All Arabidopsis thaliana (L.) Heynh. plants studied in this work were homozygous for the mutations indicated in each case. The Nottingham Arabidopsis Stock Center provided seeds of the Col-0 accession (N1092; referred to herein as WT), as well as of the alkbh9a (N693727), alkbh9b (N515591) and alkbh9c (N521775) mutants. For rosette morphology analyses and to obtain multiple mutant genetic combinations, plants were grown under sterile conditions on 150-mm diameter Petri dishes, containing 100ml half-strength Murashige and Skoog agar medium, 0.6% Gelrite Agar (Duchefa), and 1% sucrose, at 20°C±1°C, 60–70% relative humidity, and continuous illumination of ≈ 75μmol·m⁻²·s⁻¹, and transferred into pots 21days after stratification (das) as previously described (Ponce et al., 1998). Sixteen evenly spaced seeds were sown per plate. Crosses were performed as previously described (Quesada et al., 2000). Photographs of rosettes were taken on a Nikon SMZ1500 stereomicroscope equipped with a Nikon DXM1200F digital camera. Rosette leaf area was calculated using Phenovein (version r214) of the MeVisLab software (version 3.1.1; Bühler et al., 2015).

The presence and position of annotated T-DNA insertions were confirmed by PCR amplification using the LbB1.3 primer and gene-specific primers (Supplementary Table 1). For RT-PCR, total RNA was extracted from seedlings collected 15 das using TRIzol (Invitrogen, Thermo Fisher Scientific) and contaminating DNA was removed using a TURBO DNA-free Kit (Invitrogen). First-strand cDNA was synthesized using random hexamers and the Maxima Reverse Transcriptase system (Fermentas). The house-keeping ORNITHINE TRANSCARBAMYLASE (OTC) gene was used as an internal control for relative expression analysis, as previously described (Quesada et al., 1999). Three different biological replicates were used. Sequence data can be found at The Arabidopsis Information Resource¹ under the following accession numbers: ALKBH9A (At1g48980), ALKBH9B (At2g17970), ALKBH9C (At4g36090), and OTC (At1g75330).

Nicotiana benthamiana and Arabidopsis thaliana plants were grown in 6-cm diameter pots in a growth chamber with a photoperiod of 25°C-16-h light/20°C-8-h dark. Plants were mechanically inoculated with purified virions (1mg/ml) of AMV PV0196 isolate (Plant Virus Collection, DSMZ) in 15mm sodium phosphate buffer pH 7. Detection of vRNAs was carried out by northern or direct-dot blot analysis. Inoculated leaves, non-inoculated rosette leaves, upper systemic floral stems, and roots were grounded in liquid nitrogen with mortar and pestle, and total RNA was extracted from 0.1g leaf material using EXTRAzol reagent protocol (Blirt). 0,5μg of total RNA was denatured by formaldehyde treatment and analyzed by northern blot. Otherwise, for direct-dot blot analysis, 3μl of total RNA was directly blotted onto nylon membranes. vRNAs were visualized on blots using digoxigenin-labeled riboprobes to detect the four vRNAs of AMV. Synthesis of the digoxigenin-labeled riboprobes, hybridization, and digoxigenin detection procedures was carried out as previously described (Pallás et al., 1998).

Leaves of 3-week-old WT and alkbh9b plants were inoculated with AMV virion and directly printed onto a nylon membrane at 4 dpi by pressing the lower side as previously described (Más and Pallás, 1995). To analyze AMV systemic invasion in all alkbh9 mutant combinations, floral stems were cut at 12 dpi and directly pressed onto the nylon membranes. Hybridization procedures were carried out as previously described (Pallás et al., 1998).

RNA in situ hybridization using a digoxigenin-labeled probe was performed on 8μm longitudinal paraffin sections as described previously (Gómez-Mena and Roque, 2018). Plant material was fixed in a solution containing 4% paraformaldehyde and 2.5% glutaraldehyde in 0.1M phosphate buffer (pH 7.2) for 4h at 4°C and embedded in paraffin. Hybridization was performed at 55°C overnight. After hybridization and immunodetection, slides were examined and photographed with a Leica DM 5000 equipped with a digital camera. When required, several images were stitched together using the stitching function of Leica Application Suit v4.9.

The RNase sensitivity assay was performed according to Gao et al. (2021). For this, 0.1g of systemic tissue previously infected with AMV was ground in liquid nitrogen and was incubated in 100μl of PIPES (CAS:5625-37-6) buffer (50mm PIPES, pH 6.7, 0.1% Tween 20) at 37°C. Total RNA was extracted at different time points and was analyzed by northern blot.

Isolation and inoculation of WT and alkbh9b protoplasts were carried out as previously described (Yoo et al., 2007) with some modifications. Well-expanded leaves from 3-week-old plants were chosen, and their undersides were slightly rubbed with carborundum. The leaves were incubated on the enzyme solution [20mM MES (pH 5.7), 1.5% (wt/vol) cellulase R10, 0.4% (wt/vol) macerozyme R10, 0.4 M mannitol, 20mm KCl, 10mM CaCl₂, 1–5mM β-mercaptoethanol, and 0.1% BSA] for 3h at room temperature. The enzyme/protoplast solution was diluted with an equal volume of W5 solution [2mM MES (pH 5.7), 154mM NaCl, 125mM CaCl₂, and 5mM KCl] before filtration with miracloth. After that, a sucrose cushion was performed adding approximately 1ml of 20% sucrose to the bottom of the tube with a Pasteur pipette and centrifugated at 200g for 5min. Protoplasts from the interphase were recovered and diluted in 1 volume of W5 and centrifuged at 200g for 3min. After removing the supernatant, the protoplasts were resuspended in 1ml in W5 solution and kept on ice for 30min. The supernatant was removed, and the protoplasts were resuspended in MMG solution [4mM MES (pH 5.7), 0.4 M mannitol, and 15mM MgCl₂]. For the transfection process, 10μl of vRNA (10–20μg extracted from AMV virions), 200μl of protoplasts (5×10⁵ cells), and 210μl of PEG solution [40% (wt/vol) PEG4000, 0.2 M mannitol, and 100mM CaCl₂] were gently mixed in a 2-ml microfuge tube and incubated at room temperature for 7min. To stop the transfection reaction, the samples were diluted with 840μlW5 solution at room temperature and centrifuged at 100g for 2min. After removing the supernatant, protoplasts were resuspended with 1ml WI [4mM MES (pH 5.7), 0.5 M mannitol, and 20mM KCl] and incubated at room temperature for 18h. Total RNA extraction with EXTRAzol reagent protocol (Blirt) was carried out to analyze viral accumulation by northern blot analysis.

A region of 200 nucleotides from the AMV RNA1 and RNA2 was amplified with specific primers of which the antisense primer contained the T7 RNA polymerase promoter. Amplified cDNA products and a plasmid containing a fragment of the ORF AMV CP linearized with the appropriate restriction enzyme were used as templates to transcribe RNA digoxigenin-labeled probes using T7 RNA polymerase and following manufacturer’s recommendations (Takara Bio Inc. Shiga, Japan). A mixture of Dig RNA 1, Dig RNA 2, and Dig-CP (100ng/each) or only the Dig-CP (500ng) was used as a probe for northern blot and in situ hybridization assays, respectively.

In Arabidopsis, ALKBH9 members form a specific clade including ALKBH9A, ALKBH9B, and ALKBH9C genes (Mielecki et al., 2012; Martínez-Pérez et al., 2017). Using the online tool Clustal Omega, we obtained the alignment between the amino acid sequences of the AlkB domain (residues 222 to 412 in ALKBH9B) from these three proteins and human ALKBH5 (hsALKBH5, residues 85 to 284; Figure 1A). Remarkably, the AlkB domain and some amino acids described as key residues for the m⁶A demethylation activity of hsALKBH5 (highlighted in Figure 1A) are quite conserved in the three ALKBH9 Arabidopsis co-orthologues (Feng et al., 2014; Lu et al., 2014; Alvarado-Marchena et al., 2021). Thus, we observed that the known HxD/E·· H and RxxxxxR motifs are conserved between the three proteins (Figure 1A, highlighted in yellow). According to previous authors, while the HxD/E···H motif and the first arginine (R) of RxxxxxR are involved in the binding to iron [Fe (II)] and 2-oxoglutarate (2-OG), the second R (R411 in ALKBH9B) seems to be related to the substrate specificity of AlkB (Bratlie and Drabløs, 2005; Lu et al., 2014). Moreover, the N and Y residues involved in 2-OG stabilization are conserved in all ALKBH9 paralogues (Figure 1A, highlighted in red, N324 and Y326 in ALKBH9B), whereas only one of the Y residues directly binding m⁶A is conserved in ALKBH9A (Figure 1A, Y273 and Y275 in ALKBH9B, in fuchsia). Likewise, instead of the characteristic hsALKBH5 motif QKR (position 170 in ALKBH9B), which is probably involved in the binding to the substrate, ALKBH9A and ALKBH9C present QRR and QKK, respectively (Figure 1A, in blue). On the other hand, an arginine residue essential for hsALKBH5 activity (Feng et al., 2014) is also conserved between the three ALKBH9 proteins (Figure 1A, R265 in ALKBH9B, highlighted in green). Finally, a histidine residue specifically found in hsALKBH5 but not in the rest of AlkB is also present in all ALKBH9 proteins (Figure 1A, H339 in ALKBH9B, highlighted in gray). Overall, this hints at m⁶A demethylation activity of ALKBH9A and ALKBH9C that also could affect AMV propagation. Therefore, we carried out a study to check whether ALKBH9A and/or ALKBH9C might be also playing a role in AMV systemic infection.

Thus, we intercrossed the alkbh9a (SALK_204823C), alkbh9b (SALK_015591), and alkbh9c (SALK_021775) insertional single mutants (Figure 1B), obtaining all possible multiple genetic combinations: the alkbh9a alkbh9b, alkbh9a alkbh9c and alkbh9b alkbh9c double mutants, and the alkbh9a alkbh9b alkbh9c triple mutant, to which we will refer here, for simplicity, as alkbh9ab alkbh9ac, alkbh9bc, and alkbh9abc, respectively. We confirmed by PCR the presence of the corresponding T-DNA insertions in homozygosis in all single and multiple mutants (Figure 1C). We also confirmed by semiquantitative RT-PCR the absence of full-length transcripts of the disrupted genes (Figure 1D, Supplementary Figure Legends). As expected, we found that ALKBH9A expression was not detected in whole seedlings of the Col-0 wild type (referred to herein as WT); unexpectedly, however, it was very low, but detectable in the alkbh9b and alkbh9bc mutants and to a lesser extent in alkbh9c (Figure 1D; Duan et al., 2017). A compensation mechanism like this has already been observed among other paralog groups in Arabidopsis; examples are the paralog gene pairs encoding the pathogen responsive VESICLE-ASSOCIATED MEMBRANE PROTEINS721 (Kwon et al., 2008; El-Brolosy and Stainier, 2017) and VAMP722 or the nucleolar proteins NUCLEOLIN1 (a ribosome biogenesis factor) and NUC2 (Durut et al., 2014). We studied the morphological phenotype at the level of whole rosette growth progression in alkbh9 single and multiple mutant seedlings. Overall, rosettes of the mutant plants studied were barely distinguishable from WT (Figures 1E–M).

The effect on viral systemic movement of the alkbh9a, alkbh9b, and alkbh9c single mutants, the alkbh9ab, alkbh9ac, and alkbh9bc double mutants, and the alkbh9abc triple mutant was analyzed by detecting AMV accumulation in floral stems by tissue printing at 12days post-inoculation (dpi). As shown in Figure 1N, the viral systemic movement was impaired only in plants lacking ALKBH9B (highlighted in red), indicating that neither ALKBH9A nor ALKBH9C is required for the systemic invasion of the plant. This result highlights the specificity of the interaction between AMV infection and ALKBH9B.

The results shown in Figure 1N are in line with our previous work reporting that AMV succeeded in infecting only 9% of the floral stems of mutant plants at 14 dpi, whereas that percentage increased until 90–100% in WT plants (Martínez-Pérez et al., 2017). However, the evaluation of AMV infection in total non-inoculated aerial tissue by dot blot revealed that 40% of alkbh9b plants presented a systemic infection compared to 93% of WT plants at 20 dpi (Supplementary Figure 1). This inconsistency encouraged us to analyze more precisely the evolution of the systemic infection in plants. Thus, we compared the viral titers in different non-inoculated tissues (roots, rosette leaves, and floral stems). After total RNA extraction, vRNAs load in these tissues was evaluated by northern blot at 11 and 15 dpi (Figures 2A,B, respectively). Although the lower susceptibility of alkbh9b plants to AMV was evident in all cases, it was especially notable in roots and floral stems, in which we practically observed a block of the viral infection. Considering that to invade upper sink leaves and roots, viruses enter the phloem of the inoculated leaves (Hipper et al., 2013; Navarro et al., 2019), our results suggest that AMV phloem loading in alkbh9b plants is restricted.

Next, we evaluated in more detail the viral progression in the shoot apical meristems (SAM) and floral stems at 8 and 15 dpi, respectively, detecting AMV RNA4 by in situ hybridization (Figure 3, Supplementary Figure 3). At 8 dpi, no hybridization signal was observed in the SAM of WT plants, as expected accordingly to the general knowledge that viruses cannot invade the SAM (Gosalvez-Bernal et al., 2006; Wu et al., 2020). However, we could observe the presence of vRNAs patches in both developing flowers and floral stems. No signal was observed in alkbh9b plants in these same tissues (Figure 3A, Supplementary Figure 2). Furthermore, at 15 dpi, AMV RNA was still not detected in transverse sections of the inflorescence stems from alkbh9b plants, whereas the hybridization signal covered almost completely those from WT plants (Figure 3B). Overall, these observations show that AMV systemic infection is generally impaired by the absence of ALKBH9B and viral RNA is unable to reach floral stems and roots. More remarkably, these new findings suggest that ALKBH9B is involved, somehow, in the viral upload to the phloem in Arabidopsis, and thus, when this protein is not present, AMV very rarely invades the vascular tissues.

In contrast to cell-to-cell movement, as explained above, AMV needs to form viral particles to move systemically (Sánchez-Navarro et al., 2006). The diminished systemic invasion observed in alkbh9b plants could be caused by compromised stability of the AMV particles. To check this hypothesis, we evaluated the stability of viral particles obtained from WT and alkbh9b plants. It would be expected that encapsulated vRNAs were hardly processed by RNases. Hence, WT- and alkbh9b-infected leaves were homogenized with PIPES buffer and, after different times of incubation at 37°C, vRNAs levels were evaluated by northern blot assays (Supplementary Figure 3). While ribosomal RNAs were completely degraded after 30min of incubation, vRNAs from both WT and mutant plants were intact to the same extent. Therefore, the absence of ALKBH9B activity does not affect the stability of AMV viral particles.

We further evaluate both AMV cell-to-cell movement among mesophyll cells by analyzing the viral distribution in the inoculated leaves. Inoculated leaves from WT and alkbh9b plants were directly printed onto nylon membranes, and AMV RNAs were detected by hybridization with dig-AMV probe (Figure 4). While most leaf surfaces were already infected in WT plants at 4 dpi, the infection in alkbh9b leaves did not reach the petioles yet. Since this may indicate a delay in the viral infection progress, we addressed in situ hybridization experiments of sections of leaf petioles and leaf blades of inoculated leaves using a dig-RNA4 probe to detect AMV RNA 3 and sgRNA 4 (Figure 5A). In WT petioles, the hybridization signal covered most of the tissue already at 4 dpi, whereas alkbh9b mutant petioles showed small patches even at 7 dpi (Figure 5B). Moreover, when the analysis was extended to leaf blades, we found that, while AMV signal was observed all over the section in WT leaf blades, only individual patches of hybridization were seen in alkbh9b, pointing to a less efficient invasion of mesophyll cells in the inoculated leaves when this protein is not present (Figure 5C, Supplementary Figure 4). Remarkably, the lack of infection in vascular tissues from alkbh9b even when the infection has reached the directly surrounding cells of this area would explain the failure of the virus to invade the floral stems (see Figure 3, Supplementary Figure 2).

The phenotypic defect in alkbh9b plants consisting of a slower cell-to-cell movement may occur, at least partially, as a consequence of any defects in earlier stages of the viral cycle (replication, translation or RNA stability). In fact, in situ hybridization assays did not only display a less widespread infection, but also a weaker viral signal in alkbh9b plants (Figure 5, Supplementary Figure 4). To properly check this hypothesis, we inoculated isolated protoplasts from WT and alkbh9b plants with purified AMV vRNAs and compared the infection levels by northern blot analysis. The result showed that viral load was lower in protoplasts coming from mutant plants (Figure 6), suggesting that ALKBH9B is involved at least in one of these stages: viral replication, translation, or vRNAs stability.