapeutic chaperone (26). Taken together, our findings suggested the functional importance of lncRNAs in phenylalanine metabolism.

Data availability:

The raw lncRNA array data for this manuscript are available at GEO under the accession number GSE120207. All other data are available in the main text or the supplementary materials. Due to the COVID-19 pandemic, our genetically modified mouse colonies died out, and we are in the process of reestablishing our homozygous mouse populations. The skin fibroblasts and hiPSCs were obtained from the Coriell Institute with MTA and are subjected to restrictions on redistribution and sharing. Request of these cells shall be directly obtained from Coriell Institute. All other research materials are available from the corresponding author upon reasonable request.