Entire Expressed Peripheral Blood Transcriptome in Pediatric Severe Malarial Anemia

This study on severe malarial anemia (SMA: Hb < 6.0 g/dL), a leading global cause of childhood morbidity and mortality, analyzed the entire expressed transcriptome in whole blood from children with non-SMA (Hb ≥ 6.0 g/dL, n = 41) and SMA (n = 25). Analyses revealed 3,420 up-regulated and 3,442 down-regulated transcripts, signifying impairments in host inflammasome activation, cell death, innate immune responses, and cellular stress responses in SMA. Immune cell profiling showed a decreased antigenic and immune priming response in children with SMA, favoring polarization toward cellular proliferation and repair. Enrichment analysis further identified altered neutrophil and autophagy-related processes, consistent with neutrophil degranulation and altered ubiquitination and proteasome degradation. Pathway analyses highlighted SMA-related alterations in cellular homeostasis, signaling, response to environmental cues, and cellular and immune stress responses. Validation with a qRT-PCR array showed strong concordance with the sequencing data. These findings identify key molecular themes in SMA pathogenesis, providing potential targets for new malaria therapies.


INTRODUCTION
Malaria remains a signi cant global public health challenge with 247 million annual cases and 619,000 deaths 1 . The majority of the cases (234 million) and mortality (593,000) occurred in the WHO African region and are due to infections with Plasmodium falciparum 1 , mainly in children under ve of age. Kenya faces a substantial challenge with P. falciparum malaria, reporting ∼3.42 (2.48-4.64) million annual cases and ∼12,011 (10,800 − 14,000) deaths, primarily in the under-ve population 1,2 . Since the disease burden increases with transmission intensity, severe malaria remains among the leading causes of morbidity and mortality in children residing in holoendemic P. falciparum regions of Kenya, and other such regions of sub-Saharan Africa [3][4][5] . In high transmission regions, the primary manifestation of severe malaria is severe malarial anemia [SMA, hemoglobin (Hb) < 6.0 g/dL] in both the presence and absence of respiratory distress with cerebral malaria occurring only in rare (atypical) cases [5][6][7] .
The etiology of SMA is multifaceted and includes overlapping characteristics such as the destruction of infected and uninfected erythrocytes, sequestration of erythrocytes in the spleen, and suppression of bone marrow functions (reviews, see Perkins 8,9 ). Although natural immunity is acquired following repeated infections with P. falciparum [10][11][12] , innate immunity is particularly important for determining disease severity in young, malaria-naïve children. Our previous longitudinal studies in Kenyan children using a combination of candidate-gene approaches, genome-wide association studies, high-throughput genotyping, and array-based whole transcriptional pro ling revealed that the development of SMA is mediated, partially by innate immune response genes 8,13−16 . Our targeted transcriptome analyses also revealed that differentially expressed genes (DEGs) in host ubiquitination processes are a central feature of SMA pathogenesis 17 . Microarray analysis of candidate genes in whole blood has also identi ed DEGs that encode amino acid transport, phospholipid metabolic processes, and positive regulation of nitrogen compound metabolic processes in Gabonese children with SMA (< 6 years old) 18 .
Advances in high-throughput sequencing technologies and bioinformatics analyses have provided important insight into the human immune response to P. falciparum and identi ed potential vaccine candidates [18][19][20][21] . For example, studies investigating gene expression in adults during controlled human malaria infection experiments identi ed > 2,700 DEGs in the whole blood transcriptome, for which a subset of 265 genes was associated with transcription and cell-cycle regulation, phosphatidylinositol signaling, and erythrocytic development 22 . In addition, dual next-generation RNA sequencing in whole blood, which concomitantly captures host and parasite gene expression, showed that severe malaria (i.e., cerebral malaria, hyperlactatemia, or their combination) in Gambian children (< 16 years old) was associated with increased expression of granulopoiesis and interferon-γ-related genes, and suppression of type 1 interferon signaling 20 . Despite progress in de ning the human immune response to P. falciparum, a comprehensive investigation of the entire expressed transcriptome has not been reported in children who develop SMA, the group who suffers the highest global morbidity and mortality 5,6,9,23−26 .
Here, we present the top emergent biological processes, networks, and pathways for the rst entire expressed whole blood transcriptome in Kenya children (< 5 years old) from a holoendemic region of western Kenya who develop SMA as the exclusive phenotype of severe malaria.

Demographic and Clinical Characteristics of the Study Participants
Admission demographic and clinical characteristics of the children selected for RNA-Seq in whole blood are shown in Table 1. Based on the selection criteria, sex (p = 0.800), overall age (p = 0.797), and distribution within age categories (p = 0.461) were comparable. Glucose levels (p = 0.967) and auxiliary temperature (p = 0.051) were comparable between the groups. Consistent with more profound anemia in children with SMA, hematocrit (p = 1.242E-11) and red blood cells (p = 1.790E-11) were lower, while red cell distribution width (p = 4.050E-4) and mean corpuscular volume (p = 0.002) were elevated. White blood cells (p = 1.390E-4) and lymphocyte (p = 1.000E-6) counts were also elevated in children with SMA. Other hematological measures were comparable between the groups, as were parasitological indices. There was a lower distribution of HbAS and higher proportion of HbSS in the SMA group (p = 0.029), yet not signi cant after multiple test correction. Data are the median (interquartile range; IQR) unless otherwise noted. Children (n = 66) presenting with malaria at SCRH were recruited. Based on hemoglobin (Hb) levels, children were categorized into either non-severe malaria anemia (non-SMA; Hb ≥ 6.0 g/dL, n = 41) or severe malarial anemia (SMA; Hb < 6.0 g/dL, n = 25). a Fisher's exact test with exact p-values for homogeneity was performed. b Twosided Mann-Whitney-U tests were used to compare the non-SMA and SMA groups, c Group means were compared by two-sided, two-sample t-test, with equal variance. All p-values shown in bold remained below the signi cance level after multiple test correction using the Bonferroni-Holm method (familywise error rate, signi cance level 0.050). Abbreviations: MPS -malaria parasites presented as mean (standard deviation).

Differential Gene Expression and Central Regulatory Features in SMA
Differential expression analysis identi ed 3,420 up-and 3,442 down-regulated genes (padj < 0.050) in children with SMA ( Fig. 1A). Comparison of the DEGs revealed 992 genes that were uniquely expressed in non-SMA, 328 in SMA, and 15,592 co-expressed genes (Fig. 1B). A non-supervised hierarchical cluster analysis of the top 1000 DEGs identi ed unique co-regulated gene clusters in children with SMA, and computed the distribution of clinical variables (i.e., parasitemia, sickle cell status, and age) in each of the groups (Fig. 1C).
Canonical pathway maps for direct functional interactions were generated to gain insight into the networks for two of the major clusters, one down-regulated (n = 156, cluster 1) and one up-regulated (n = 114, cluster 2). The network for the down-regulated genes (cluster 1) was IRF1↔IL-1β↔Caspase-1↔Caspase-4↔FasR with the transcription factor, IRF1, as the central divergence hub (36 direct interactions) and IL-1β as the central convergence hub (27 direct interactions, Fig. 1D). The top process associated with the functional interactions for cluster 1 was Response to Stress (p = 2.404E-08). Cluster 2 (up-regulated) generated a TCF7L1↔E2A↔RING2 network with two transcription factors, TCF7L1, as the central divergence hub (26 direct interactions) and E2A, as a secondary divergence hub (10 functional interactions). The central convergence hub in cluster 2 was RING2 with 5 functional interactions (Fig. 1E). Regulation of Metabolic Process was the top represented process for cluster 2 (p = 2.671E-09).
Collectively, these results suggest that SMA is characterized by enhanced stress responses and perturbations in metabolic processes.

Altered Leukocytic Immune Cell Pro les in SMA
To determine if leukocytic immune pro les differed in children who developed SMA, a bioinformatic approach was implemented using CIBERSORTx. Although there was interindividual variability, 10 immune cell types were differentially expressed at p < 0.050 ( Fig Results from the GO and Reactome enrichment analyses converge on common biological processes and suggest that SMA is characterized by altered neutrophil responses, perturbations in autophagy and endosomal pathways, and activation of ubiquitin-related processes. KEGG Canonical Pathways Implicated in SMA: Functional classi cation of the DEGs between SMA and non-SMA for the KEGG pathways is shown in Fig. 4A. The top-ranked pathways were: (i) cellular processes -endocytosis (padj = 3.380E-05); (ii) environmental information processing -TNF signaling (padj = 3.829E-04); (iii) genetic information processing -protein processing in ER (padj = 7.880E-09), which was also the most signi cant of all pathways; (iv) metabolism -inositol phosphate metabolism (padj = 4.369E-02); and (v) organismal systems -FcγR-mediated phagocytosis (padj = 3.829E-04).
Collectively, the KEGG results suggest that children with SMA exhibit signi cant alterations in immune response triggering, cellular recycling processes, and protein regulation.
Endoplasmic Reticulum Quality Control Dysfunction in SMA: To further characterize the pathogenesis of SMA, we focused on the top emergent pathway generated from the KEGG database: Protein Processing in Endoplasmic Reticulum (ER; Fig. 4B). Children with SMA had elevated expression of ERManI and PERK and decreased expression of HSP40, EDEM, PDIs, and TRAP, indicating that the ER quality control system is impaired, leading to potential protein misfolding and cellular dysfunction. This is consistent with upregulation of several transcripts in the ubiquitin ligase complex in the ER membrane (i.e., Ubx, gp78, and Derlin1) and dysregulation in the ER cytoplasm (i.e., elevated HSP40, RBX1, and Skp1, and decreased UbcH5 and FBP). Because of misfolded or damaged proteins, the ER-associated degradation pathway was activated, as illustrated by increased Otu1 and RAD23 expression. Further indications of ERassociated cellular stress and attempts to maintain cellular homeostasis in children with SMA are demonstrated by elevated Sec62/Sec63 complex subunits and MKK7.

Deciphering Cellular Stress Responses in SMA
Of the top canonical pathways that emerged from MetaCore™, we focused on positive regulation of WNT/ β-catenin signaling to further de ne cellular and immune stress responses in SMA (Fig. 4D). Children with SMA had reduced expression of WNT, but increased expression of the WNT receptor, LRP5/LRP6, and its target, Axin, which was also signaled by increased expression of ZBED3, SIAH2, PP2C, PP1-cat, and tankyrases. This transcriptional pattern indicates enhanced proteasomal degradation in the context of reduced protection from degradation (i.e., down-regulation of NKD1 and NKD2), despite increased expression of Dsh signaled by up-regulation of the WNT receptor (i.e., Frizzled), transmembrane receptor (i.e., ITGB1), IRS-2, and USP9X, accentuated by decreased phosphorylation from RIPK4 and reduced binding by GRB2 and β-aresstin2. Children with SMA also had increased expression for SMAD3, SMAD4, TBLR1, Makorin-1, Trabid, UBE2B, RNF220, USP7, PKB, and USB47 in the context of decreased expression in β-catenin, indicating activation of the non-canonical WNT signaling pathway. This pattern of expression also suggests cellular stress responses and attempts to regulate protein degradation and stabilize cell survival through alterations in ubiquitination and de-ubiquitination processes.
Validation of Whole Blood Transcriptome Data using a Targeted Gene Array: To validate the RNA-Seq results, we utilized a targeted gene approach with a qRT-PCR array that captured 84 genes involved in ubiquitination. A heatmap cluster analysis showed strong concordance in the fold-change and directionality in the two platforms (Fig. 5A). Cluster analysis of the signi cant DEGs in the qRT-PCR array showed a signi cant correlation with the same genes captured from RNA-Seq (r = 0.834, p < 0.001; Fig. 5B). To further compare ndings from the two datasets, a work ow was generated in MetaCore™ to identify common (shared) genes that mapped to pathways, GO processes, and process networks. The combined data objects (gene set) from the two platforms revealed the following top-ranked features: (i) pathway map -Proteolysis Ubiquitination (padj = 2.193E-11); (ii) GO process -Protein ubiquitination (padj = 1.083E-12); and (iii) process network -Proteolysis Ubiquitin -proteasomal proteolysis padj = 2.250E-05) (Supplemental Fig. 1A-C).

DISCUSSION
Severe life-threatening malaria is represented by distinct and overlapping disease features (one or more) of the following: impaired consciousness, prostration, multiple convulsions, acidosis, hypoglycemia, SMA, renal impairment, jaundice, pulmonary edema, signi cant bleeding, shock, and hyperparasitemia [27][28][29][30][31][32][33] . Identifying common gene pathways/networks that encompass the diverse pathophysiological landscape of severe malaria (i.e., mixed phenotype) has presented signi cant challenges, likely because distinct biological processes may not share common networks. The clinical manifestations of severe malaria and the age at which they present are largely driven by P. falciparum endemicity 9,34 . The overwhelming majority of life-threatening severe malaria occurs in holoendemic P. falciparum transmission areas of sub-Saharan Africa in children under ve years who develop SMA, making this severe manifestation a leading cause of childhood deaths in such regions 8,9 .
A major advantage of studies in holoendemic malaria regions, such as western Kenya, is that children have a distinct pathophysiological presentation of SMA, making the discovery of gene-disease relationships more feasible. Our previous studies have identi ed innate immune response genes that in uence the pathogenesis of SMA, largely through imparting changes in soluble mediators of in ammation 8, 9,15,35,36 . However, this is the rst investigation to examine the entire expressed peripheral blood transcriptome in children whose primary phenotype of severe disease is SMA. Findings from this study identify novel biological pathways and process networks that converge on perturbations in cellular and immune stress responses, illustrating that the pathogenesis of SMA is complex and multi-faceted.
Molecular patterns in children with SMA were initially identi ed by hierarchical clustering of the top 1000 DEGs. Relationships that emerged from the analysis were then further deciphered by creating canonical process networks for several gene clusters. Cluster 1 (IRF1↔IL-1β↔Caspase-1↔Caspase-4↔FasR) contained a set of down-regulated genes that are central to host in ammasome activation, cell death, and innate immune responses 37 . Children with SMA had a decrease in the transcription factor, IRF1, and downstream targets (i.e., IL-1β, capsases 1 and 4, and FasR) in the absence of a Nod-like receptor family (i.e., NLRP1, NLRP3, and NLRC4) response [37][38][39] . This suggests an inability to initiate pyroptosis and dysregulated in ammasome. Cluster 2 (TCF7L1↔E2A↔RING2) was up-regulated in SMA and contained a family of genes involved in cellular stress responses 40,41 . This set of genes also emerged in the topranked pathway (i.e., positive regulation of WNT/β-catenin signaling) suggesting that SMA is characterized by increased cellular stress responses via enhanced proteasomal degradation in the context of reduced protein degradation through the non-canonical WNT signaling pathway.
Inheritance of sickle cell trait (HbAS) has protective effects against the development of severe malaria [42][43][44][45][46] . As expected in a holoendemic region, there was a higher proportion of HbSS carriage (sickle cell disease) in the SMA group. Since hierarchical clustering analysis indicated more pronounced gene dysregulation in children with HbSS, the analyses were repeated without these children. This resulted in an identical network of down-regulated genes (cluster 1), but a different set of up-regulated genes (TAL1↔LYL1↔EKLF1↔HMBS↔RHD) that are involved in the regulation and maturation of erythrocytes and Hb production (review, see Love 47 ). Thus, there are both overlapping and distinct molecular pro les in children who develop SMA with and without inheritance of HbSS (Supplemental Fig. 2A-C).
Results from the immune pro ling with CIBERSORTx indicate that children with SMA have a decreased antigenic response, reduced immune priming, and an enhanced polarization towards cellular proliferation and repair. The hematological patterns captured by the CBC, although not as speci c, parallel results obtained from the immune cell pro ling. Consistency between the two independent methods supports the reliability of the observed immune alterations in SMA.
To gain further insight into SMA pathogenesis, we used a combination of functional enrichment analysis platforms to identify convergent patterns amongst central themes. One distinct feature of SMA was neutrophil activation and degranulation. Although not speci cally in children with SMA, previous studies suggest that neutrophil activation is more pronounced in severe malaria 20,48−53 . An additional molecular convergence among children with SMA was perturbations in autophagy. Altered autophagy, to our knowledge, has not been described in human malaria pathogenesis, but autophagy defects appear common in various other infectious diseases (review, see Deretic 54 ).
The transcriptome in children with SMA also revealed dysregulation in proteasome-mediated activity. This nding parallels our earlier studies that were the rst to report DEG in ubiquitin-related processes as a feature of SMA 17,55 . Canonical pathway analyses showed additional molecular features of SMA including changes in cellular homeostasis, signaling, environmental response, and various molecular mechanisms that regulate cellular and immune stress responses. A key emergent pathway in SMA was protein processing in the ER with DEGs involved in protein folding, export, targeting, ERAD, and the ubiquitin ligase complex. While these pathways have not been described for the human response to malaria, similar pathways are operational in human malaria parasites and are being explored as novel therapeutic targets 56-58 .
Changes in endocytosis-mediated pathways were also witnessed in SMA, indicating signi cant changes to cellular structure as a molecular theme, as has been described in experimental murine cerebral malaria models 59 . The central role of cellular stress response disruptions in SMA is further supported by dysregulation in the WNT/β-catenin signaling pathway. While changes in this pathway have been implicated in various diseases and are known to be affected by antimalarial drugs 60-64 , its role in immune response to malaria is a novel nding. Validation of the transcriptome data was performed using a targeted human ubiquitination qRT-PCR array technology containing 84 genes. We compared 15 signi cant DEGs previously identi ed in children with SMA 17 , and observed a high level of concordance and identical directionality from the two platforms. Moreover, enrichment analyses using the full datasets for both platforms revealed highly consistent and shared process networks and pathways', providing further validation and con rmation that disruptions in the ubiquitin proteasome system is a fundamental feature of SMA pathogenesis.
In conclusion, an unbiased RNA-Seq analysis capturing the entire expressed blood transcriptome identi ed key molecular aspects of SMA pathogenesis, such as changes in neutrophil responses, autophagy, endosomal pathways, and activation of ubiquitin-related processes and cellular stress responses. Strengths of this study are the extensive clinical characterization of the cohort which allowed for the exclusion of co-infections known to in uence the immune response 65-67 , and a robust sample size. Limitations of the study include potential generalizability to other forms of severe malaria, such as cerebral malaria, which is likely a distinct pathogenesis. To determine the impact of the observed transcriptional changes on protein expression, we are currently performing proteomics in the cohort presented here. Collectively, results presented here highlight the complexity and multi-dimensional nature of SMA. An improved understanding of this complexity can guide the development of novel targeted therapies for improved clinical outcomes.

Study region and participants
This study was conducted at Siaya County Referral Hospital (SCRH), located in a holoendemic P. falciparum transmission region in western Kenya where SMA is among the main causes of childhood morbidity and mortality in the community 5,6,25,26,68−70 . Individuals inhabiting the study area are predominantly from the Luo group (> 96%), an ethnically homogeneous population 71 . Children presenting at SCRH with symptoms of infectious diseases, and who met the following inclusion criteria; auxiliary temperature ≥ 37.5°C (axillary), age 0-59 mos., distance to hospital ≤ 25 km, and parent/guardian willing and able to provide signed informed consent and attend follow-up visits (14 days later), were approached for enrollment into the study. Children presenting with suspected non-infectious diseases were excluded. Based on inclusion/exclusion criteria, 565 children were enrolled into the acute febrile cohort (3/2017 to 9/2020). For children requiring hospitalization and those released as outpatients, venipuncture blood samples (3-4mL) were collected on admission into the study (day 0), prior to treatment with antimalarials or other medication. The study was approved by the Kenya Medical Research Institute Scienti c and Ethics Review Unit, the University of New Mexico Institutional Review Board, and the Maseno University Scienti c and Ethics Review Committee. Written informed consent was provided by the parents/legal guardian of the study participants.

Laboratory procedures
At enrollment, demographic and clinical data were collected, and a physical examination performed. Giemsa-stained thin blood smears were prepared, examined, and asexual malaria parasite densities determined 25 . Complete blood counts (CBCs) were determined using a DxH 500 hematology analyzer (Beckman-Coulter). Since we have shown that co-infections in uence malarial anemia severity in the Siaya community, all children were tested for HIV-1 and blood-borne bacterial infections according to our previously described methods 65-67 . Parents/legal guardians of participating children received pre-and post-test HIV&AIDS counseling. To further characterize potential causes of anemia, sickle-cell trait status was determined by alkaline cellulose acetate electrophoresis (Helena BioSciences).

Study participant selection for RNA-Seq
For selection of samples for the RNA-Seq from the 565 enrolled study participants, children with malaria were strati ed into two groups based on Hb levels (i.e., Hb ≥ 6.0 g/dL and Hb < 6.0 g/dL), and then matched according to age and sex. Further selection criteria for the RNA-Seq included omitting children with any detected co-infections (e.g., HIV, tuberculosis, bacteremia, etc) 65-67 . This selection strategy yielded 41 children with non-SMA (Hb ≥ 6.0 g/dL) and 25 children with SMA (Hb < 6.0 g/dL).

RNA isolation, library construction, and sequencing
Approximately 500µL of whole blood collected from venipuncture prior to treatment was stabilized with Trizol® (Thermo Fisher Scienti c Inc.), immediately frozen in liquid nitrogen, and then subsequently

Bioinformatics analysis
Raw data were quality-controlled and ltered using fastp 72  Differential expression analysis: Differential expression analysis of the two clinical conditions was performed using edgeR (3.16.5). For this study, adjusted p-values (padj) of < 0.050 plus absolute log 2 -(fold-change) of > 1 were used as the threshold for DEGs. To identify the correlation between different genes, samples were clustered using expression level FPKM utilizing the hierarchical clustering distance method.
Leukocytic immune cell pro ling: The relative percentage of different immune cell types/subtypes in whole blood was imputed using CIBERSORTx 77,78 . This analytical tool processes gene expression data from a bulk admixture of different cell types to estimate the abundance of member cell types in a mixed cell population 79  Encyclopedia of Genes and Genomes (KEGG) using the R package 'ClusterPro le' 81 . For all analysis, padj < 0.050 were considered signi cant enrichment. Con rmation and further discovery of the ndings were implemented by using MetaCore™ (https://clarivate.com/products/metacore/) to identify DEGs that mapped to GO processes, process networks, and pathways.

Validation of transcriptome pro les
To validate the transcriptome data, we used data generated from a targeted gene approach that measured transcript expression levels of 84 key genes involved in the ubiquitination process (Human Ubiquitination Pathway RT² Pro ler PCR Array kit, Qiagen). Total RNA was isolated from whole blood samples from children with non-SMA (Hb ≥ 6.0 g/dL, n = 23) and SMA (Hb < 6.0 g/dL, n = 21) who were not included in the RNA-Seq. Cluster analysis was used to compare the signi cantly expressed genes in the quantitative reverse transcription polymerase chain reaction (qRT-PCR) array with identical genes generated from the RNA-Seq. Convergence between the two datasets was determined by the 'compare experiments work ow' in MetaCore™ for GO processes, process networks, and pathways.
Declarations Figure 1 RNA-seq data for Kenyan children presenting with non-SMA (Hb≥6.0g/dL, n=41) and SMA (Hb<6.0g/dL, n=25  (Fig. 1D), while the TCF7L1 (TCF3), E2A, RING2 network shows up-regulated (n=114) genes in cluster 2 (Fig. 1E). The blue-shaded circles show downregulated genes and the red-shaded circles are up-regulated genes. The details of symbols used in these gures are available at: https://portal.genego.com/legends/MetaCoreQuickReferenceGuide.pdf. and SMA (Hb<6.0 g/dL, n=25) groups. *Indicates signi cant differences (p<0.050) in immune cell proportions between the two groups determined using two-sided, two-sample t-tests with Welch correction. (B) Relative proportion (%) of expression for the immune cell types that differed signi cantly between children with non-SMA and SMA. Bivariate analysis was performed using two-sided, two-sample t-tests with Welch correction and presented as mean (SEM) for the non-SMA and SMA groups.   Validation of the transcriptome data. A Human Ubiquitination Pathway RT 2 Pro ler PCR Array kit (Qiagen, LLC-USA, Germantown, MD, United States) was used to measure the expression of genes involved in the ubiquitination process (A) Heat map shows a graphical and color-coded representation of fold regulation (Log 2 ) comparison of signi cant DEGs using the RT-qPCR array versus data generated from the whole blood transcriptome analysis. The Y-axis shows the gene names. The X-axis shows the assay type. The darkest purple represents the lowest fold change, and the brightest yellow represents the highest fold change in children with SMA relative to non-SMA. (B) Correlation scatter plot of the signi cantly expressed ubiquitination process genes in the RT-qPCR analysis (Y-axis) versus the transcriptome data (X-axis). There was a strong positive correlation of the DEGs in SMA cases (r=0.834, p<0.001).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryInformation.docx