use prefix []or [-]not [+]and [=]has feature [!]exclude feature ie. 'interleukin-6 -animal +phenotypic =protein !tumor'

Displaying 2 papers, 1 pages, start at 1, 51 Hits
23 section matches

Abstract

TP53 gene is known as the "guardian of the genome" as it plays a vital role in regulating cell cycle, cell proliferation, DNA damage repair, initiation of programmed cell death and suppressing tumor growth. Non uniform usage of synonymous codons for a specific amino acid during translation of protein known as codon usage bias (CUB) is a unique property of the genome and shows species specific deviation. Analysis of codon usage bias with compositional dynamics of coding sequences has contributed to the better understanding of the molecular mechanism and the evolution of a particular gene. In this study, the complete nucleotide coding sequences of TP53 gene from eight different mammalian species were used for CUB analysis. Our results showed that the codon usage patterns in TP53 gene across different mammalian species has been influenced by GC bias particularly GC 3 and a moderate bias exists in the codon usage of TP53 gene. Moreover, we observed that nature has highly favored the most over represented codon CTG for leucine amino acid but selected against the ATA codon for isoleucine in TP53 gene across all mammalian species during the course of evolution.

Introduction

TP53 gene encodes tumor protein p53 which is known as the "guardian of the genome" as it plays a vital role in maintaining genomic stability by preventing mutation in the genome [1] . The p53 primarily acts as transcription factor and stands out as a key player in restricting tumor cell invasion that includes the ability to induce cell cycle arrest, DNA repair, senescence and apoptosis [2] . Mutation in p53 results in abnormal proliferation of cells that leads to the formation of tumor development and so TP53 gene is cataloged as tumor suppressor gene [3] .
The nucleus of a cell is the main store house of tumor protein p53 where it binds to DNA. When any damage occurs in the DNA of a cell by some external agents like toxic chemicals, radiation, exposure to sun light or ultra violet rays, p53 plays the crucial role in activating other genes and inhibits cell cycle to repair the damage [4] . In case of failure of DNA repair, the tumor protein p53 prevents the cell from dividing and provokes signals to a wide variety of genes that contribute to TP53 mediated cell death i.e., apoptosis [5] .
The present study was undertaken in order to perform a comparative analysis of codon bias and compositional dynamics of codon usage patterns in TP53 gene across eight different mammalian species using nucleotide chemistry (GC contents) and several genetic indices namely effective number of codons (ENC), relative synonymous codon usage (RSCU), and relative codon usage bias (RCBS) etc. Our analysis has given a novel insight into the codon usage patterns of TP53 gene that would facilitate better understanding of the structural, functional as well as evolutionary significance of the gene among the mammalian species.

Results and Discussion

Codon usage patterns in TP53 genes across mammalian species Correlation coefficient between codon usage and GC bias was analyzed using heat map ( Fig. 1 ) in order to find out the relationship between the codon usage variation and the GC constraints among the selected coding sequences of TP53 genes. In our analysis, nearly all codons ending with G/C base showed positive correlation with GC bias and nearly all A/T-ending codons showed negative correlation with GC bias. But, 8 G/C-ending codons (ATC, ACG, TAC, TTG, TCC, CAC, GTG, GGG) showed negative correlation with GC bias whereas 6 A/T-ending codons (AAT, ATT, TGT, CGA, GTA, GGA) showed positive correlation with AT bias although statistically not significant (p>0.05). Two G-ending codons i.e. TCG for serine and CTG for leucine amino acid showed strong positive correlation (p<0.01) with GC 3s , indicating that codon usage has been influenced by GC bias due to GC 3s . Interestingly, we observed that the codon ATA encoding isoleucine amino acid was not favored by natural selection in TP53 genes across mammalian species during the course of evolution. Thus, scanning the codon usage pattern provides the basis of the mechanism for synonymous codon usage bias and has both practical as well as theoretical significance in gaining clues of understanding molecular biology [34] .

G/C-ending codons are favored by TP53 gene across mammalian species

We analyzed the nucleotide composition of coding sequences from TP53 genes ( Table 1 ) which revealed that mean value of C (361.50) was the highest followed by G (306.75), A (270.88) and T (227.50) among all the selected mammals. The mean percentage of GC and AT compositions was 57.3% and 42.7% respectively. Thus, the overall nucleotide composition suggested that the nucleotide C and G occurred more frequently compared to A and T in the coding sequences of TP53 gene across the mammalian species. The nucleotide composition at the third position of codon (A 3 ,T 3 ,G 3 ,C 3 ) showed that the mean values of C 3 and G 3 were the highest followed by T 3 and A 3 . The GC 3 values (ranged from 58.7%-70.6%, mean = 65.1%, SD = 0.040) was compared with that of AT 3 values (ranged from 29.4%-41.3%, mean = 34.9%, SD = 0.040) in the coding sequences of TP53 genes. The average percentage of GC contents at the first and second codon positions (GC 12 ) was found in the range of 52.6% to 54.9% with a mean value of 53.4% and a standard deviation (SD) of 0.008. Therefore, nucleotide composition analysis suggested that GC-ending codons might be preferred over AT-ending codons in the coding sequences of TP53 genes across the selected mammalian species. Further, we calculated the occurrence of frequently used optimal codons (Fop) for each amino acid as suggested by Lavner and Kotler (2005) [14] . The frequency was allied with statistical analysis to find out the highest and lowest frequently used codon. Our results showed that the most frequently used codons were G/C-ending for the corresponding amino acid (Fig. 2) in TP53 genes across mammalian species.

Relative synonymous codon usage in TP53 gene across mammals

The relative synonymous codon usage values of 59 codons for TP53 gene across eight mammalian species were analyzed excluding the codons ATG (methionine) and TGG (tryptophan). In our calculation RSCU value greater than 1.0 represents that the particular codon is used more frequently and less than 1.0 represents the less frequently used codon for the corresponding amino acid. The RSCU value greater than 1.6 indicates over represented codon for the corresponding amino acid. The overall RSCU values in the selected coding sequences of TP53 gene revealed that 25 codons were most frequently used among the 59 codons and the most predominantly used codons were G/C-ending compared to A/T-ending (Table 2) . Besides, it was observed that C-ending codon was mostly favored compared to G-ending codon in the coding sequence of TP53 gene among the selected mammalian species. Our results showed marked similarities as reported by Dass et al., (2012) in serotonin receptor gene family from different mammalian species [35] . Further, clustering analysis of RSCU values (Fig. 3) depicted that the codon GCC, CGC (except Rattus norvegicus), ATC (except Tupaia chinensis), CTG, ACC (except Rattus norvegicus, Macaca mulatta), GTG (except Felis catus) were displayed as the over represented codons (RSCU>1.6). The highest RSCU value was found for the codon CTG for leucine amino acid in all TP53 genes across mammalian species. The codon ATA showed the RSCU value zero because natural selection has not favored this codon in TP53 gene across mammalian species.
Codon usage patterns of TP53 gene correspond to phylogeny of mammalian species
We have performed a neighbor joining tree analysis based on Kimura 2-parameter (K2P) distances of the coding sequences in TP53 gene across mammalian species (Fig. 4) . We observed that codon usage patterns in TP53 genes have significant similarities among the closely related mammalian species. The gene TP53 in H. sapiens showed resemblance to the TP53 gene in M. mulatta, Similarly, TP53 of F. catus resembled to that of C. lupus and M.unguiculatus with R. norvigicus. Generally, genes with similar functions exhibit similar patterns of codon usage frequency [36] . Our analysis further suggested that the coding sequence of TP53 gene share similar patterns of codon usage bias across eight mammalian species.

Selection pressure over TP53 gene across mammalian species

The ENC values of the coding sequences ranged from 52 to 59 with a mean of 55.5±2.33 indicating relatively smaller variation in the codon usage of TP53 gene across eight mammalian species. However, the GC 3s values ranged from 0.59 to 0.71 with a mean value of 0.65±0.040. Significant negative correlation (Pearson r = -0.979, p<0.01) was observed between ENC and GC 3s . Moreover, a plot of ENC vs GC 3s revealed that the ENC values had negative correlation with the GC 3 content (Fig. 5 ) and comparatively lower ENC was linked to higher GC 3s values. All the selected coding sequences of TP53 gene across the selected mammalian species had Overall frequency of optimal and non optimal codon used in TP53 genes among mammals. Red color coding represents optimal used codons with corresponding amino acid.
doi:10.1371/journal.pone.0121709.g002 a higher predominance of G/C-ending codons. It suggested that GC 3s values determined the codon usage pattern in the coding sequences of TP53 gene [33] . Nabiyouni et al., (2013) reported that eukaryotic organisms with very high GC-contents have high GC 3 -composition while organisms with low GC-content have low GC 3 -composition in the genome [37] . We also calculated GC 3 skew values which ranged from 0.000 to -0.094, indicating that GC 3 composition at the third position of codon might have played an important role in the codon usage bias [38] . Negative GC skew was observed in all the coding sequences of TP53 gene which revealed that the abundance of C over G [39] . In addition, lower values of the frequency of optimal codons (FOP) and the effective number of codons (ENC) along with higher GC contents suggested that a moderate bias exists in the usage of synonymous codons [33] for TP53 gene in different mammalian species. Predominant codon usage bias was observed in TP53 gene of M.unguiculatus compared to other mammalian species (Table 3) . RCBS value of a gene can be used as an effective measure of predicting gene expression and its value depends on the patterns of codon usage along with nucleotide compositional bias of a gene [20] . The distribution of RCBS values for TP53 gene across eight mammalian species is shown in figure below (Fig. 6) . The RCBS values ranged from 0.006 to 0.065 with a mean value of 0.039 and a standard deviation (SD) of 0.021. In our analysis, low mean RCBS value suggested that there exists a low codon bias for TP53 gene associated with low expression level [20] .

Conclusions

In brief, our results showed that codon usage in TP53 gene in mammals has been influenced by GC bias, mainly due to GC 3s . The majority of frequently used codons were G/C ending in which C-ending codons were mostly favored compared to G-ending codons for the corresponding amino acid. The most over-represented codon was CTG encoding the amino acid leucine in the TP53 gene of all the selected mammalian species. We further observed that the codon ATA encoding isoleucine was selected against by nature in TP53 genes across the mammalian species under study during the course of evolution. The codon usage pattern for TP53 in H. sapiens showed resemblance to that of M. mulatta; similarly, F. catus to C. lupus and M. unguiculatus to R. norvigicus. Moderate codon bias was observed for the TP53 gene in different mammalian species. The codon usage patterns in the coding sequence of TP53 gene across different mammalian species showed significant similarities, suggesting that the evolutionary pattern might be similar. According to Yang and Nielsen (2008) , codon bias in mammals is mainly influenced by mutation bias and the selection on codon bias is weak for nearly neutral synonymous mutations [40] . From the outstanding work of Grantham et al., (1980 Grantham et al., ( -1981 on "genome hypothesis" it was evident that species specific genes share similar spectrum of codon usage frequency [41, 42] . The present study revealed that specific gene of closely related species with similar functions exhibit similar patterns of codon bias across different mammals as evident from the previous work of Dass et al., (2012) [35] . To the best of our knowledge, this is the first report on the codon usage pattern in TP53 gene across the mammalian species. Since our analysis has given better insights into the codon usage, it may have theoretical value in further understanding the molecular evolution of TP53gene.

Sequence Data

The complete nucleotide coding sequences (cds) for TP53 gene having perfect start and stop codon, devoid of any unknown bases (N) and perfect multiple of three bases, were retrieved from National Center for Biotechnology Information (NCBI) GenBank database (http://www. ncbi.nlm.nih.gov). Finally, we selected eight coding sequences for TP53 gene that fulfill the above mentioned criteria in different mammalian species and used in our CUB analysis (Table 4 ).

Software Used

The above mentioned genetic indices were estimated in a PERL program developed by SC (corresponding author) to measure the CUB on the selected coding sequences of TP53 genes in different mammalian species. All statistical analyses were carried out using the SPSS software. Cluster analysis (Heat map) of correlation coefficient of codons with GC3 and the RSCU values of codons among the eight mammalian species were clustered using a hierarchical clustering method implemented in NetWalker software [48] .

Effective Number of Codons (ENC) Analysis

ENC is generally used to quantify the codon usage bias of a gene that is independent of the gene length and number of amino acids [43] . This measure was computed as per Wright (1990) to estimate the extent of CUB exhibited by the coding sequences of TP53 gene across the selected mammalian species:

Relative Synonymous Codon Usage (RSCU) Analysis

RSCU is defined as the observed frequency of a codon divided by the expected frequency if all codons are used equally for any particular amino acid [46] . RSCU values of codons for each of the selected coding sequence of TP53 gene was calculated as follows:

Computation of Gene Expression

Gene expression was estimated through RCBS which can be defined as the overall score of a gene indicating the influence of relative codon bias (RCB) of each codon in a gene [20] . The RCBS value of each coding sequence of TP53 gene was calculated as follows:

Introduction

Unequal usage of synonymous codons that encode the same amino acid during translation of a gene into protein is known as codon usage bias (CUB). Some codons in a synonymous group are used more frequently whereas others less frequently in the genome of an organism [6, 7] . CUB is a unique property of the genome and it may vary between genes from the same genome or within a single gene [8, 9] .
The study of codon usage bias acquires significance in biology not only in the context of understanding the process of evolution at molecular level but also in designing transgenes for increased expression, discovering new genes [29] based on nucleotide compositional dynamics, detecting lateral gene transfer and for analyzing the functional conservation of gene expression [30] . Codon usage bias may be superimposed on the effect of natural selection. The amount of protein produced from the mRNA transcript may vary significantly since the translational properties of alternate synonymous codons are not equivalent [31] . Several studies have further shown that codon usage bias is associated with highly expressed genes as some codons are used more often than others in the coding sequences [32] . Moreover, literature suggested that a gene can be epitomized not only by the sequence of its amino acid but also by its codon usage patterns shaped by the balance between mutational bias and natural selection [33] . As a consequence of selection pressure within a gene, differentiation in codon bias may arise between species of the same genus.

Effective Number of Codons (ENC) Analysis

Where, F k ( k = 2, 3, 4 or 6) is the average of the F k values for k-fold degenerate amino acids. The F value denotes the probability that two randomly chosen codons for an amino acid with two codons are identical. The values of ENC ranged from 20 indicating strong codon bias in the gene using only one synonymous codon for the corresponding amino acid, to 61indicating no bias in the gene using all synonymous codons equally for the corresponding amino acid [43] .

Frequency of Optimal Codon (Fop) Analysis

Fop is a measure of codon usage bias in a gene [44] . Fop values represent the ratio of the number of optimal codons used to the total number of synonymous codons [22] . The Fop value ranges from 0.36 for a gene showing uniform codon usage bias to 1 for a gene showing strong codon usage bias [45] . Fop value for each selected coding sequence was calculated using the formula given by Lavner and Kotler (2005) [14] .

Introduction

The advent of whole genome sequencing in different organisms and the easily accessible nucleotide database from NCBI (GenBank) have attracted much attention of the scientific community to study CUB in gaining clues for understanding the molecular evolution of genes and genome characterization.
Previously, several studies were conducted on synonymous codon usage bias in a wide variety of organisms including prokaryotes and eukaryotes [10] [11] [12] [13] [14] [15] [16] , and till date in many organisms the codon usage patterns have been interpreted for diverse reasons. Many genomic factors such as gene length, GC-content, recombination rate, gene expression level, or modulation in the genetic code are associated with CUB in different organisms [17] [18] [19] [20] [21] . In general, compositional constraints under natural selection or mutation pressure are considered as major factors in the codon usage variation among different organisms [8, [22] [23] [24] [25] . Moreover, studies revealed that mutation pressure, natural or translational selection, secondary protein structure, replication and selective transcription, hydrophobicity and hydrophilicity of the protein and the external environment play a major role in the codon usage pattern of organisms [26] . In unicellular and multicellular organisms it was observed that, preferred synonymous codons/optimal codons with abundant tRNA gene copy number rise with gene expression level within the genome that supports selection on high codon bias confirmed by positive correlation between optimal codons and tRNA abundance [18, 22, 27] . Urrutia and Hurst (2003) reported weak correlation between gene expression level and codon usage bias within human genome though not related with tRNA abundance [19] . However, Comeron (2004) observed that in human genome, highly expressed genes have preference towards codon bias favoring codons with most abundant tRNA gene copy number compared to less highly expressed genes [28] .
26 section matches

Abstract

We report here that in rat and human neuroprogenitor cells as well as rat embryonic cortical neurons Zika virus (ZIKV) infection leads to ribosomal stress that is characterized by structural disruption of the nucleolus. The anti-nucleolar effects were most pronounced in postmitotic neurons. Moreover, in the latter system, nucleolar presence of ZIKV capsid protein (ZIKV-C) was associated with ribosomal stress and apoptosis. Deletion of 22 C-terminal residues of ZIKV-C prevented nucleolar localization, ribosomal stress and apoptosis. Consistent with a casual relationship between ZIKV-C-induced ribosomal stress and apoptosis, ZIKV-C-overexpressing neurons were protected by loss-of-function manipulations targeting the ribosomal stress effector Tp53 or knockdown of the ribosomal stress mediator RPL11. Finally, capsid protein of Dengue virus, but not West Nile virus, induced ribosomal stress and apoptosis. Thus, anti-nucleolar and pro-apoptotic effects of protein C are flavivirus-species specific. In the case of ZIKV, capsid protein-mediated ribosomal stress may contribute to neuronal death, neurodevelopmental disruption and microcephaly.
Recent work has suggested that ZIKV protein NS4A and, to lesser extent, NS4B may impair NP proliferation by inhibiting the growth promoting AKT pathway 12 . In addition, Tp53 has been proposed to mediate ZIKV-induced apoptosis but its ZIKV-related activation mechanism remains obscure 13 . Thus, experiments are needed to determine which ZIKV proteins induce cell death of the developing brain cells and what the mechanisms behind their anti-survival effects in that cellular context are.
Congenital microcephaly is caused by insufficiency of neurogenesis that may originate from (i) depletion of neuroprogenitor cells (NPCs) due to their apoptosis and/or pre-mature differentiation, (ii) inhibition of NPC proliferation, and, (iii) apoptosis of newly generated neurons 4-6 . ZIKV has a high potential to infect human or rodent NPCs and to a lesser extent developing neurons 7-11 . ZIKV reduces NPC proliferation, induces their premature differentiation and activates apoptosis of NPCs and immature neurons. However, a question remains as to what are the mechanisms behind such cytotoxic effects of ZIKV.

Several lines of evidence suggest that RS may be involved in pathogenesis of neurodevelopmental diseases. First, RS activates RPL11/Tp53-dependent apoptosis of immature cortical neurons 29 . Second, acephalic mouse fetuses were produced when Pol1 was inhibited selectively in Nestin-positive NPCs 30 . Similarly, just a single injection of the anti-ribosomal drug 5-fluorouracil to pregnant rats induces microcephaly of the offspring 31 . Finally, recent description of the nucleolar proteome from the developing rat brain included 9 proteins whose human counterparts are mutated in microcephaly syndromes 32 . Several of these proteins are expected to participate in brain ribosomal biogenesis as verified experimentally for a newly identified RBF, LARP7 32 . Hence, RS may contribute to neuroteratogenic effects of various mutations, toxins and infectious agents by inhibiting neurogenesis and/or activating apoptosis of immature neurons. Therefore, the current study was initiated to examine role of RS in ZIKV-mediated damage of the developing brain.

Nucleolar localization of the ZIKV capsid protein (ZIKV-C) in neural cells.

Neuronal apoptosis in response to overexpressed ZIKV-C is mediated by the ribosomal stress pathway. In neurons that were transfected with Fl-ZIKV-C but not DN-Tp53 (Tp53-DD) a plasmid DNA dose-dependent increase of apoptosis was observed (Fig. 5a,b) . No apoptotic response was observed with other variants of ZIKV-C including C(anch) and C(1-73) as well as ZIKV-M (Fig. 5c) . Finally, pro-apoptotic effects of ZIKV-C were similar to those after knockdown of the Pol1 co-activator, TIF1A (Fig. 5c) . Thus, in case of ZIKV-C there is a good correlation between nucleolar localization, disruption of nucleolar NPM1 and ability to induce neuronal apoptosis.
As ribosomal stress-induced neuronal apoptosis involves pro-apoptotic activity of Tp53 and its ribosomal stress-specific regulator, RPL11 27 their role in neurotoxicity of ZIKV-C was investigated. Overexpression of Fl-ZIKV-C stimulated Tp53-driven transcription as demonstrated using a Tp53-driven luciferase reporter construct (Tp53-Luc, Fig. 5d ). Importantly, ability of the Tp53-Luc to detect Tp53 activity was confirmed using previously validated shRNAs against rat Tp53 39 . In neurons, such shRNAs reduced reporter activity under basal conditions or after treatment with the Tp53 activator nutlin (Supplementary Fig. S8 ).
When Fl-ZIKV-C was co-expressed with several dominant negative mutants of Tp53 (DN-Tp53) ZIKV-C-induced apoptosis was reduced (Fig. 5e ). While such protective effects suggest a role of Tp53 in ZIKV-C-induced apoptosis, one should note that many DN-Tp53 mutants may have gain of function activities beyond inhibition of wild type Tp53 (WT-Tp53) including cross inhibition of Tp53 relatives such as p63 or p73 and modulation of several other transcriptional regulators 40 . However, such gain of function effects are often mutant-specific 40 . Hence, similar inhibition of ZIKV-C-induced apoptosis by each of the three different DN-Tp53s that were tested here (two missense mutations: R174H /human/ and V135A/mouse/ and a miniprotein Tp53-DD that lacks amino acid residues 15-301 of mouse WT-Tp53) suggests that their protective effects were mediated by inhibition of WT-Tp53. Indeed, shRNAs against the RS-specific Tp53 regulator RPL11 or Tp53 itself abolished apoptotic response to ZIKV-C (Fig. 5f ).

Discussion

At least in cultured cells, role of RS in mediating cytopathic effects of ZIKV is supported by recent pharmacological as well transcriptomic evidence that Tp53 contributes to ZIKV-induced apoptosis 13 . However, it is conceivable that other components of cellular ZIKV response including activation of toll-like receptor signaling, inhibition of Akt, or centrosomal/mitochondrial pathology co-operate with RS to produce ZIKV-associated brain malformations 12, 35, 61 . Hence, future experiments are needed to determine relative neuropathogenic contribution of RS to ZIKV-induced neurodevelopmental disruption.

Materials.

Plasmids. The following plasmids were previously described: EF1α-LacZ (EF1α promoter-driven β-galactosidase), CMV-driven DN-Tp53s (R175H 68 , V135A 69 and Tp53-DD 70 ); shLuc 71 ; shL11 27 , chicken β-actin promoter-driven-EGFP-RPL4 53 ; PG13 (Tp53 response element-driven luciferase reporter plasmid) 72 , EF1α-Renilla-luciferase 73 . Synthetic cDNAs encoding flaviviral proteins were custom synthesized by BioBasic (NY, USA) and cloned into mammalian expression vector pBact-16-pl 74 downstream of the chicken β-actin promoter via HindIII-SpeI sites. These constructs included ZIKV strain BeH819966 mature capsid protein 1-104 (ZIKV-C), anchored capsid protein 1-122 (ZIKV-C/anch/), capsid deletion mutant 1-82 (ZIKV-C/1-82/), capsid deletion mutant 1-73 (ZIKV-C/1-73/), membrane glycoprotein M 1-75 (ZIKV-M), protein NS4A 1-127 (ZIKV-NS4A, GenBank genome accession: KU365779), DENV type 2 strain 'New Guinea C' mature capsid protein 1-101 (DENV-C, GenBank genome accession: KM204118), and WNV Type 1 A strain New York 99 mature core protein 1-105 (GenBank genome accession: HQ596519). All cloned proteins were tagged with 3xFlag tag on their N-terminal end; an additional ZIKV-C construct without any tag was also prepared. Two previously validated shRNA target sequences for rat Tp53 were cloned into pSuper shRNA expression vector (Supplementary Table S3) 39 . permeabilization in 0.5% NP-40/PBS and 1 h blocking in 5% goat serum, 0.1% Triton X-100/PBS. The following primary antibodies were used: mouse monoclonal anti-nucleophosmin-1/B23 antibody (1:750, Sigma), rat-anti Pes1 (clone 8E9, 1:200, HelmholtzZentrum Munchen, Core facility Monoklonale Antikorper, German Research Center for Environmental Health, Munich, Germany), humanized 4G2 monoclonal anti-Flavi-E (1:500, a gift from Dr. Nobuyuki Matoba, University of Louisville), rabbit anti-β-galactosidase (1:1000, MP), and rabbit anti-GFP (1:1000, MBL). The following secondary antibodies were used Alexa Fluor 488 goat anti-mouse IgG, Alexa Fluor 488 goat anti-rat IgG, Alexa Fluor 488 goat anti-rabbit IgG, Alexa Fluor 594 anti-human IgG, and Alexa Fluor 594 anti-rabbit IgG (all Invitrogen, in all cases dilution was 1:300).

Although flaviviruses replicate on the ER membrane 14, 15 , at least a fraction of some flaviviral proteins is found in the nucleus of the infected cells, including nucleolar presence of WNV-, JEV-, or Dengue virus (DENV) capsid protein (protein C) 16 . Being structural components of flaviviruses, capsid proteins are very abundant in the infected cells 17 . They interact with multiple host proteins including ribosomal biogenesis factors (RBFs) [18] [19] [20] [21] . Some of these interactions may be toxic to the infected cells. For instance, immature form of WNV-C was shown to cause nucleolar sequestration of the Tp53 inhibitor MDM2/HDM2 leading to Tp53-mediated apoptosis in a human cell line 22 .
The nucleolus, where flaviviral capsid proteins are often found, is a center of ribosomal biogenesis 23 . Dysregulation of ribosomal biogenesis triggers ribosomal stress (RS) 24 . In most cases, such a response relies on interactions between the Tp53 ubiquitin ligase MDM2/HDM2 and ribosomal components that are no longer incorporated into ribosomes 24 . Specifically, RPL11, RPL5, and 5S rRNA, acting as a complex (the 5S ribonucleoprotein particle /5S RNP/), bind and inhibit MDM2, stabilizing Tp53 and producing cell cycle arrest and/or apoptosis 25, 26 . When RNA-Polymerase-1 (Pol1), which initiates ribosomal biogenesis, is inhibited, nucleolar structure is disrupted 23 . Thus, loss of the granular component (GC) of the nucleolus including nucleoplasmic dispersion of its prominent protein components such as nucleophosmin-1 (NPM1/B23) or PES1 provides a convenient way of monitoring this type of RS 27 . However, the RS-Tp53 pathway may also be activated without structural disruption of the nucleolus and with normal-or increased activity of Pol1 27, 28 .

Nucleolar localization of the ZIKV capsid protein (ZIKV-C) in neural cells.

As capsid proteins of several non-ZIKV flaviviruses have a high potential to interact with host RBFs and localize to the nucleoli 16,18-21 , anti-nucleolar effects of ZIKV may be caused by ZIKV-C. To evaluate such a possibility, Flag (Fl)-tagged ZIKV-C was overexpressed in otherwise intact rat embryonic neurons. In these studies, a synthetic ZIKV-C cDNA gene was used to produce ZIKV-C with amino acid sequence identical to that of three ZIKV strains with direct links to microcephaly; it differs by 1 or 3 amino acid residues from PRV or MR766, respectively ( Supplementary Fig. S5 ). The dominant-negative mutant of Tp53 (DN-Tp53) was also co-expressed to protect cells from a potential RS and RS-mediated apoptosis that may have been expected in response to ZIKV-C. Most of the mature Fl-ZIKV-C (ZIKV-C) was present in the perikaryal cytoplasm (i.e. cell soma cytoplasm, Fig. 3a and Supplementary Fig. S6 ). However, in many cells a fraction of Fl-ZIKV-C was also found in nuclei including enrichment in NPM1-positive nucleoli ( Fig. 3a and Supplementary Fig. S6 ). Mostly cytosolic signal was observed for other variants of ZIKV-C including the immature, membrane anchored ZIKV-C (ZIKV-C/anch/) and C-terminal deletion mutants (ZIKV-C/1-73/ and ZIKV-C/1-82/) that lacked the C-terminal basic RNA binding domain and a portion of the α-helical domain 4 (α4, Fig. 3a ). Almost exclusively cytosolic signal was found in case of two other ZIKV proteins (NS4A and M, Fig. 3a ).
As currently there are no available antibodies against mature ZIKV-C it is unclear whether in ZIKV-infected neural cells, ZIKV-C is nucleolar. However, such localization is expected based on data from non-neuronal cells that were infected with related flaviviruses including JEV or DENV 37,38 . Nucleolar disruption by the overexpressed ZIKV-C. As compared to cells that were transfected with an empty cloning vector, cells receiving increasing doses of Fl-ZIKV-C plasmid DNA displayed progressive reductions in nucleolar NPM1 signal intensity as well as NPM1-defined nucleolar territory (Fig. 4a -c and Supplementary Fig. S7 ). However, complete loss of nucleolar NPM1 was not observed; average number of NPM1-positive nucleoli per cell was unaffected either (Fig. 4d) . Conversely, NPM1 signal in the nucleoplasm was increased (Fig. 4a,b) . When untagged ZIKV-C was overexpressed, similar reductions in nucleolar NPM1 staining were observed as with Fl-ZIKV-C (Fig. 4e) . These anti-nucleolar effects of ZIKV-C overexpression were similar to those after knockdown of the Pol1 co-activator TIF1A whose depletion is a well established stimulus for pro-apoptotic RS in neurons ( Fig. 4a and f-h) 29 . No significant effects on nucleolar morphology were observed after overexpression of ZIKV-C variants with reduced ability to reside in the nucleolus (Fig. 4f-h) . Moreover anti-nucleolar effects of ZIKV-C were specific ; to avoid apoptosis due to possible ZIKV-C-mediated ribosomal stress, an expression plasmid for a dominant negative mutant of Tp53 was also added (Tp53-DD, 0.15 μg plasmid DNA/3.5*10 5 cells). Representative images of co-immunofluorescence for Flag and the nucleolar marker NPM1 revealed strong perikaryal expression of all ZIKV proteins including predominantly cytosolic localization and nucleolar enrichment of a fraction of Fl-ZIKV-C in many but not all Fl-positive cells (more images of Fl-ZIKV-C immunofluorescence are presented in Supplementary Fig. S6. (b) Quantification of nucleolar enrichment of Fl-ZIKV proteins. To equalize apparent differences in expression efficiency between Fl-ZIKV-C and other constructs, a β-gal expression plasmid was added to the transfections to provide a consistent transfection marker (150 ng plasmid DNA/3.5*10 5 cells, all other components as described for panel (a), see text for more details). Nucleolar enrichment analysis was performed in β-gal/Fl-double-positive cells. Nucleolar enrichment was observed for Fl-ZIKV-C and, to lesser extent, for the immature, membrane anchored version of Fl-ZIKV-C (C/anch/). Nucleolar enrichment was rare for other constructs including C-terminal deletion mutants of ZIKV-C (C/1-73/ and C/1-82/). Data represent averages of 6 sister cultures from three independent experiments; NS, p > 0.05; *p < 0.05; **p < 0.01 (u-test). (c,d) hNPCs or SH-SY5Y cells were transfected with Fl-ZIKV-C (150 ng plasmid DNA/10 5 cells) and its localization was analyzed by Fl immunofluorescence 48 h later; nucleolar enrichment was confirmed by co-transfection of a nucleolar/ribosomal marker GFP-RPL4 (hNPCs, 150 ng plasmid DNA/10 5 cells) or co-immunostaining for NPM1 (SH-SY5Y cells). (e) Neurons were transfected as in (a), and treated with a Pol1-specific inhibitor, BMH21 as indicated. Nucleolar enrichment of both NPM1 and Fl-ZIKV-C was disrupted by BMH21 (nucleolar enrichment of Fl-ZIKV-C was present in 50.6 ± 0.6% or 15.8 ± 5.3% control-or BMH21-treated cells, respectively as determined in two independent experiments). Fig. 3b with β-gal or empty cloning vector (EV, pBact-16-pl) used as a transfection marker or a negative control, respectively; additional controls included shRNAs targeting Renilla luciferase (shLuc) or the Pol1 coactivator Tif1a (shTif1a, used as a positive control for RS); all analyses were performed 48 h after transfection. (a) Representative images of transfected (i.e. β-gal-positive neurons, arrows) that were co-immunostained for NPM1; 150 ng plasmid DNA of EV or ZIKV-C/3.5*10 5 cells or 300 ng plasmid DNA of shTif1a/3.5*10 5 cells were used. Fl-ZIKV-C or shTif1a reduced NPM1 signal in the nucleolus while increasing it in the nucleoplasm (more images are in Supplementary Fig. S7). (b,c) Plasmid DNA dose-dependent reduction of intensity-and territory of the nucleolar NPM1 signal (one-way ANOVA, factor DNA dose, p < 0.001). (d) Fl-ZIKV-C did not affect nucleolar number. (e) Similar reductions of NPM1 signal in the nucleolus after transfections of Fl-ZIKV-C or untagged ZIKV-C. (f-h) Plasmid dosage was as in (a). Nucleolar NPM1 signal was affected by ZIKV-C or shTif1a but not the membrane bound precursor form ZIKV-C(anch) or C-terminal deletion mutants of mature ZIKV-C or other ZIKV proteins (f,g); nucleolar number was reduced only by shTif1a (h). (i-j) In situ run on assay revealed reduction of nascent RNA signal in nucleoli of ZIKV-C-transfected neurons suggesting lower activity of Pol1. Two days after transfections (as in (a)), cells were incubated with 5-ethynyluridine (5-EU) for 2 h, fixed and 5-EU-labelled nascent RNA was detected using Click-It chemistry. Then, β-gal immunofluorescence was performed to identify transfected cells (arrows). In nucleoli of Fl-ZIKV-C-transfected neurons, whole nucleus-normalized accumulation of nascent RNA was moderately reduced (j); however, Pol1 inhibition may be potentially underestimated as average whole nucleus signal was also reduced (Supplementary as neither ZIKV-NS4A nor ZIKV-M affected neuronal nucleoli (Fig. 4f-h) . Finally, in situ run on assay revealed that such effects were associated with reduced nucleolar accumulation of nascent RNA suggesting lower activity of Pol1 (Fig. 4i,j, Supplementary Table S2 ). Although nucleolar accumulation of nascent RNA was visible in similar fractions of ZIKV-C-or empty vector-transfected cells (64.6-or 67.4%, respectively), relative intensity of the nucleolar signal was down by 27% (Fig. 4i,j, Supplementary Table S2) . Therefore, at least in post-mitotic neurons overexpression of ZIKV-C is sufficient to perturb nucleolar structure and reduce ribosomal biogenesis.
ZIKV-C-mediated disruption of the nucleolus is specific to post-mitotic neurons. Although in proliferating cells including hNPCs or SH-SY5Y a fraction of Fl-ZIKV-C was present in the nucleoli (Fig. 3c,d) , its overexpression had only relatively limited effects on nucleolar NPM1 staining ( Supplementary Fig. S9) . Moreover, there were no significant effects on nucleolar nascent RNA accumulation or Tp53-driven transcription ( Supplementary Fig. S9 ). Therefore, unlike in post-mitotic neurons, ZIKV-C overexpression appears to be insufficient to induce the RS-Tp53 pathway in proliferating cells.
In neurons, DENV-C, but not WNV-C, disrupts nucleolus and induces apoptosis. As ZIKV-related flaviviruses including DENV and WNV have a neurotropic potential one could wonder if their capsid proteins may engage similar neurotoxic mechanisms as ZIKV-C 1,41,42 . Both DENV-C and WNV-C are highly basic and have similar domains as ZIKV-C with ZIKV-C sequence identity at 39.4% or 46.2%, respectively ( Supplementary Fig. S5) 17 . Indeed, besides strong cytosolic presence, overexpressed Fl-tagged DENV-C or WNV-C were also observed in neuronal nucleoli (Fig. 6a) . However, both frequency of such cells as well as relative nucleolar enrichment as compared to the whole nucleus was greater for DENV-C than for WNV-C ( Fig. 6b and Supplementary Fig S10) . Moreover, while WNV-C had only minor effects on nucleolar NPM1 signal intensity but not nucleolar NPM1 territory, DENV-C reduced those parameters at least as much as ZIKV-C (Figs 4a-g and 6d-f ). However, there was no effect on number of nucleoli (Fig. 6g ). In addition, DENV-C, but not WNV-C, reduced nucleolar accumulation of nascent RNA (Fig. 6h,i, Supplementary Table S2 ). The reduction was moderate (0.75 fold empty vector control) but significant and was reminiscent of that observed with ZIKV-C (compare Figs 6i to 4j). These anti-nucleolar effects were correlated with DENV-C-induced neuronal apoptosis (Fig. 6j) . No increase in apoptosis was observed with WNV-C (Fig. 6j) . Finally, overexpression of DENV-C or WNV-C increased Tp53-driven transcription to 4.8-or 2-fold of empty vector control (Fig. 6k) . In WNV-C-transfected neurons, such a nucleolar disruption-unrelated activation of Tp53 may be due to direct inhibition of MDM2 by WNV-C 22 . However, it appears to be insufficient to activate neuronal apoptosis. Taken together, similarly to ZIKV-C, DENV-C, but not WNV-C, engages cytotoxic RS to induce apoptosis in postmitotic neurons. Hence, pro-RS potential of flaviviral capsid proteins appears to be virus species-specific and may be related to distinct interactions with host cell proteome.

Discussion

Viral interactions with host cell nucleolus are observed with many RNA viruses including positive strand RNA viruses such as Flaviviridae 43 . Thus, specific viral proteins including the capsid protein traffic through the nucleolus interacting with nucleolar proteins which may be recruited to support viral replication. For instance, NPM1, or, the RNA helicase DDX56, or, the RNA binding protein nucleolin bind to capsid proteins and contribute to viral particle production of JEV, WNV, or, DENV, respectively [19] [20] [21] . Consistent with such observations, mutations that target residues critical for viral protein localization to the nucleolus often compromise viral titers and give rise to attenuated viral strains as reported for JEV or the arterivirus porcine reproductive and respiratory syndrome virus (PRRSV) 37, 44 . One could also expect that such a hijacking of cellular RBFs could perturb cellular ribosomal biogenesis and induce RS leading to RS-mediated cytopathic effects including cell cycle arrest and/or Tp53-mediated apoptosis 24 . However, to the best of our knowledge, there were no prior reports of RS contribution to cytopathic effects of any virus. Hence, the current manuscript provides the first piece of evidence for such a role of RS as exemplified in ZIKV-infected neurons. increased Tp53-driven transcription as determined by activity of a co-transfected p53-driven firefly luciferase reporter plasmid. (e,f) As expected for ribosomal stress-mediated apoptosis that involves the RPL11-Tp53 pathway, ZIKV-C-transfected neurons were protected by co-transfection of DN-Tp53 variants, or previously validated shRNAs against Rpl11 27 , or rat Tp53 39 . The shTp53 plasmids also reduced activity of the Tp53-driven luciferase reporter in either vehicle-or nutlin-treated neurons confirming their ability to inhibit Tp53 in that system ( Supplementary Fig. S8 ). Note that as compared to a control cDNA expression vector (pcDNA3.1, (e)) control shRNA (shLuc, (f)), increased baseline apoptosis in non-ZIKV-C-transfected neurons (16.6% or 28.3%, respectively, p < 0.05, u-test). Due to such baseline increases, ZIKV-C was relatively less pro-apoptotic in neurons that received shLuc than pcDNA3.1 (1.3-vs. 1.6 fold, (f) vs. (e) ). Data represent averages ± SEM of 4 (b) or 6 (d) sister cultures from two independent experiments, or 6 sister cultures from three independent experiments (c,e,f); NS, p > 0.05; *p < 0.05; **p < 0.01 (u-test).
Interestingly, possible negative effects on ribosomal biogenesis including ZIKV-like nucleolar disruption were reported for several neuropathogenic RNA viruses. For instance, the poliovirus 3Cpro protease inhibits Pol1 co-activators such as upstream binding factor (UBF) and selectivity factor-1 (SL1) 45 . Schmallenberg virus is a representative of Orthobunyaviridae which has documented neuroteratogenic effects in ruminants; interestingly, its non-structural protein NS localizes to the nucleolus leading to nucleoplasmic translocation of NPM1 which suggests block of Pol1 46 . Similar redistribution of NPM1 was reported with the Newcastle virus which causes extensive apoptotic cell death in the brains of infected chickens 47, 48 . Therefore, as in the case of ZIKV-infected neurons, one could anticipate that at least these pathogens may induce RS-mediated cytopathic effects. (a-c) In neurons and SH-SY5Y cells nucleolar enrichment appeared to be stronger for DENV-C than WNV-C including higher fraction of cells displaying such an enrichment (b), and, higher fluorescence intensity (FI) in the nucleolus (FI quantifications in Supplementary Fig. S10). (d-g) In β-gal-positive neurons that were co-transfected with expression plasmids for β-gal and flaviviral Cs, DENV-C but not WNV-C reduced nucleoplasm-normalized NPM1 signal intensity in the nucleolus and NPM1-positive nucleolar territory (arrowheads in (d), black bars in (e,f)); arrows point a Fl-WNV-C overexpressing neuron. However, number of NPM1-positive nucleoli was unaffected (g). (h,i) In situ run on assay revealed that whole nucleus-normalized nucleolar accumulation of nascent RNA was reduced in DENV-C but not WNV-C-transfected neurons. In both cases, nascent RNA signal was also reduced in whole nuclei suggesting inhibition of extranucleolar transcription and/or lower uptake of 5-EU into cells (Supplementary Table S2 ). Hence, anti-Pol1 effects of DENV-C may be potentially underestimated. (j) Increased neuronal apoptosis in response to DENV-C-but not WNV-C. (k) Tp53 reporter assay was performed as for Fig. 5d . DENV-C and WNV-C activated Tp53-driven transcription; the activation was stronger with the RS-inducing DENV-C than WNV-C. Data represent averages ± SEM of 6-(b,j) or 9 sister cultures (k) from three independent experiments, or, at least 51 cells-from three independent experiments (e-g,i); NS, p > 0.05; *p < 0.05; **p < 0.01, ***p < 0.001 (u- test in (b,j,k) ; one-way ANOVA with Tukey's posthoc tests in (e-g,i) ).

Quantification of apoptosis.

Cell survival assay. Cell survival assay using luminometric ATP level measurements was performed in ZIKV-infected hNPCs as previously described 75 . Transcription assay. Dual luciferase reporter assay was performed after co-transfection of the p53-driven-Firefly luciferase construct PG13 and the EF1α-Renilla-luciferase plasmid. Luciferase activities were measured using commercial kits (Promega) and the Berthold Orion II luminometer as previously described 76 . Transcriptional activity was expressed as a Renilla-normalized luciferase activity.

Nucleolar disruption in ZIKV-infected neural cells.

Pathological evidence as well as data from mouse models indicate that postmitotic cortical neurons may become infected with ZIKV and are highly sensitive to ZIKV-induced apoptosis 3, 11, 34 . Therefore anti-nucleolar effects of ZIKV infection were evaluated in embryonic cortical neurons. Consistent with previously published cell culture data 7, 35 , neuronal infection rate was relatively low with 1.2 or 1.4% cells being Flavi-E-positive at 3 dpi with PRV or MR766, respectively (Fig. 2a) . However, infection with either strain of ZIKV was associated with increased neuronal apoptosis (Fig. 2a,b) . NPCs (rNPCs) or monolayer cultures of human iPSC-derived NPCs (hNPCs) were infected with ZIKV strains MR766 or PRVABC59 at MOI 0.1. The rNPCs were grown as neurospheres for 3 days post infection (dpi), dispersed and cultured as a monolayer for 16 h to enable microscopic analysis at a single cell level; the hNPCs were maintained in a monolayer culture. After fixation, co-immunofluorescence staining was performed for the ZIKV infection marker (the flaviviral E protein, Flavi-E) and the nucleolar marker nucleophosmin-1 (NPM1/B23); nuclear DNA was counterstained with Hoechst-33258. Additional NPC data on ZIKV infection and cytotoxicity are presented in Supplemenary Figs S1 and S2. (a) Representative images depicting non-apoptotic ZIKV-infected rNPCs (i.e. lacking apoptotic chromatin condensation). Dotted lines mark nuclear contours of these cells; note reduced fluorescence intensity (FI) of NPM1. (b) At least MR766 infection increased fraction of non-apoptotic cells without NMP1-positive nucleoli. (c,d) Quantification of NPM1 signal confirmed ZIKV-induced reduction of fluorescence intensity (FI) as well as nucleolar territory. (e) Nucleolar number was unaffected by ZIKV. (f) Nucleolar stress in ZIKV-infected hNPCs as revealed by reduced nucleolar FI of NPM1 signal at dpi 1 (representative images are shown in Supplementary Fig. S2 ). (g) At dpi 4, there was also a significant reduction in NPM1-defined nucleolar territory of PRV-infected cells. (h) Nucleolar number was unaffected. At dpi 4, cells that survived MR766 infection showed similar anti-nucleolar effects as those infected with PRV (Fig. S3 ). Immunofluorescence staining for an additional marker of the nucleolar GC, PES1 confirmed negative effects of ZIKV infection on hNPC nucleoli ( Supplementary Fig. S4 ). Data represent two independent experiments including 2 sister cultures/experiment in Co-staining for Flavi-E, NPM1, and, DNA revealed that in many Flavi-E-positive-but non-apoptotic neurons both nucleolar NPM1 intensity and size of the nucleolus appeared to be reduced (Fig. 2c ). Infection with MR766-but not PRV increased fraction of a-nucleolar cells with most NPM1 translocated to the nucleoplasm and lack of a clearly identifiable nucleolus (Fig. 2c,d) . Moreover, in MR766-or PRV-infected neurons that maintained nucleolar enrichment of NPM1, the nucleus-normalized fluorescence intensity of nucleolar NPM1 as (e,f) Quantification of nucleolar NPM1 signal confirmed ZIKV-induced reduction of its intensity as well as territory. (g) Nucleolar number was reduced by MR766 but not PRV. In (b,d) data represent 6 sister cultures from three independent experiments; in (e-g), images of at least 60 randomly selected individual cells with no signs of apoptotic chromatin condensation were analyzed for each condition; such images were collected from two independent experiments; error bars are SEM. Data were analyzed by u-test (b,d) or one-way ANOVA and Tukey's post-hoc tests (e-g), NS, p > 0.05; *p < 0.05; ***p < 0.001. well as NPM1-defined nucleolar territory was reduced (Fig. 2e,f) . However, MR766 effects on these parameters were greater than those of PRV (MR766 vs. PRV, p < 0.001 or p < 0.01 for nucleolar NPM1 intensity or territory, respectively, Tukey's post-hoc test; Fig. 2e ,f). Also, number of nucleoli per nucleus declined in response to MR766 but not PRV (Fig. 2g) .

Discussion

However, nucleolar disruption-associated RS may not be a general feature of a host cell response to viral infection. First, some viruses activate rather than inhibit rRNA synthesis as demonstrated in human hepatoma cells that were infected with the flavivirus hepatitis C virus (HCV) 49 . Second, analysis of nucleolar proteome in cell lines that were infected with such RNA viruses as the coronavirus infectious bronchitis virus or influenza virus A revealed surprisingly small changes in nucleolar proteins unlike massive shifts of nucleolar proteomic landscape in response to the nucleolodisruptive transcriptional inhibitor Actinomycin D 50, 51 . Nevertheless, focusing attention on nucleolar morphology as a correlate of RS may underestimate the occurrence of virally-induced RS. For instance, inhibition of post-transcriptional processing of rRNA due to depletion of certain ribosomal proteins or RBFs as well as increase of ribosomal biogenesis that is associated with oncogenic transformation may induce RS despite relatively normal appearance of the nucleolus 27, 28 . Similar pro-RS effects may be expected in cases of viral infections that upregulate Pol1 and/or perturb rRNA processing without blocking rRNA transcription 49, 52 .

Statistical analysis.

Data availability statement. The datasets generated during the current study are available from the corresponding author on reasonable request.

Nucleolar disruption in ZIKV-infected neural cells.

Rat NPCs (rNPCs) were infected with two different strains of ZIKV including the African strain MR766 and the Puerto Rico strain PRVABC59 (PRV). Neurosphere formation was disrupted by either strain of ZIKV as early as at 2 days post infection (dpi, Supplementary Fig. S1 ). It was accompanied by cell attachment to the plates and disappearance of neurospheres suggesting NPC differentiation ( Supplementary Fig. S1 ). These effects were associated with moderate infection of rNPCs as, at least, at 4 dpi, only 8.3-10% cells were positive for flaviviral E protein (Flavi-E) dependent on ZIKV strain ( Supplementary Fig. S1 ). In addition, increased apoptosis was observed in MR766infected rNPCs ( Supplementary Fig. S1 ). Hence, our findings are consistent with reports of disrupted neurogenesis in ZIKV-infected human NPC (hNPC) neurosphere cultures and suggest that the virus perturbs growth and/or maintenance of mammalian NPCs 9, 33 .
ZIKV-induced nucleolar abnormalities were also present in human induced pluripotent stem cell-derived NPCs (hNPCs). In this system, MR766 or PRVABC59 led to relatively rapid declines of cell viability with at least 75% reductions at 3-or 5 dpi, respectively ( Supplementary Fig. S2 ). Infection rates were also higher than in rat NPCs (maximum fractions of infected cells of 84% or 49% with the MR766 or PRV at 2-or 4 dpi, respectively, Supplementary Fig. S2 ). At 1 dpi, when Flavi-E signal was detected in 11% or 0.25% cells that were infected with MR766 or PRV, respectively, each ZIKV strain reduced nucleolar NPM1 signal intensity by nearly 30% (Supplementary Fig. S2 and Fig. 1f) . Importantly, at that time hNPC viability was unaffected by ZIKV ( Supplementary Fig. S2 ). Similar reduction of NPM1 signal intensity and additional nucleolar shrinkage were present in surviving cells that were infected with PRV for 4 days (Fig. 1f,g) . Nucleolar number was not significantly altered by ZIKV at any time (Fig. 1h ). In addition, fraction of cells with absence of nucleolar NPM1 did not change after ZIKV infection except a slight increase from 0 to 3.2% in MR766-infected hNPCs at 4 dpi. At that time, MR766-infecetd cells which contained NPM1-positive nucleoli displayed similar reductions of nucleolar NPM1 as at 1 dpi ( Supplementary Fig. S3 ). Finally, in ZIKV-infected hNPCs, anti-nucleolar effects were also evident when nucleolar GC was visualized by immunostaining with a human-specific antibody against the ribosomal biogenesis factor PES1 ( Supplementary Fig. S4 ). These findings suggest that in ZIKV-infected NPCs, structural integrity of the nucleolus is disturbed.

Discussion

Our studies have revealed (i) presence of RS including nucleolar disruption in brain cells that were infected with ZIKV, and, (ii) ability of the overexpressed capsid protein ZIKV-C to localize to the nucleoli, induce RS and activate RS-mediated neuronal apoptosis. Therefore, inhibition of ribosomal biogenesis is directly involved in neurocytopathic effects of ZIKV and may lead to neurodevelopmental defects that are observed after fetal infection with that pathogen. Table S2 ). At a low concentration of 33 nM, ActD abolished all nascent RNA signal in nucleoli validating its specificity. Data represent averages of at least 39 cells/condition from two-(b-e,j) or three independent experiments (f-h); NS, p > 0.05; *p < 0.05; ***p < 0.001 (one-way ANOVA and Tukey's post-hoc tests).
Importantly, presence of RS may be determined not only by the viral species but also by differential host cell sensitivity to anti-nucleolar effects of viruses. In support of that notion, we observed greater extent of ZIKV-induced nucleolar disruption in post-mitotic neurons than in proliferating NPCs (Table 1) . Moreover, overexpression of ZIKV-C was sufficient to induce RS in neurons but not hNPCs or SH-SY5Ys (Fig. 4 vs. Supplementary Fig. S9 ). One could speculate that such a differential sensitivity could be determined by unequal expression of RBFs that are targeted by viral proteins. As neuronal differentiation is associated with reduced Pol1 activity and downregulation of such targets for flaviviral C proteins as NPM1 or NCL 53 , post-mitotic neurons may be particularly sensitive to ZIKV-induced RS. If correct, such a scenario would predict that although ZIKV is most infectious to NPCs, at least its RS-related toxicity would be most pronounced in NPC progeny as it enters the neuronal differentiation fate. Such a differentiation stage-specific cytotoxicity could explain why the ZIKV-related microcephaly is often accompanied by symptoms of cranial collapse suggesting massive brain cell death that follows initially successful neurogenesis 54 . Future studies are needed to directly test whether neuronal differentiation-associated changes in RBF expression determine sensitivity to ZIKV-induced cell death and shape neuropathogenic cascades that underlie ZIKV-mediated damage of the developing brain.
Recent work has shown that liquid liquid phase transition (LLPS) plays a critical role in formation of the granular component (GC) of the nucleolus 36 . LLPS is mediated by multivalent interactions of NPM1 acidic amino acids with basic amino acids of other nucleolar proteins and NPM1's RNA binding sites with rRNA. Interestingly, that process can be disrupted by arginine-containing dipeptide repeats (DPRs) that arise due to ALS/FTLD-associated intronic expansion of the C9ORF72 gene 58 . DRPs which are believed to be key contributors to neuronal death in C9ORF72-related cases of neurodegeneration, impair dynamics of nucleolar components and compromise ribosomal biogenesis. Although DRPs are nucleolar; they also disrupt dynamics and function of other membraneless organelles such as RNA stress granules or nuclear pores. In both cases, negative impact on these structures is believed to be due to disruption of protein-protein interactions that involve basic amino acids 58 . Therefore, it is tempting to speculate that the highly basic ZIKV-C may (i) initiate similar molecular mechanism of RS as DRPs, and, (ii), like them, may have negative effects on other LLPS-dependent cellular structures including stress granules, Cajal's bodies and/or nuclear speckles.