Citrus are widely used for the consumption of fruit flesh and juices. Modern citrus varieties were generated over thousands of years by intraspecific and interspecific crosses of a handful of species combined with clonal propagation1,2,3. Acidity is a major trait determining the taste and use of citrus fruits and selection by breeders and producers has generated a broad palette of sour and “sweet” (i.e., non-sour) varieties of lemons, oranges, pummelos, and other citrus fruits4

For a sour taste, food/liquid should have (i) a high concentration of free H+ ions (low pH), which is sensed by acid-sensitive cells in taste buds, presumably via an H+-selective channel5 and (ii) a certain pH-buffering capacity to prevent the liquid from being neutralized by the saliva. The acidity (low pH) of Citrus fruits is determined by the pH of the vacuoles in juice vesicle cells, which can be as low as 2 in sour lemons and lime6,7,8. The steep proton gradient across the vacuolar membrane (tonoplast) drives massive transport of citrate into the vacuole via a mechanism that is only partially understood9,10. As citrate enters the vacuole in dissociated form (citrate3−), it increases its buffer capacity, which contributes to the sour taste11, but does not lower the pH. How juice vesicle cells can hyperacidify their vacuoles to such an extreme extent remained elusive in spite of extensive biochemical work8,9,12,13,14,15.

In most plant cells, the cytoplasm is about neutral and the vacuolar lumen mildly acidic. The (moderate) pH gradient across the tonoplast is generated by vacuolar-ATPases (V-ATPases)16,17, which are complex multi-subunit proton pumps found in both animals and plants18,19. V-ATPases translocate 2–4 protons per hydrolyzed ATP (H+/ATP = 2–4) depending on the pH on both sides of the membrane20,21 and may in theory acidify vacuoles down to pH ≈ 3.5, when operating without kinetic inhibition, which it rarely if ever occurs in vivo, and in its “lowest gear” (H+/ATP = 2). Further acidification to pH < 3, as in lemon vacuoles, would require H+/ATP ratios <2 (ref. 20). Acid lemons indeed contain a proton-pump activity with H+/ATP = 1 stoichiometry8,12,14,15. However, the nature of this fruit-specific proton pump has remained elusive ever since, because it could not be completely purified13,15.

In petunia, mutations in one of the seven PH loci (PH1PH7) reduce the acidity of petal vacuoles and petal homogenates, resulting in a blue flower color22,23. PH3 and PH4 encode transcription factors of the WRKY and MYB family, respectively, which together with the helix–loop–helix protein ANTHOCYANIN1 (AN1) and the WD-repeat protein AN11, form a complex (WMBW) that activates genes involved in vacuolar acidification24,25,26,27. PH1 and PH5 are the major downstream genes involved in vacuolar hyperacidification28. PH5 encodes a P3A-ATPase proton pump that resides in the tonoplast instead of the plasma membrane where other P3A-ATPases reside29,30. PH1 is a P3B-ATPase similar to bacterial Mg transporters and also resides in the tonoplast. It has no known transport activity on its own but can bind to PH5 and promote PH5 proton-pumping activity28 and has an additional role in membrane/protein trafficking to the vacuole31.

PH1 and PH5 homologs are widely conserved among flowering plants, including species without colored petals, suggesting that their function is not confined to flower pigmentation30,32. Since CitPH5 homologs are expressed in sour lemons and oranges33,34,35 and P-ATPase proton pumps can theoretically generate steeper proton gradients (because H+/ATP = 1), we investigated whether the PH5/PH1 complex might be the proton pump that acidifies Citrus vacuoles to such extreme extent.

Analyzing a collection of lemon, orange, and pummelo varieties (Supplementary Table 1), we found that PH1 and PH5 homologs are highly expressed in all acidic (low pH) fruits but are downregulated in non-acidic (high pH) fruits, due to inactivating mutations in CitAN1 (sweet lemon and sweet oranges) or regulatory mutations that inactivate CitAN1, CitPH3, and/or CitPH4 expression.


CitPH1 and CitPH5 expression in Faris lemons

Lemon trees of the variety ‘Faris’ produce branches bearing either sour or sweet (non sour) fruits, enabling a comparison of sweet and sour fruits grown in the same conditions33. However, sweet and sour fruits are not necessarily isogenic, because ‘Faris’ is a graft chimera in which the L1 tunica layer of the shoot meristem derives from an unknown variety related to ‘Millsweet’ limetta and the L2 layer from a standard sour lemon33. Branches of ‘Faris’ trees carrying sour fruits (Fso) have purple immature leaves, a purple blush on the lower petal epidermis, and dark spots at the chalazal end of the seeds, whereas the branches bearing sweet fruits (Fsw) lack purple pigmentation on leaves, petals, and seeds (Fig. 1a). The juice of Fsw fruits is less acidic (pH 5.1) than that of Fso or ‘Frost Lisbon’ (Fli) (pH 2.5), a standard lemon with strong sour taste, and Fsw fruits contain less titratable acid than Fso or Fli fruits (Fig. 1b). There was, however, no correlation between soluble solid content (Brix) and juice taste (Fig. 1b), indicating that juice pH and/or titratable acid rather than the sugars, which are major components of total soluble solids, determine the different taste of Fsw and Fso fruits.

Fig. 1
figure 1

Analysis of ‘Frost Lisbon’ (Fli), ‘Faris’ sour (Fso), and ‘Faris’ sweet (Fsw) lemons. a Phenotypes of Fli, Fso, and Fsw fruits, flowers, and leaves. Scale bars represent 1 cm. b pH, titratable acid, and soluble solid content (°Brix) in juice from (nearly) mature Fli, Fso, and Fsw fruits (mean ± SE). c Real-time RT-PCR of CitPH1, CitPH5, and CitMAC9F1 mRNA in juice vesicles from (nearly) mature Fli, Fso, and Fsw fruits. d Real-time RT-PCR of CitAN1 mRNA in Fli, Fso and Fsw juice cells, using primers that amplify the 5’, middle, and 3’ part of the mRNA. e Structure of the CitAN1 alleles and their transcripts in Fso and Fsw juice vesicles. f Real-time RT-PCR of CitPH3, CitPH4, and CitAN11 mRNA in juice vesicles from (nearly) mature Fli, Fso, and Fsw fruits. Values in c, d, f are mean ± SE; n = number of samples from different fruits × number of technical replicates of each. Source data are provided as a Source Data file

In melon (Cucumis melo) and tomato (Solanum lycopersicum), which both lack PH1 and PH5 homologs30, the (mild) acidification of the fruit flesh requires a membrane transporter of unknown function, known in melon as PH or SOUR36. We identified the SOUR homolog of Citrus (CitSO) and found that CitSO mRNAs in Fsw and Fli fruits are expressed at similar levels and lack mutations with an obvious effect on the expression and/or activity of the encoded protein, indicating that CitSO does not contribute to the differences in their acidity (Supplementary Figs. 1 and 2).

Analysis of the Citrus limon homologs CitPH1 and CitPH5 identified previously30 revealed that both are expressed in low-pH Fli and Fso lemons but not in high-pH Fsw lemons (Fig. 1c, Supplementary Fig. 3a). Expression of the Citrus homolog of MAC9F1, a petunia gene of unknown function that is activated by the same transcription factors as PH1 and PH537, is also downregulated in Fsw juice vesicles similar to CitPH1 and CitPH5 (Fig. 1c).

These findings indicate that (i) the CitPH1-PH5 heterodimeric proton pump is involved in the hyperacidification of Fso and Fli fruits and (ii) that the reduced acidity of Fsw fruits results from a mutation in a regulatory gene controlling the expression of CitPH1, CitPH5, and other genes like CitMAC9F1.

Inactivation of CitAN1 causes the reduced acidity in Fsw

To identify the mutation(s) responsible for the reduced expression of CitPH1 and CitPH5 in Fsw fruits, we identified Citrus homologs of the transcription regulators AN1, AN11, PH4, and PH3 (Supplementary Fig. 4), which drive PH1 and PH5 expression in petunia petals and seeds28,29. Public RNA-seq data38 indicate that in the Citrus sinensis (sweet orange) variety ‘Valencia’ CitPH4 is predominantly expressed in fruits, like CitPH5 and CitMAC9F1, whereas CitAN1 is also expressed in flowers, where it most likely drives anthocyanin synthesis in concert with other MYB proteins (Supplementary Fig. 5).

Reverse transcriptase–polymerase chain reaction (RT-PCR) experiments using primers complementary to the 5′ and 3′ end of the CitAN1 coding sequence revealed that Fli and Fso juice vesicles express full-size CitAN1 mRNAs (Fig. 1d; Supplementary Fig. 3b), which encode a functional protein (see below). Both CitAN1 alleles of Fso juice vesicles consist of seven exons, of which one (exon 4) is skipped in a fraction of the mature CitAN1 mRNA (Fig. 1e; Supplementary Fig. 3), and are distinguishable by sequence polymorphisms in CitAN1 (single-nucleotide polymorphism (SNPs) and a 1661-bp transposon insertion in the 5′ flanking region) and in the genes immediately upstream (CitFAR-like) and downstream (CitTFIIH-like) (Supplementary Figs. 68).

Fsw fruits, by contrast, lack full-size CitAN1 mRNAs and instead express truncated CitAN1 transcripts, which span the 5′ and middle region of the mRNA but lack the 3′ part (Fig. 1d, e and Supplementary Fig. 3). Analysis of genomic and 3′-RACE cDNA fragments revealed that the citan1Fsw allele(s) contain an identical 1.3-kb deletion with breakpoints in exon 7 and 143 bp downstream the normal polyadenylation site, resulting in a transcript that lacks exon 8 and most of exon 7 to terminate at a cryptic polyadenylation site 57 bp downstream from the deletion breakpoint (Fig. 1e and Supplementary Fig. 9). Sequencing of PCR products from CitAN1 and the flanking genes CitFAR-like and CitTFIIH-like revealed no sequence polymorphism (Supplementary Fig. 8). This indicates that in Fsw fruits the genomic CitAN1 region is either homozygous or hemizygous over a larger deletion in the sister chromosome that spans CitAN1, CitFAR-like, and CitTFIIH-like.

Transcripts of the PH4 homolog CitPH4 accumulate at similar levels in Fso and Fsw fruits (Fig. 1f), excluding that the high-pH phenotype results from downregulation of CitPH4. Fso and Fsw fruit both contain two distinct CitPH4 alleles distinguishable by few SNPs and triplet repeats of variable length, resulting in several amino acid replacements and polyglutamine (Q) tracks of variable length, which may affect protein function and/or stability (Supplementary Figs. 1011). The sequences of CitAN11Fso and CitAN11Fsw did not reveal differences with an obvious negative effect on Fsw protein activity (Supplementary Fig. 12), and in fruits, mRNA expression levels of the two alleles are comparable (Fig. 1e).

The finding that juice vesicles of Fso and Fsw fruits contain distinct alleles for multiple genes, including CitAN1, CitPH4, CitAN11, CitFAR-like, and CitTFIID-like, suggests that the occurrence of sweet and sour fruits results from atypical periclinal cell division(s) by which a daughter of a cell in the L2 meristem layer, containing the genome of a standard sour lemon, invaded L1 and displaced the mutant ‘Millsweet’-like L1 cells, or vice versa, as in other chimeras39,40, rather than the somatic reversion of an unstable (epi)allele.

To examine whether CitPH1 and CitPH5 are transcriptionally activated by CitAN1 and CitPH4 and whether the deletion in CitAN1 and/or the polymorphisms in CitPH4 coding sequence impair the transcription of CitPH1 and CitPH5 in Fsw lemons, we performed transient expression in protoplasts from ph4 petunia petals, where endogenous PH1 and PH5 promoters are inactive. Therefore, we expressed various combinations of CitAN1Fso, CitAN1Fsw, and green fluorescent protein (GFP) fusions of CitPH4Fso or CitPH4Fsw from the constitutive 35S promoter and measured the expression of red fluorescent protein (RFP) reporter genes that were translationally fused to the CitPH1Fso or CitPH5Fso promoters (Fig. 2a). To identify transformed cells and normalize the expression levels for transformation efficiency, we co-transformed p35S:AHA10-GFP, which encodes a GFP fusion of the Arabidopsis PH5-homolog AHA10 that localizes in tonoplast of the central vacuole32, and the nuclear localized GFP-CitPH4.

Fig. 2
figure 2

Activation of CitPH1 and CitPH5 promoters by CitAN1 and CitPH4. a Structure of effector, reporter, and control genes used for transient transformation of protoplast from the epidermis of petunia ph4 petals. b Relative expression (real-time RT-PCR of effector (CitAN1 and CitPH4) and reporter (pCitPH1:RFP and pCitPH5:RFP) mRNA expression in petal protoplasts transformed with different effectors [mean ± SE from 2 technical PCR replicates of 4 independent transformation experiments (n = 4 × 2)]. Expression of the co-transformed 35S:AHA10-GFP gene was used for normalization. c Confocal micrographs of petunia ph4 epidermal petal protoplasts, with anthocyanins in the central vacuole, from stage 6 flowers, 24 h after transformation of 35S:AHA10-GFP, the reporters pCitPH1:RFP, pCitPH5:RFP, and different combinations of the effectors 35S:GFP-CitPH4Fso, 35S:GFP-CitPH4Fsw, 35S:CitAN1Fso, and 35S:CitAN1Fsw. Note that GFP-CitPH4 expression is visible in the nucleus (green arrows). Images shown are representative of tens to hundreds of cells observed. Size bars equal 10 µm. Source data are provided as a Source Data file

Expression of either CitAN1 or GFP-CitPH4 alone was insufficient to induce the CitPH1 or CitPH5 reporters in protoplasts (Fig. 2b, c; Supplementary Fig. 13). However, co-expression of CitAN1 and GFP-CitPH4 strongly induced pCitPH1:RFP and pCitPH5:RFP expression (Fig. 2b) in most if not all cells expressing GFP-CitPH4 and AHA10-GFP (Fig. 2c) and was independent from other (petunia) regulators of the anthocyanin/pH pathway as it also occurred in white mesophyll cells (Supplementary Fig. 13) where endogenous AN and PH genes are not expressed26,27,29. The CitPH4FSo and CitPH4Fsw alleles activated the CitPH1 and CitPH5 reporters with similar efficiency, whereas the truncated AN1 protein encoded by an1Fsw proved unable to induce pCitPH1:RFP or pCitPH5:RFP expression.

These findings indicate that the mutation in citan1Fsw is responsible for the reduced CitPH1 and CitPH5 expression in the Fsw fruits and very likely also the loss of anthocyanins in leaves and flowers on the branches bearing Fsw fruits, whereas the polymorphisms in the CitPH4Fsw and CitPH4Fso have little or no effect. The reduced expression of CitPH3 in Fsw juice cells (Fig. 1f) is likely due to the citan1Fsw mutation, since in petunia petals PH3 is (partially) regulated by AN1 and PH4 and essential for transcription of PH1 and PH527.

CitPH1 and CitPH5 expression in other sweet lemons

To support that the reduced acidity of Fsw fruit was caused by mutation in CitAN1 and consequent loss of CitPH1 and CitPH5 expression, rather than by independent mutations affecting unrelated pathway(s), we analyzed additional sweet and sour lemon varieties (Fig. 3a). ‘Schaub’ rough lemon is a non-edible citrus variety, unrelated to standard lemons and generally used as rootstock. ‘Millsweet’ limetta (Citrus limetta) and unnamed Sweet lemon (Citrus limettioides) bear sweet-tasting fruits, which have a similar soluble solid content (Brix) but have reduced juice acidity and titratable acid content. as well as CitPH1, CitPH5, and CitMAC9F1 mRNA expression compared to ‘Schaub’ and ‘Frost Lisbon’ fruits (Fig. 3b, c).

Fig. 3
figure 3

Analysis of ‘Schaub’, ‘Amber’, ‘Millsweet’, and Sweet lemon varieties. a Fruits, flower, and leaf phenotypes of the analyzed varieties. Size bars represent 1 cm. b pH, titratable acid, and soluble solid content (°Brix) in juice from different varieties (mean ± SE). c Real-time RT-PCR of CitPH1, CitPH5, and CitMAC9F1 mRNA juice vesicles. d Real-time RT-PCR of CitAN1, CitPH3, CitPH4, and CitAN11 mRNA in juice vesicles. Values in c, d are mean ± SE; n = number samples from different fruits × number of technical replicates of each. For c, d, full-size nearly mature fruits were analyzed. Relative mRNA expression levels from ‘Frost Lisbon’ were taken from Fig. 1 and are repeated here for comparison. Source data are provided as a Source Data file

CitAN1, CitPH3, and CitPH4 transcripts are low in abundance in ‘Millsweet’ limetta and Sweet lemon compared to low-pH fruits, whereas CitAN11 is expressed at similar levels as in low-pH fruits (Fig. 3d). ‘Schaub’, sweet ‘Amber’, and sour ‘Amber’ lemons have two CitAN1 alleles with wild-type structure (Fig. 4), differing only for a few SNPs in CitAN1 and the flanking genes CitFAR-like and CitTFIIH-like (Supplementary Figs. 89). ‘Millsweet’ instead contains one allele with a wild-type structure and a truncated allele (citan1Msw) with a 1.3-kb deletion. The deletion in citan1Msw has the same breakpoints as that in citan1Fsw (Fig. 4, Supplementary Fig. 9) suggesting that both alleles originate from the same deletion event, which apparently occurred in the distant past as citan1Fsw and citan1Msw and the flanking CitFAR-LIKE and CitTFIIH-like genes acquired since then several polymorphisms (Supplementary Figs. 89). As the first plant containing this 3’ deletion allele of CitAN1 was most likely heterozygous, homozygotes appeared later in progenies segregating for the deletion allele from various crosses, because of which today’s varieties can be homozygous (Fsw) or heterozygous (Msw) for this allele.

Fig. 4
figure 4

Structure and expression of CitAN1 alleles in distinct lemon, orange, and pummelo varieties. Exons are indicated by rectangles with 5’ and 3’ untranslated regions at reduced height. Deleted sequences are indicated with dotted red lines and pink filling and transposon insertions by triangles with different colors. The hooked arrow marks the transcription start. The red dotted hooked arrows in ‘Millsweet’ and Sweet lemon indicate (inferred) reduced transcription caused by reduced activity of an upstream regulator(s). Gene structures confirmed by sequencing are depicted as dark gray lines (introns and flanking DNA), filled rectangles (exons), or filled triangles (transposon insertion). Gene structures inferred from polymerase chain reaction alone are marked by gray lines and white filling

PCR analysis and partial sequencing showed that Sweet lemon contains a CitAN1 allele of normal size and a truncated allele with a deletion in 3′ end that may be similar to the deletion in citan1Fsw and citan1Msw (Supplementary Fig. 14). One of the Sweet lemon alleles, presumably the full-size allele, contains in its upstream region the same 1.7-kb transposon insertion as found in the ‘Faris’ sour and ‘Frost Lisbon’ alleles (Fig. 4; Supplementary Fig. 6). Given that, in ‘Millsweet’ limetta and Sweet lemon fruits mRNAs from both the full-size and a truncated CitAN1 allele are downregulated, in spite of their very different origin, it is most likely that their reduced expression is caused by a mutation in an upstream trans-acting factor, rather than independent cis-acting mutations in each allele.

‘Amber’ is a chimera, like ‘Faris’, that originated as a variant tree in a grove of ‘Eureka’ lemons. Some branches on the ‘Amber’ tree bear pale green leaves, white flowers, and sweet-tasting fruits with “amber”-colored flesh, while other branches bear purplish young leaves, flowers with a purple blush, and fruits with yellow flesh and sour taste (Fig. 3a). Sweet ‘Amber’ lemons have a low juice pH similar to sour ‘Amber’, ‘Schaub’, and ‘Frost Lisbon’ lemons and express similar CitPH1 and CitPH5 mRNA levels as sour ‘Amber’ fruits and only slightly less than in ‘Frost Lisbon’. Expression of the CitAN1 allele(s) in ‘Amber’ sour fruits, which contain the same transposon insertion as CitAN1Fso, and those in sweet ‘Amber’ fruit, which lack this insertion (Supplementary Fig. 6), differs less than two-fold, indicating that this transposon has little or no effect on the expression of CitAN1 and downstream genes. Thus ‘Amber’ sweet is essentially a low-pH variety that owes its sweet taste not from increased juice pH, but from the reduced buffer capacity, as measured by the amount of titratable acid (Fig. 3b), which might result from genetic defects in distinct pathways affecting, for example, citrate transport into the vacuole.

Taken together, we found that all varieties with low-pH lemons (‘Frost Lisbon’, ‘Schaub’, sour ‘Amber’, sweet ‘Amber’) express relatively high levels of CitPH1 and CitPH5 mRNA, whereas in all varieties with strongly reduced acidity (‘Faris sweet’, ‘Millsweet’, Sweet lemon) CitPH1 and CitPH5 expression is reduced due to independent mutations that affect upstream transcription regulators. This strongly supports the view that CitPH1 and CitPH5 are essential for the hyperacidification of vacuoles in lemon juice vesicles and sour taste of the fruit.

CitAN1 and CitPH4 are downregulated in high-pH oranges

Next, we extended our investigation to orange varieties (Fig. 5a). Sweet (Citrus sinensis) and sour oranges (Citrus aurantium) are both hybrids originating from pummelo and mandarin1,3, with tastes ranging from sharply sour to insipidly sweet. We analyzed ‘Pineapple’ orange, a variety with moderate acid and sweet-rich taste, three varieties with a bland (non-sour) taste (‘Lima’, ‘Vaniglia’, and ‘Orange of Heaven’), and sour oranges from two trees growing on different locations on Gran Canaria Island (small and large fruits) and one from Ostia, Italy. ‘Pineapple’ orange juice vesicles express CitPH1 and CitPH5 and CitMAC9F1 at similar levels as sour oranges, which correlates with the low pH of the fruit juice, whereas the titratable acid content is much lower than in sour oranges (Fig. 5b, c).

Fig. 5
figure 5

Analysis of sweet and sour orange varieties. a Fruits, flower, and leaf phenotypes of the analyzed varieties. Size bars represent 1 cm. b pH, titratable acid, and soluble solid content (°Brix) in juice from full colored mature fruits (mean ± SE; n ≥ 3). Note that titratable acid was not determined for Ostia (nd). c Real-time RT-PCR of CitPH1, CitPH5, and CitMAC9F1 mRNA in juice vesicles from full colored mature fruits. d Real-time RT-PCR of CitAN1, CitPH3, CitPH4, and CitAN11 mRNA in juice vesicles from full colored mature fruits. Values in c, d are mean ± SE; n = number samples from different fruits × number of technical replicates of each. Source data are provided as a Source Data file

The juice of ‘Lima’, ‘Vaniglia’, and ‘Orange of Heaven’ oranges has higher pH and lower titratable acidity than juice from the sour varieties from Ostia, Gran Canaria1 (GC1) and GC2, whereas the soluble solid content (Brix) is similar and therefore does not contribute to the taste differences (Fig. 5b). The high pH of ‘Lima’, ‘Vaniglia’, and ‘Orange of Heaven’ fruits correlates with strongly reduced amounts of CitPH1, CitPH5, and CitMAC9F1 mRNAs and approximately four-fold less CitPH3 mRNA as compared to the low-pH fruits from ‘Pineapple’ orange and the trees from Ostia and Gran Canaria (GC1, GC2), possibly due to the strongly reduced levels of CitAN1 and CitPH4 mRNA (Fig. 5c, d). RNA-seq data revealed that CitPH5, CitPH4, and CitAN1 mRNAs are also downregulated in the ‘Sucarri’ and ‘Bintang’ sweet oranges35.

The low-pH oranges from Gran Canaria (GC1 and GC2) and Ostia all have two CitAN1 alleles, which differ for a few SNPs and the presence/absence of the same transposon insertion as in CitAN1Fso and CitAN1Aso (Supplementary Figs. 67; Fig. 4). The citan1 allele(s) of the high-pH ‘Lima’ and ‘Orange of Heaven’ oranges are disrupted by the insertion of a 6.9-kb transposable element containing 39-bp inverted repeats (MudR-like element) in intron 2. The citan1 allele(s) of ‘Vaniglia’ are disrupted by the insertion of a 5.3-kb copia-like TCS1 retrotransposon in exon 6 (Fig. 4). The downregulation of CitPH4 is most likely due to an independent mutation, as CitPH4 expression in lemons is independent from CitAN1 (Fig. 1).

No relevant differences in the expression of CitAN11 were detected in fruits of low- and high-pH orange varieties, while CitPH3 is low expressed in high-pH oranges as compared to the low-pH ones, consistent with its (partial) regulation by CitAN1 and CitPH427. Because the CitPH4 alleles from these varieties contain no mutations in the coding sequence (Supplementary Fig. 10), it is conceivable that transcription of CitPH4 (and possibly CitAN1) transcription is reduced by additional mutations in higher-rank regulator(s).

These data show that CitPH1 and CitPH5 are also in oranges responsible for hyperacidification in juice vesicles and that independent mutations in CitAN1 and an (unknown) upstream regulator of CitPH4 have been selected to obtain fruit varieties with decreased acidity.

Analyses of acidless pummelos and rangpur limes

We further broadened our survey to a group of pummelo accessions (Fig. 6a). Pummelos (C. maxima) are non-hybrid citrus fruits1,3. They are similar to large grapefruit, variable in sweetness, and native of Southeast Asia but now grown in tropical and subtropical areas all over Asia and the Pacific Islands as well as in California and Florida.

Fig. 6
figure 6

Analysis of sweet and sour pummelo and rangpur lime varieties. a Fruits of the analyzed pummelo varieties. Size bars represent 1 cm. b pH, titratable acid, and soluble solid content (Brix) in juice from the analyzed varieties (mean ± SE). c Real-time RT-PCR of CitPH1, CitPH5, and CitMAC9F1 mRNA in juice vesicles. d Real-time RT-PCR of CitAN1, CitPH3, CitPH4, and CitAN11 mRNA in juice vesicles. e Fruits, flower, and leaf phenotypes of the analyzed rangpur limes. Size bars represent 1 cm. Values in c, d are mean ± SE; n = number samples from different fruits × number of technical replicates of each. Fruits analyzed in bd were mature with a slight green blush (pummelos) or full colored mature (rangpurs and rangpur × pummelo hybrids). Images of Punk and ‘Chandler’ fruits were provided by and reproduced with permission of the UC-Riverside Citrus Variety collection. Source data are provided as a Source Data file

‘Pin Shan Kong Yau’ (PSKY) and an unnamed ‘Kao Panne’ pummelo (Punk) are acidic pummelos with blond fruit flesh (Fig. 6a), whereas the unrelated Siamese pummelo P2240 (also known as ‘Siamese sweet’) is an acidless pummelo with blond flesh that is commonly used in breeding to reduce acidity in the progeny. The “sweet” (non-sour) taste of P2240 fruits correlated with reduced acidity and titratable acid content of the fruit juice and reduced CitPH1, CitPH5, and CitMAC9F1 mRNA levels in juice vesicles, as compared to the low-pH PSKY and Punk fruits (Fig. 6b, c). In the pink and mildly sweet-tasting fruits of ‘Chandler’, a hybrid of P2240 and an acidic accession (P2241), the juice pH as well as CitPH1, CitPH5, and CitMAC9F1 mRNA levels are similar to those of the low-pH PSKY and Punk fruits. This indicates the acidification of pummelo fruits also depends on CitPH1 and CitPH5 and that the causative mutation(s) that reduce the acidity and expression of CitPH1 and CitPH5 in P2240 fruits (are) recessive.

Since the reduced acidity of P2240 fruits is associated with downregulation of at least three genes (Fig. 6c), the causative mutation(s) most likely affect one of more upstream transcription activators. Because the CitAN1, CitPH3, and CitPH4 alleles of P2240 express in juice vesicles similar mRNA amounts as those from Punk and ‘Chandler’ (Fig. 6d), and because their coding sequences do not contain polymorphisms with obvious negative effects on protein activity (Supplementary Figs. 11 and 15), the causative mutation likely affects an unknown transcription factor that operates in concert with CitAN1, CitPH3, and CitPH4, which may be encoded or controlled by an unknown locus (acitric) within 1.2 cM of the RFLP marker RFZ20 (ref. 41), which maps to genome sequences on chromosome 2.

The rangpur limes (Citrus limonia) ‘Philippine’ and ‘Weirick’ have flowers and leaves with anthocyanin pigments (Fig. 6e). While soluble solid content (Brix) is similar, the acidity and amount of titratable acid is reduced in juice from sweet tasting ‘Weirick’ fruit, compared to the sour fruits from ‘Philippine’ (Fig. 6b). This correlates with strongly reduced mRNA expression levels of CitPH1, CitPH5, and CitMAC9F1, most likely due to reduced activity of an upstream transcription factor (Fig. 6c). CitAN1 and CitPH4 transcripts accumulate at similar levels in sour ‘Philippine’ and sweet ‘Weirick’ juice vesicles, whereas CitPH3 mRNA was essentially abolished in ‘Weirick’ fruits (Fig. 6d). PH3 is in petunia, like other WMBW complex components (AN1, PH4, AN11), essential for PH1, PH5, and MAC9F1 expression27, suggesting that the strongly reduced CitPH1 and CitPH5 expression in ‘Weirick’ fruits is most likely caused by abolished CitPH3 expression. We found no obvious defects in the coding part of the ‘Weirick’ CitPH3 allele (Supplementary Fig. 16) or any large rearrangements in its promoter that may account for the reduced CitPH3 transcription or mRNA processing. Hence, we infer that the reduced acidity and titratable acid is most likely due to a mutation in an (unknown) upstream regulator(s), which strongly reduces the expression of CitPH3 and its target genes CitPH1, CitPH5, and CitMAC9F1.

The hybrid ‘Weirick’ × P2240 (7D-76-03) had sour fruits with high titratable acid, indicating that loss of acidity in ‘Weirick’ and P2240 is due to recessive mutations in distinct genes. Since 7D-76-03 is no longer available, we analyzed progeny with sour (6C-32-01) and sweet fruits (6C-32-02) from the cross ‘Chandler’ × 7D-76-03 in which segregation of acidity is observed. Whereas soluble solid content (Brix) is similar in both fruits, juice acidity, titratable acid content, and the expression of CitPH1, CitPH5, and CitMAC9F1 are all reduced in 6C-32-02 fruits compared to 6C-32-01. The latter is probably due to mutations originating from P2240 rather than from ‘Weirick’, as CitPH3 expression in 6C-32-02 is strongly increased compared to ‘Weirick’ (Fig. 6d).


How juice cell vacuoles in sour citrus fruits can be acidified to such an extreme extent is a long-standing question, as vacuolar proton pumps capable of generating the required steep pH gradient across the tonoplast were unknown. We have shown that juice vesicles of Citrus varieties with acidic (low pH) fruits express CitPH1 and CitPH5, encoding two interacting P-ATPases that constitute a vacuolar proton pump, while CitPH1 and CitPH5 expression levels are drastically decreased in fruit varieties with reduced acidity (high pH). The downregulation of CitPH1 and CitPH5 in distinct fruits results from independent mutations in multiple genes required for CitPH1 and CitPH5 expression, such as, for example, CitAN1, or upstream regulators thereof, suggesting that these are the causative mutations for the loss of acidity rather than being linked to them. These results indicate that the long-sought vacuolar proton pump is a P-ATPase complex, encoded by CitPH1 and CitPH5, previously identified for its role in the pigmentation of flowers and seeds. The coupling ratio of P-ATPases (H+/ATP = 1), the sensitivity of PH1/PH5 activity to vanadate, and its insensitivity to bafilomycin28 fit perfectly with the biochemical properties of the proton-pump activity that is expressed in sour (low pH) fruits but absent in sweet fruit varieties8,12,13,14,15.

While our results show that the large pH differences between acidic (low pH) and acidless (high pH) varieties within a Citrus group (lemons, oranges, or pummelos) are due to differences in CitPH1 and CitPH5 expression, the cause(s) of the much smaller pH variations between acidic (non-mutant) varieties of different Citrus groups or the even smaller pH differences within a group remain unclear. Given that the analyzed varieties are not isogenic and not grown under identical circumstances, such small pH differences may originate from small differences in the expression of CitPH1/CitPH5 or other transporters, like CitSO or various vacuolar antiporters that import solutes in exchange for protons17.

It is noteworthy that fruits with strongly reduced CitPH1 and CitPH5 expression all contain reduced amounts of titratable acids, which is in Citrus mostly citric acid. This provides in vivo support of biochemical data, which indicated that most of the citrate transport into vacuoles is driven by the H+ gradient across the tonoplast (ΔpH), while only a small part relies on ATP-driven transporters9. Hence, we infer that CitPH1 and CitPH5 promote sour taste by (i) hyperacidifying vacuoles resulting in low pH of the fruit (juice) and by (ii) generating the steep pH gradient required for the import and sequestration of citrate and conjugate bases of other acids into the vacuole. The latter increases pH-buffering capacity of the juice and prevents the juice pH from being neutralized by saliva before the low pH can be sensed.

Citrus varieties have been subject to cultivation and selection for several thousands of years3. Our data show that citrus varieties with reduced acidity arose multiple times independently in different citrus lineages through mutations disrupting the expression of genes that encode transcription activators of CitPH1 and CitPH5. The finding that the citan1Fsw and citan1Msw alleles contain several SNPs suggests that the inactivating 3′ deletion and, hence, varieties with reduced acidity, arose in the distant past, possibly hundreds if not thousands of years ago during early stages of Citrus domestication. It is, however, difficult to give a precise timing without a better estimate of mutation rates in (inactivated) Citrus genes that are not under selection.

Inspection of public RNA-seq data42 indicates that Malus domestica homologs of PH1 (MDP0000319016) and PH5 (MDP0000303799) are expressed in developing apples and RT-PCR data30 indicate that the Vitis vinifera homologs VvPH1 and VvPH5 are expressed in developing grape berries. Hence, PH1 and PH5 and genes encoding upstream transcription activators are likely to be important determinants of the acidity and taste in many other fruits besides Citrus.

Taken together, our genetic data show that a vacuolar proton pump consisting of the P-ATPases PH1 and PH5 is required for the hyperacidification of vacuoles in juice vesicles and the very sour taste of Citrus fruits and juices and that over thousands of years of Citrus breeding “sweet” (non sour) tasting varieties were obtained many times via independent mutations in distinct transcription regulators driving CitPH1 and CitPH5 expression. This opens the way to develop molecular markers for fruit acidity and taste to speed up the breeding in Citrus and other fruit crops, most of which are trees or shrubs with long generation times.


Plant material

Lemon, sweet orange, pummelo, and rangpur lime fruits were collected from trees belonging to the Citrus Variety Collection or from other citrus orchards on the Agricultural Experiment Station at the University of California, Riverside, CA. Sour oranges were collected from a tree in Ostia, Italy (Ostia orange) and from trees grown in the botanical garden Jardin Botanico Canario Viera y Clavijo (GC1) and from the town Agüimes (GC2) in Gran Canaria, Spain. For transient expression assays and isolation of protoplasts, we used petals from the petunia line V74 (ph4 mutant) grown under normal greenhouse conditions.

Fruit taste parameters

Vesicles were excised from the fruits and frozen in liquid nitrogen. After grinding 800 mg of frozen pulp and dissolving in 6 ml distilled water, pH was immediately measured with a pH meter (Consort P901). Brix was determined by directly reading the juice on a refractometer (Marius, Poland). The acid content (expressed as citric acid equivalents) was evaluated by titration using sodium hydroxide (0.1 M) and a phenolphthalein pH indicator. To deliver and measure the volumes, we use a volumetric buret.

Citrus homologs of petunia AN and PH genes and melon SO

To identify Citrus homologs of petunia and melon genes, we searched Citrus genome sequences at Phytozome ( and Citrus sinensis Annotation Project ( with BLAST and confirmed homology by phylogenetic analysis30. Protein sequences were aligned using MUSCLE, and after curation by GBLOCKS phylogenetic trees were constructed with maximum likelihood (PHYML) using online tools43. For comparison of DNA and proteins, sequences were aligned with Clustal-Omega ( and MUSCLE ( respectively, optimized by hand using Aliview44, and colored using BOXSHADE (

CitAN1, CitPH4, CitPH3, and CitAN11 alleles from citrus varieties (Supplementary Table 2) were isolated by PCR with primers amplifying the entire genomic sequence (for CitAN1 gDNA primers 7642 and 7590, for CitAN1 cDNA primers 7642 and 6592, for CitPH4 gDNA and cDNA primers 6572 and 6573, for CitPH3 cDNA primers 6611 and 6612, for CitAN11 cDNA primers 6613 and 6615). PCR fragments were then directly sequenced or cloned into pDONR P1-P2 by Gateway cloning system. CitSO cDNAs were amplified by RT-PCR from juice vesicle RNA using primers 8932 and 8933.

To isolate CitAN1 promoters, PCR fragments were generated using primers covering the 4-kb promoter (primers 7674 and 7473) and used directly for sequencing.

To isolate the flanking sequences of the CitAN1 gene, we designed primers to amplify fragments (1000 bp) of the FAR-like gene located 29 kb upstream of the start of the AN1 coding sequence (primers 8066 and 8068) and of the TFIIH gene (570 bp) located 617 bp downstream of the AN1 stop codon (primers 8063 and 8065). PCR fragments were directly sequenced and nucleotide polymorphisms were depicted using the IUPAC code. Primer sequences are shown in Supplementary Tables 37.

DNA and RNA isolation from Citrus vesicles

For DNA and RNA extraction, 800 mg of frozen vesicles were ground in liquid nitrogen and 7.5 ml of preheated (65 °C) extraction buffer (2% (w/v) CTAB, 2% (w/v) PVP (molecular weight 30,000–40,000), 25 mM EDTA pH 8.0, 2 M NaCl, 100 mM Tris-HCl (pH 8.0), 2% β-mercaptoethanol) was added to the frozen powder. To (re)adjust the solution to pH 7–8, we added for citrus vesicles from very acidic fruits (pH 2–2.5) approximately 800 μl 1 M Tris/HCl pH 9.0 and for less acidic or acidless fruits 400 μl Tris/HCl pH 9.0 and verified the pH using pH paper. The samples were then incubated at 65 °C for 15 min and extracted twice with chloroform–isoamyl alcohol (24:1), precipitated with 2-propanol, resuspended in sterile water, extracted with phenol/chloroform (1:1), precipitated with NaOAc and 2-propanol, washed in 70% ethanol, and resuspended in RNAse free water. Total RNA was precipitated using 1 volume of 4 M LiCl. The RNA pellet was washed with 70% ethanol, dissolved in sterile water, and quantified by measuring optical density (OD)260/280. To ensure the absence of genomic DNA, we performed a DNAse treatment.

Total DNA was obtained from the supernatant left after LiCl precipitation of RNA by precipitation with 2-propanol, washed with 70% ethanol, dissolved in water, and checked by agarose electrophoresis and OD260/280.

Expression analysis

To identify the truncated AN1 transcript in Fsw juice vesicles, we amplified the 3′ cDNA ends by 3′-RACE (5′/3′-RACe KIT 2nd generation; Roche). RT products from Fso and Fsw AN1 were amplified with a primer complementary to the 5′ untranslated region (UTR) (primer 6577) and an adaptor primer complementary to the poly(A) tail (primer 64a). Two nested PCRs were then performed using gene-specific primers: nested1 PCR: primer 6577 designed on the 5′ UTR and adaptor primer 65a (tail adaptor); nested2 PCR: primer 6579 designed on the exon 5 sequence and adaptor primer 65a.

RT-PCR analysis of CitPH1, CitPH5, CitAN1, CitPH3, CitPH4, CitAN11, and CitACTIN were performed as described previously24,26,45, using primers shown in Supplementary Table 3. cDNA products were amplified using primers specific for PH1 (primers 6524 and 4383), PH5 (primers 6530 and 6529), AN1 (primers 6623 and 6593, for the Fsw allele: primers 6623 and 6592), PH3 (primers 6538 and 6540), PH4 (primers 6544 and 6546), AN11 (primers 6614 and 6615), and ACTIN (primers 58a and 59a). All primer sequences are reported in Supplementary Table 3.

Quantitative RT-PCR was performed with an QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific) using the SensiMix (Bioline QT650–05) as described before27, using primers shown in Supplementary Table 8. Relative expression was calculated by normalizing against CitANKYRIN, CitANNEXIN2, and CitRIBOSOMAL PROTEIN S10, which are the most constantly expressed genes in broad range of tissues and Citrus species33, and CitACTIN11. For ‘Schaub’, ‘Amber’ sour, ‘Amber’ sweet, Sweet lemon, GC1, GC2, and Ostia, one biological replicate was analyzed and for all other varieties two or three fruits. For each fruit, two quantitative PCR reactions (technical replicates) were performed.

Gene constructs and transient expression in protoplasts

To generate 35S:CitAN1FsocDNA, the 2.2-kb full-size cDNA AN1 was amplified from Fso cDNA with primers 6623 and 6593 and used to generate a Gateway Entry clone by BP reaction with pDONR P1-P2 (Gateway system; Life Technologies, Invitrogen, Carlsbad, CA, USA) and then recombined into the pK2GW7.0 overexpression vector.

35S:CitAN1FswcDNA construct was obtained by amplifying the PCR fragment of Fsw cDNA, with primers 6623 and 6595 and cloning the 1.3-kb cDNA fragment in pDONR P1-P2. The insert was then recombined into pK2GW7.0 by Gateway recombination. The 35S:GFP-CitPH4Fso construct was generated by amplification of the PH4Fso cDNA with a 3′-RACE PCR primer (6572) and a 65a (oligo dT), cloning in pDONR P1-P2, and recombining in pK7WGF2.0 by Gateway system. The 35S:GFP-CitPH4Fsw construct was obtained similarly from cDNA of Fsw juice vesicles.

Promoter:RFP constructs of the target genes were generated as follows: a 3.3-kb PH1 promoter fragment was amplified with primers 7326 and 7328, cloned in pDONR P1-P2, and recombined into pWSK by Gateway system; a 3-kb PH5 promoter fragment was amplified with primers 7194 and 7195 and cloned with the same procedure. In 35S:AtAHA10-GFP, the 35S promoter drives expression of a translational fusion of the coding sequence of the AHA10 gene (including all introns) and GFP30,32. All primer sequences are reported in Supplementary Table 9. For isolation and transformation of petunia protoplasts28,31,32, chopped corollas of 15 open flowers (stage 6–7) from line V74 (ph4) were incubated in the dark for 16 h at room temperature in 0.2% (w/v) macerozyme R-10, 0.4% (w/v) cellulose R-10 in TEX buffer (3.1 g/l Gamborg’s B5 salts (Sigma-Aldrich), 500 mg/l 4-morpholineethanesulfonic acid (MES), 750 mg/l CaCl2, 250 mg/l NH4NO3, 136.9 g/l sucrose, pH 5.7) and centrifuged for 10 min at 700 RPM. The floating protoplasts were recovered by removing the underlying layers, washed twice in TEX buffer, and left for 2 h at 4 °C. After centrifugation (10 min, 700 RPM), floating protoplasts were recovered and resuspended in 5 ml MMM solution (0.1% (w/v) MES, 0.5 M Mannitol, 15 mM MgCl2). For transformation, 300 µl protoplast suspension, 20 µg plasmid DNA, and 300 µl PEG solution (0.4 M Mannitol, 0.1 M Ca(NO3)2, 40% (w/v) Polyethyleneglycol 4000, pH 8) were mixed, and after 1 min 2 ml of TEX buffer was added. After 2 h at room temperature, 5 ml W5 buffer (154 mM NaCl, 125 mM CaCl2, 5 mM KCl, 5 mM glucose) was added and protoplasts were pelleted for 5 min at 700 RPM, resuspended in 2 ml TEX buffer, and kept for 16 h in the dark at room temperature and then analyzed by confocal microscopy and RNA isolation.

Total RNA was extracted from protoplasts using the NucleoSpin® RNA II Kit (Macherey-Nagel, Düren, Germany). In the expression analysis after transient transformation, AHA10 mRNA was used as reference to normalize the data.

GFP, RFP, and anthocyanin fluorescence were imaged with a LSM Pascal Zeiss confocal microscope28,31,32.

Reporting Summary

Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.