Abstract
Eukaryotic initiation factor 5A (eIF5A) is an essential factor with a unique amino acid, hypusine, required for its activity. Hypusine is formed exclusively in eIF5A by a post-translational modification involving two enzymes, deoxyhypusine synthase (DHPS) and deoxyhypusine hydroxylase (DOHH). Each of the three genes, Eif5a, Dhps or Dohh is required for mouse embryonic development. Variants in EIF5A or DHPS were recently identified as the genetic basis underlying certain rare neurodevelopmental disorders in humans. To investigate the roles of eIF5A and DHPS in brain development, we have generated four conditional knockout mouse strains using the Emx1-Cre or Camk2a-Cre strain and examined the effects of temporal- and region-specific deletion of Eif5a or Dhps. The conditional deletion of Dhps or Eif5a by Emx1 promotor driven Cre expression (E.9.5, cortex and hippocampus) led to gross defects in forebrain development, reduced growth and premature death. On the other hand, the conditional deletion of Dhps or Eif5a by Camk2a-promoter driven Cre expression (postnatal, mainly in the CA1 region of hippocampus) did not lead to global developmental defects; rather, these knockout animals exhibited severe impairment in spatial learning, contextual learning and memory, when subjected to the Morris Water Maze test and a contextual learning test. In both models, the Dhps knockout mice displayed more severe impairment than their Eif5a knockout counterparts. The observed defects in brain, global development or cognitive functions most likely result from translation errors due to a deficiency in active, hypusinated eIF5A. Our study underscores the important roles of eIF5A and DHPS in neurodevelopment.
Significance eIF5A, an essential translation factor, is the only protein that undergoes a unique posttranslational modification, that converts lysine to hypusine by conjugation of the aminobutyl moiety from the polyamine spermidine. Hypusine biosynthesis occurs by two enzymatic steps involving deoxyhypusine synthase (DHPS) and deoxyhypusine hydroxylase (DOHH). Mutations in EIF5A or DHPS have been associated with rare neurodevelopmental disorders in humans. To understand the mechanisms underlying the pathogenesis of the disease, we generated mutant mice with brain-specific deletions of Eif5a or Dhps. The Eif5a and Dhps conditional knockout mice exhibited impairment in brain development, growth and cognitive functions. These animal models may serve as useful tools in the development of therapies against the eIF5A- or DHPS-associated neurodevelopmental disorders.
Introduction
The eukaryotic initiation factor 5A (eIF5A) is the only cellular protein that is activated by a unique posttranslational modification that forms an unusual amino acid, hypusine [Nε-(4-amino-2-hydroxybutyl)lysine] (1). Hypusine is essential for the activity of this factor. It is formed in the eIF5A precursor by two consecutive enzymatic steps (Fig 1) (2). The first enzyme, deoxyhypusine synthase (DHPS) (3), catalyzes the transfer of the aminobutyl moiety from the polyamine spermidine to one specific lysine residue of the eIF5A precursor to form an intermediate, deoxyhypusine [Nε-(4-aminobutyl)lysine] residue, which is subsequently hydroxylated by deoxyhypusine hydroxylase (DOHH) (4) to complete the synthesis of hypusine (Fig.1). Homozygous, whole-body deletion of any of these three genes, Eif5a, Dhps or Dohh, in mice causes early embryonic lethality (5, 6) and postnatal conditional deletion of Eif5a or Dhps leads to inhibition of organ development in mice (7, 8).
Polyamines (putrescine, spermidine and spermine) are essential for eukaryotic cell growth and regulate a vast array of cellular activities (9-11). Polyamine homeostasis is tightly regulated by intricate mechanisms at multiple levels including biosynthesis, catabolism and transport. The majority of cellular polyamines are bound to RNA and the most important function of polyamines appears to be the regulation of translation as polycations (10, 12) and also as a component of hypusine in eIF5A. As hypusine is vital for eIF5A activity and cell proliferation, hypusine synthesis defines a critical function of polyamines in eukaryotic cell growth (13).
In contradiction to its nomenclature, eIF5A facilitates translation elongation rather than translation initiation (14-16). In yeast, eIF5A promotes translation elongation broadly at ribosome stall sites including sequences encoding polyproline stretches and it also enhances translation termination (14, 17). eIF5A binds to the 80S ribosome between the peptidyl-tRNA site and the exit tRNA site (18). The hypusine side chain of eIF5A stabilizes the binding of the peptidyl tRNA to the 80S ribosome and facilitates peptide bond synthesis. Two or more eIF5A isoform genes have been identified in many eukaryotic organisms from fungi to humans. In human and mouse, eIF5A2 shares 84% and 82% amino acid sequence identity with eIF5A1 (usually termed eIF5A). Both isoforms effectively undergo hypusine modification in cells (19). However, only eIF5A1 is constitutively expressed in all mammalian cells and tissues whereas the eIF5A2 isoform mRNA expression appears to be tissue-specific in brain and testis (20). The eIF5A2 protein is normally undetectable in most mammalian tissues and cells, but increased expression of this isoform or eIF5A has been associated with various human cancers (21-23). The Eif5a2 homozygous knockout mouse develops and grows normally, suggesting that eIF5A2 is dispensible for mouse development (24).
DHPS is known to be totally specific for eIF5A (eIF5A1 and the eIF5A2 isoforms); no other cellular protein is modified by DHPS. The exclusive specificity is based on the requirement for a macromolecular interaction between DHPS and the nearly intact N-domain of eIF5A. A potential role of eIF5A and DHPS in neuronal growth and survival was first suggested in studies that used the neuronal cell line PC12 and rat primary hippocampal cultures in vitro (25). In these studies, a reduction of hypusinated eIF5A by using a DHPS inhibitor or DHPS RNAi attenuated neurite outgrowth and neuronal survival (25). Only recently, definitive genetic evidence for their importance in human neurodevelopment was reported (26, 27). From whole genome exome sequencing and genetic analysis, biallelic DHPS variants were identified as the cause of a rare autosomal recessive neurodevelopmental disorder (27). More recently, germ line, de novo, heterozygous EIF5A variants were also reported to be associated with a neurodevelopmental disorder (26). The patients carrying biallelic DHPS variants, or heterozygous EIF5A variants, share common phenotypes including intellectual disability and developmental delay. Among the five DHPS variant patients, four have facial dysmorphism, one has microcephaly and four have clinical seizures. Of the seven EIF5A variant patients, all display facial dysmorphism and five of them with microcephaly. Thus, a decrease in the biologically active, hypusinated form of eIF5A appears to interfere with proper neurodevelopment.
To further investigate the roles of eIF5A and DHPS in brain development, we have generated four mouse strains in which either Eif5a or Dhps is deleted in a temporally and spatially specific manner using the Emx1-Cre or the Camk2a-Cre line. Phenotype analyses revealed severe morphological defects in the brain, growth retardation and reduced viability in mice with Emx1-Cre mediated deletion of Eif5a or Dhps, and impaired cognitive functions in mice with Camk2a-Cre mediated deletion of Eif5a or Dhps.
Results
Generation of four conditional knockout strains: Eif5afl/fl;Emx1-Cre (Eif5aEmx), Dhpsfl/fl;Emx1-Cre (DhpsEmx), Eif5afl/fl;Camk2a-Cre (Eif5aCamk2a) and Dhpsfl/fl;Camk2a-Cre (DhpsCamk2a)
Emx1-Cre mediated knockout of Eif5a or Dhps was achieved by two-step breeding. First, Eif5a fl/fl (8) or Dhps fl/fl (7) mouse was mated with Emx1-IRES-Cre mouse (28) to generate either Eif5a fl/+;Emx1-Cre or Dhps fl/+;Emx1-Cre mouse, which was mated again with mice carrying their respective homozygous floxed allele to produce either Eif5a fl/fl; Emx1-Cre or Dhps fl/fl;Emx1-Cre mice. Camk2a-Cre mediated knockout of Eif5a or Dhps was achieved as above by the two-step breeding, using the Camk2a-Cre transgenic strain T29-1 (29). These four conditional knockout (CKO) mice are referred to as Eif5aEmx, DhpsEmx, Eif5aCamk2a and DhpsCamk2a, in the rest of the paper. The genotypes of the CKO strains were confirmed by PCR as shown in Fig. S1.
The effects of temporal- and region-specific knockout of Eif5a or Dhps in the brain on growth and survival of mice
In the Emx1-Cre driven knockout strains, the Eif5a or Dhps gene is downregulated in the neurons of developing rostral brain including the cerebral cortex, and hippocampus, beginning at E9.5 and continuing throughout postnatal life. On the other hand, in the Camk2a-Cre driven CKO strains, the expression of the target gene is abolished postnatally (beginning at P15 – P 21 and continuing through adulthood) in the Camk2a expressing neurons in the CA1 regions of hippocampus (29). Differential phenotypes were observed in all four CKO strains. Both male and female groups of Eif5aEmx pups grew significantly slower than the control Eif5afl/fl pups (Fig. 2 A and B). There was little difference in the growth rates between the male and female groups of the Eif5aEmx mice, whereas in the control Eif5afl/fl group, the males were consistently heavier than the female counterparts (Fig. 2 A and B). Moreover, survival was reduced in Eif5aEmx mice compared with control (Fig. 2C). The average body weights of both the male and female Eif5aEmx mice was reduced compared to the control Eif5afl/fl mice throughout the period examined (Fig. 2 A, B, D, E).
At birth, the DhpsEmx mice appeared to be similar in size to control mice (Dhpsfl/+, Emx1-cre, Dhpsfl/fl, Dhpsfl/+). However, the postnatal growth of DhpsEmx mice was significantly impaired (Fig. 3A) and nearly arrested by day 12, while the control mice continued to grow. All DhpsEmx pups died before 4 weeks after birth (Fig. 3B). On day 24, DhpsEmx mice were much smaller than Dhpsfl/fl mice with the average whole body weight less than 50% of the control mice (Fig. 3C).
Unlike the deletion of Eif5a or Dhps in the Emx1 expressing neurons, deletion of either gene in the Camk2a expressing neurons did not result in significant inhibition of growth and no visible signs of developmental defects were observed in the first 3 months. However, both Eif5aCamk2a and DhpsCamk2a mice lost viability between 2-9 months of age (Fig. S2).
The effects of deletion of Dhps or Eif5a on brain development and morphology
The deletion of Eif5a or Dhps also exerted variable impacts on brain development in the four CKO strains (Figs. 4 and 5). Gross brain images revealed quite similar morphological defects in the brains of Eif5aEmx and DhpsEmx mice (Fig. 4 A and C), even though DhpsEmx mice displayed more serious defects in growth and survival than Eif5aEmx mice. The average brain weight of the Eif5aEmx mice at 4 months was less than that of controls (0.25 g vs 0.48 g, respectively). The same gross lesions shown in Fig. 4A were observed in all Eif5aEmx brains (1 and 4 month old mice). The abnormal brain morphology included the loss of the cerebral cortex, hippocampus, corpus collosum, internal capsule and portions of the lateral ventricles and the opening of the third ventricle to the meninges. However, we could not detect cellular changes in the microscopic images of the remaining part of the Eif5aEmx brain at 4 months (Fig. 4 I vs J). The average weight of the DhpsEmx brains was less than half of the control brain (0.19g vs 0.431g) on day 24. Each of the four DhpsEmx brains examined showed the same gross abnormality (Fig. 4C), similar to that of Eif5aEmx brain (Fig. 4A). In the DhpsEmx brain, the rostral portion of the cerebral cortex was missing or thinned. The deformity also included agenesis of the corpus collosum, hippocampus, internal capsule and the distal portion of the cerebrum overlying the mid-brain. The lumen and the roof of the third ventricle were missing. Microscopic images of remaining DhpsEmx brain cerebrum showed the neurons enlarged and vesiculated (black arrows, Fig. 4 K), not found in the control brain (Fig. 4 L).
In contrast to the Eif5aEmx and DhpsEmx mice, Eif5aCamk2a and DhpsCamk2a mice appeared to grow normally and their gross brain images were indistinguishable from those of control mice (Fig. S3). However, microscopic examination revealed that DhpsCamk2a mice had extensive neuronal necrosis of the cerebral cortex and hippocampus at 4 months (Fig. 5 C, G and K). These regions contained necrotic neurons (black arrows) and neurons with enlarged nuclei with extensively vesiculated chromatin (white arrows, Fig 5 G and K). Similar cellular changes were not observed in the Eif5aCamk2a brains (Fig. 5 E and I).
Impaired cognitive functions in the Eif5aCamk2a and DhpsCamk2a mice
We first examined the effects of deletion of Eif5a or Dhps in Eif5aCamk2a and DhpsCamk2a mice on cognitive functions by the Morris water maze test (MWM) (Fig. 6). In this test, the mouse relies on visual cues to navigate to a submerged escape platform. Spatial learning was assessed by daily repeated trials for six days. The latencies to reach the hidden platform for the two controls, Eif5afl/fl and Dhpsfl/fl, were 37.02 and 38.61 sec, respectively, on day 1 and were shortened to 9.52 and 9.93 sec, respectively, by day 6 (Fig. 6 A and B). On the other hand, the latencies of Eif5aCamk2a and DhpsCamk2a mice on day 1 (42.71 and 53.66 sec, respectively) were longer than those of the respective controls suggesting a poor baseline performance. Furthermore, the improvements of Eif5aCamk2a and DhpsCamk2a mice from day 1 to day 6 (reduction of latency by 53% and 38.5%, respectively) were much less than those of the respective control mice (reduction of latency by 74.3% and 75.1%, respectively), suggesting impaired learning in both CKO mice, with DHPSCamk2a showing greater impairment than EIF5ACamk2a. The analyses of swim distance to the hidden platform also provided a similar indication of learning disability of the two CKO strains (Fig. 6 C and D). The swim distances were similar for all four groups on day 1, but were significantly longer for the Eif5aCamk2a and DhpsCamk2a mice than their respective controls on consecutive days. The improvement indicated by a shortening of the swim distance from day 1 to day 6 was worse for the CKO groups compared to their controls and DhpsCamk2a consistently underperformed Eif5aCamk2a mice. These results provide strong evidence that both Eif5aCamk2a and DhpsCamk2a mice are impaired in spatial learning and that the impairment is more severe in DhpsCamk2a than in Eif5aCamk2a mice (Fig 6. A, B, C and D).
After the completion of the six day trials, reference memory was evaluated by a probe trial after the removal of the hidden platform. The % time occupancy in the target quadrant (NW) (Fig. 6 E and F) and and the number of crossings into the small area that previously contained the removed platform were measured (Fig.6 G and H). The Eif5afl/fl and Dhpsfl/fl control groups occupied the target NW quadrant for a significantly longer time, (37.04% and 37.21% time in quadrent, respectively) than in three other quadrants (∼ 20 % in each quadrant). In contrast, the preference to occupy the target quadrant was significantly reduced in Eif5aCamk2a mice compared to the controls (30.18% vs 37.21%) and in DhpsCamk2a mice compared to the controls (23.20% vs 37.04%), suggesting the impaired memory in both CKO groups. The average numbers of crossings into platform area were lower in the CKO mice than in the controls (7.42 and 4.94, respectively, for Eif5afl/fl and Eif5aCamk2a mice and 6.29 and 2.08, respectively, for Dhpsfl/fl and DhpsCamk2a mice). Both the platform occupancy and the platform area entry data provide clear evidence for memory impairment in the CKO mice, DhpsCamk2a mice being more deficient than Eif5aCamk2a mice.
Then a contextual learning (cued fear conditioning) test was carried out as outlined in the top panel of Fig. 7. The baseline freezing and novel context baseline freezing were low and no significant differences were observed among the four groups of mice. However, contextual freezing time was significantly reduced in Eif5aCamk2a mice (to 48 % of the control Eif5afl/fl value) and DhpsCamk2a (to 40% of Dhpsfl/fl value) (Fig. 7 A and B). Auditory cued freezing was also reduced in Eif5aCamk2a mice (to 64% of the control) and in DhpsCamk2a mice (to 78% of the control), but not as much as the contextual freezing. Taken together, the data in Fig. 6 and Fig. 7 clearly demonstrate the impairment in spatial learning, memory and contextual learning in mice in which Eif5a or Dhps is deleted in the Camk2a expressing neurons of cortex and hippocampus.
Discussion
Recent genetic studies have provided evidence that certain variants in the EIF5A or DHPS gene are associated with rare neurodevelopmental disorders in humans. Furthermore, individuals with DOHH variants who display similar developmental delay and intellectual disability have also been identified (Ziegler A. et al, Unpublished results), underscoring the importance of each step of the hypusine modification pathway and thereby the critical role of the hypusinated eIF5A in neurodevelopment in humans. These findings prompted us to generate the four CKO mouse models, with temporal- and region-specific deletion of either Eif5a or Dhps in the forebrain and to assess the impact on development. Different phenotypes in brain development, growth, survival and cognitive functions were observed in these CKO strains depending on the targeted gene and the Cre-driver. Although the CKO strains do not harbor the same variants of the EIF5A and the DHPS genes as those of the affected human individuals, it is interesting to note that certain features of the human neurodevelopmental disorders, including intellectual disability, developmental delay, reduced growth, and shortened lifespan, are reflected in the phenotypes of these CKO mice.
The deletion of Eif5a or Dhps in the Emx1-expressing neurons from the mid-embryonic stage resulted in gross morphological abnormalities in brain; cerebral cortex was thinned or missing and hippocampus, corpus collosum and the internal capsule portions of the ventricles were also missing due to agenesis (Fig 4, A and C). These results indicate that both Eif5a and Dhps are essential for the embryonic and postnatal development of cortex and hippocampus. Although gross brain defects were similar in the Eif5aEmx and the DhpsEmx mice, the deleterious effects on growth and viability were more severe in DhpsEmx mice than in Eif5aEmx mice; Approximately 67% of Eif5aEmx mice survived longer than 3 months whereas all DhpsEmx mice died within four weeks after birth. The behavioral tests were not performed on these groups of CKO mice because of their short life spans and premature and unpredictable death, especially of DhpsEmx mice.
The postnatal ablation of Eif5a or Dhps in the Camk2a-expressing neurons did not cause gross changes in the brain compared with those found the Eif5aEmx and the DhpsEmx mice, suggesting that the development of cortex and hippocampus was unaltered. Although no significant growth inhibition or visible defects were found in Eif5aCamk2a or DhpsCamk2a mice, they both died prematurely between 2-9 months. Furthermore, they displayed concrete evidence of impairment in cognitive functions (Figs. 6 and 7), DhpsCamk2a being more affected than Eif5aCamk2a mice. We speculate that this is due to compromised neuronal activities in the hippocampus. Although the genetic alterations in these CKO mice are different from those in the affected human individuals, these cognitively impaired CKO mice hold potential utility in the future development of chemical or biological therapeutics for human neurodevelopmental disorders caused by variants of EIF5A, or DHPS.
Common phenotypes between Eif5aEmx and DhpsEmx mice, and between Eif5aCamk2a and DhpsCamk2a mice strongly suggest that a common pathway underlies the impairment in both CKO mice. However, it is hard to explain why the ablation of Dhps is more detrimental than that of Eif5a, as is evident in all the observed phenotypes. This is counter-intuitive, as the hypusinated eIF5A is the direct player in translation elongation, whereas DHPS is a modifier of eIF5A activity. One possibility may be that eIF5A2 isoform (modified to the hypusine form) can partially compensate for the loss of eIF5A in the targeted neurons of the mouse brain. However, we did not find clear evidence for accumulation of eIF5A2 isoform protein in brain tissues of control or CKO mice (data not shown). Furthermore, commercial eIF5A2 antibodies often crossreact with eIF5A, making it difficult to detect a very low level of eIF5A2 in the presence of abundant eIF5A. The whole-body knockout of eIF5A is embryonic lethal in mouse, suggesting that eIF5A2 is not induced upon knockout of eIF5A and that eIF5A2 cannot compensate for the loss of eIF5A during early embryonic development (6). Another possibility is the interference of eIF5A(Hpu) activities by unhypusinated eIF5A precursors that accumulate upon depletion of spermidine or upon inhibition of DHPS. Although the unhypusinated eIF5A precursors were inactive and did not appear to interfere with the activity of hypusinated eIF5A in the methionyl-puromycin synthesis in vitro (30), their potential effects on translation need to be reevaluated in vivo. It is possible that the eIF5A precursors still associate with the 80S ribosome through interactions not involving the hypusine residue (18) and interefere with the action of hypusinated eIF5A in mammalian cells and tissues. In such a case, the potential interference by the eIF5A precursors that may have accumulated in DhpsEmx and DhpsCamk2a brains could explain their more deleterious phenotypes. In the case of human patients, a heterozygous de novo EIF5A variant with partial activities causes clinical phenotypes (26), suggesting that proper neuronal function in humans cannot tolerate even a partial loss (<50%) of eIF5A activity. The detrimental effects of heterozygous EIF5A variants may not be simply due to a reduction in active eIF5A, but may also be compounded by the interference by the eIF5A variants. The molecular basis underlying the better survival and performance of the Eif5a CKO mice than the Dhps CKO mice warrants further investigation.
The implication of variants of EIF5A or DHPS in human neurodevelopmental disorders is not surprising in view of the fact that variants in a number of other factors in the translation machinery such as alanyl tRNA synthetase (AARS) and eukaryotic translation elongation factors 2 (EF2) and 1a2 (EF1a2) have been associated with neurodevelopmental disorders (31). Errors during translation elongation can lead to production and accumulation of aberrant proteins that are toxic to neural cells. In human individuals with variants in EIF5A, or DHPS, major clinical symptoms were associated with neurodevelopment (26, 27), suggesting that neuronal systems are most vulnerable to a deficiency of hypusinated eIF5A. Global proteomics analyses provided evidence that depletion of eIF5A in mammalian cells led to endoplasmic reticulum stress, unfolded protein response and upregulation of chaperone expression (32). In addition to these general effects of eIF5A depletion, it is also possible that there are key regulatory factors of brain development that may be specifically dependent on eIF5A for translation. Future studies will be directed towards elucidation of molecular mechanisms underlying these neurodevelopmental disorders stemming from a reduction in bioactive, hypusinated eIF5A and the identification of downstream effectors of eIF5A.
Materials and Methods
Mouse strains used and detailed methods on mouse maintenance, genotyping, histochemical analysis, Morris water maze test and the contextual learning test are described in SI Appendix, Materials and Methods.
Author Contributions
M.H.P. designed research, R.K.K, A.S.H, M.F.S. and D.S. performed experiments and/or analyzed the data. T.L.M. and R.G.M. provided the Eif5afl/fl and Dhpsfl/fl mice and expertise. M.H.P. wrote the paper with input from coauthors.
Acknowdgements
We thank Dr. Edith Wolff, Dr. Roman Szabo and Andrew Cho (National Institute of Dental and Craniofacial Research, National Institutes of Health) for helpful advice and suggestions, and Dr. Hans Johansson (Biosearch Technologies, CA) for critical reading of the manuscript, and the National Institute of Dental and Craniofacial Research Veterinary Resource Core for excellent care of mice, and the National Heart, Lung and Blood Institute Murine Phenotyping Core and Morteza Peiravi for conducting the behavioral tests.
This work was supported by the intramural research program of the National Institute of Dental and Craniofacial Research, National Institutes of Health (M.H.P.), grants from the Juvenile Diabetes Research Foundation 5-CDA-2016-194-A-N (T.L.M) and NIH 1R01DK121987-01A1 (T.L.M.), and NIH R01 DK060581, R01 DK125906, and R01 DK105588 (all to R.G.M.).
The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.