ArticlePDF Available

The genome of the glasshouse plant noble rhubarb (Rheum nobile) provides a window into alpine adaptation

July 2023
Communications Biology 6(1)

DOI:10.1038/s42003-023-05044-1

License
CC BY 4.0

Authors:

Feng Tao

Chinese Academy of Sciences

Boas Pucker

Technische Universität Braunschweig

Tianhui Kuang

Chinese Academy of Sciences

Bo Song

Chinese Academy of Sciences

Show all 13 authorsHide

Glasshouse plants are species that trap warmth via specialized morphology and physiology, mimicking a human glasshouse. In the Himalayan alpine region, the highly specialized glasshouse morphology has independently evolved in distinct lineages to adapt to intensive UV radiation and low temperature. Here we demonstrate that the glasshouse structure – specialized cauline leaves – is highly effective in absorbing UV light but transmitting visible and infrared light, creating an optimal microclimate for the development of reproductive organs. We reveal that this glasshouse syndrome has evolved at least three times independently in the rhubarb genus Rheum. We report the genome sequence of the flagship glasshouse plant Rheum nobile and identify key genetic network modules in association with the morphological transition to specialized glasshouse leaves, including active secondary cell wall biogenesis, upregulated cuticular cutin biosynthesis, and suppression of photosynthesis and terpenoid biosynthesis. The distinct cell wall organization and cuticle development might be important for the specialized optical property of glasshouse leaves. We also find that the expansion of LTRs has likely played an important role in noble rhubarb adaptation to high elevation environments. Our study will enable additional comparative analyses to identify the genetic basis underlying the convergent occurrence of glasshouse syndrome.

Glasshouse plant morphology and its evolution in Rheum a Representatives of glasshouse plants. A–C: Rheum nobile, R. alexandrae var. 1 and R. alexandrae var. 2 (Polygonaceae). D, E: Ajuga lupulina var. lupulina and A. lupulina var. major (Lamiaceae). F–G: Saussurea obvallata, S. involucrata and S. velutina (Asteraceae). b Phylogeny of Rheum estimated from 132 orthologs using coalescent method with local posterior probabilities shown in node. The glasshouse lineages are labeled in red. Node bars represent 95% highest posterior densities (HPD) of divergence time estimated using MCMCTree.

…

The translucent leaves in the noble rhubarb function as glasshouse a A mature noble rhubarb plant in the Hengduan Mountains (N28.53°, E99.95°; Elev. 4500 m) with insert showing a morphological comparison between a photosynthetic basal leaf and a translucent leaf. b Spectral characteristics of bracts and leaves in light transmittance, showing the different patterns of light transmittance by bracts and leaves. Significance was statistically tested at six wavelengths (250–750 nm) using T test (n = 3 biologically independent samples; ***p < 0.001). c Temperature change inside and outside of the inflorescence glasshouse in a 48 h interval at mid-August in the eastern Himalayas (N29.40°, E94.90°; Elev. 4200 m). d The HPLC spectrum of UV-absorbing substances extracted from five tissue types; s: the flavonol standard isovitexin; 1–5: the compounds isolated from tissues. e The contents of the five flavonol species isolated from noble rhubarb tissues. Fruits and roots only contained traces of the flavonols and are not shown. In boxplots, the center, lower, and upper lines depict the median, 25th, and 75th percentile respectively, and whiskers represent maxima and minima.

…

Genome architecture and gene family expansion and contraction a Multi-dimensional display of genomic components of the noble rhubarb genome. The density was calculated per 100 Kb. b Frequency distributions of synonymous substitution rates (Ks) between homologous gene pairs in syntenic blocks of R. nobile-R. nobile, F. tataricum-F. tataricum and R. nobile-F. tataricum. Both species are members of the family Polygonaceae. c Number of significantly expanded (+) and contracted (-) gene families in nine Caryophyllales species with high-quality genome sequences available. d REVIGO clusters of overrepresented GO terms for significantly expanded gene families in the noble rhubarb genome. Each bubble represents a summarized GO term from the full GO list by reducing functional redundancies, and their closeness on the plot reflects their closeness in the GO graph, i.e. the semantic similarity. e KEGG enrichment of metabolic genes that were significantly expanded in the noble rhubarb genome.

…

Evolutionary dynamics of LTRs a Histogram of intact LTR insertions in R. nobile (top), Ru. hastalutus (middle) and F. tataricum (bottom). b Time-calibrated phylogeny of Rheum and its close relatives Rumex and Oxyria with the proliferation of LTRs labeled with red shapes. c Proportion of intact LTR insertions within the coding region of annotated genes and downstream (5 kb) or upstream (5 kb) of annotated genes in R. nobile (Rhno) and Ru. hastalutus (Ruha). d LTR insertions were enriched in the flanking regions of coding sequences (1 kb) in Rhno but not in Ruha. e, f Violin plots of the divergences (log10 transformed) between gene pairs at expression and tissue specificity levels (Wilcoxon rank sum test). Boxplots represent the median, 25th, and 75th percentile, respectively.

…

Morphological transition a The leaves were classified into three zones based on their morphology. b The phenotype of representative leaves in each zone. c Significantly differentially expressed genes (P < 0.05, |log2(fold-change)| > 1) from different comparisons between leaf types.

…

Figures - available from: Communications Biology

This content is subject to copyright. Terms and conditions apply.

Access to this full-text is provided by Springer Nature.

Learn more

Content available from Communications Biology

This content is subject to copyright. Terms and conditions apply.

ARTICLE

The genome of the glasshouse plant noble rhubarb

(Rheum nobile) provides a window into alpine

adaptation

Tao Feng 1,2,3,9, Boas Pucker 4,5,6,9, Tianhui Kuang2,9, Bo Song2,9, Ya Yang 7, Nan Lin1,3,

Huajie Zhang 1,3, Michael J. Moore 8, Samuel F. Brockington 4, Qingfeng Wang 1,3, Tao Deng2✉,

Hengchang Wang 1,3✉& Hang Sun 2✉

Glasshouse plants are species that trap warmth via specialized morphology and physiology,

mimicking a human glasshouse. In the Himalayan alpine region, the highly specialized glass-

house morphology has independently evolved in distinct lineages to adapt to intensive UV

radiation and low temperature. Here we demonstrate that the glasshouse structure –specialized

cauline leaves –is highly effective in absorbing UV light but transmitting visible and infrared light,

creating an optimal microclimate for the development of reproductive organs. We reveal

that this glasshouse syndrome has evolved at least three times independently in the rhubarb

genus Rheum. We report the genome sequence of the ﬂagship glasshouse plant Rheum nobile

and identify key genetic network modules in association with the morphological transition to

specialized glasshouse leaves, including active secondary cell wall biogenesis, upregulated

cuticular cutin biosynthesis, and suppression of photosynthesis and terpenoid biosynthesis. The

distinct cell wall organization and cuticle development might be important for the specialized

optical property of glasshouse leaves. We also ﬁnd that the expansion of LTRs has likely played

an important role in noble rhubarb adaptation to high elevation environments. Our study will

enable additional comparative analyses to identify the genetic basis underlying the convergent

occurrence of glasshouse syndrome.

https://doi.org/10.1038/s42003-023-05044-1 OPEN

1CAS Key Laboratory of Plant Germplasm Enhancement and Specialty Agriculture, Wuhan Botanical Garden, Chinese Academy of Sciences, Wuhan 430074,

China. 2CAS Key Laboratory for Plant Biodiversity and Biogeography of East Asia, Kunming Institute of Botany, Chinese Academy of Sciences, Kunming, Yunnan

650201, China. 3Center of Conservation Biology, Core Botanical Gardens, Chinese Academy of Sciences, Wuhan, Hubei 430074, China. 4Department of Plant

Sciences, University of Cambridge, Tennis Court Road, Cambridge CB2 3EA, UK. 5CeBiTec & Faculty of Biology, Bielefeld University, Universitaetsstrasse,

Bielefeld 33615, Germany. 6Institute of Plant Biology & BRICS, TU Braunschweig, 38106 Braunschweig, Germany. 7Department of Plant and Microbial Biology,

University of Minnesota, Twin Cities, St. Paul, MN 55108, USA. 8Department of Biology, Oberlin College, Oberlin, OH 44074, USA.

These authors contributed

equally: Tao Feng, Boas Pucker, Tianhui Kuang, Bo Song. ✉email: dengtao@mail.kib.ac.cn;hcwang@wbgcas.cn;sunhang@mail.kib.ac.cn

COMMUNICATIONS BIOLOGY | (2023) 6:706 | https://doi.org/10.1038/s42003-023 -05044-1 | www.nature.com/commsbio 1

1234567890():,;

Content courtesy of Springer Nature, terms of use apply. Rights reserved

Tertiary and Quaternary uplift of mountains has exposed

organisms to demanding alpine conditions and accelerated

the evolution of alpine biotas1,2. Plants have responded to

harsh alpine environments with a high degree of specialization1,3,

leading to diverse life forms including cushion plants, giant

rosettes and succulents3. In Himalayan regions, the world’s most

species-rich temperate alpine zone, specialized morphologies have

evolved in response to hostile environmental conditions including

low temperature, high solar radiation, strong winds and a short

growing season1. Specialized morphologies include woolly plants

(plants covered by dense woolly hairs), nodding plants (plants

with ﬂowers facing the ground) and glasshouse plants. Glasshouse

plants are perhaps the most striking, with inﬂorescences sheltered

by semi-translucent leaves that create a warmer interior and can

be compared to the glass in a greenhouse4–6. The most prominent

Himalayan alpine plant is noble rhubarb [Rheum nobile Hook.f.

& Thomson, Polygonaceae)], which is the ﬂagship glasshouse

plant adapted to high elevations (>4000 m)7.

In contrast to sympatric competitors that generally are dwarf or

prostrate, noble rhubarb grows to heights of up to 2 m when in

ﬂowering, making it highly conspicuous in the alpine region. The

remarkable glasshouse-like morphology was assumed to be essen-

tial for the plants to cope with low temperature and strong UV

radiation at high elevation8–10. This adaptive trait has also been

found to be important for the mutualism between the plants and

the pollinating seed-consuming Bradysia fungus gnats, providing

shelter for adult oviposition and larva development11–13.

Elucidating the genetic basis of adaptive traits is a central goal of

evolutionary genetics14. Genomic modiﬁcation associated with

adaptation to high elevations has been well documented in

animals15 but has been less explored in plants relative to the high

diversity of alpine ﬂora. But as iconic plants of the alpine landscape,

glasshouse plants have become the focus of greater ecological and

evolutionary interests8,10,12,13,16–19. Studies on the ecological

function of the specialized structures of alpine plants, such as

cushion-like leaf canopy20–23, hairy leaves and inﬂorescences24–26,

leafy bracts8,12,13, and nodding capitula27,28 have revealed that

these traits are particularly efﬁcient in heat-trapping. For example,

the temperature within the leaf canopy of the cushion plant Silene

acaulis (L.) Jacq. is typically 15 °C higher than the ambient tem-

perature during clear summer days20. Such thermal beneﬁts are

essential for growth, development, metabolism and reproduction of

plants that inhabit the consistently cold, windy environments of

alpine regions25. In addition, the downward orientation of ﬂowers

and glasshouse-like leaves are assumed to be helpful in protecting

the sensitive reproductive parts from UV radiation and frequent

storms10,12,13,27–29. Despite great progress toward understanding

the functional ecology of these extremophiles, the genetic basis

facilitating their fascinating adaptation is poorly understood, owing

to the lack of genomic information.

Here, we report the genome sequence of the ﬂagship glasshouse

plant noble rhubarb and integrate comparative genomic, tran-

scriptomic, and phytochemical data to provide insights into the

evolution of glasshouse morphology. We show that the glasshouse

leaves function as solar radiation ﬁlters that absorb UV light to

provide photoprotection for reproductive organs, and reveal that

this adaptive morphological syndrome has cryptically evolved in

at least three lineages in rhubarb. By deciphering the transition

pattern of the glasshouse syndrome, both in morphology and in

transcriptomic proﬁles, we identiﬁed key genetic network mod-

ules underlying the developmental differentiation of glasshouse

leaves from normal leaves. The data presented in this study lay

the foundation for further deciphering the genetic mechanisms

underlying the origin and evolution of glasshouse morphology.

Results

Multiple origins of the glasshouse syndrome in Rheum L.

Glasshouse plants are unique in the Eastern Asian alpine biome

and are recorded in several phylogenetically distant plant taxa,

including the ﬂowering plant families Lamiaceae, Asteraceae,

and Polygonaceae (Fig. 1a). The most notable glasshouse-like

morphology is found in Rheum L. (Polygonaceae), such as noble

rhubarb, a wild relative of commercial rhubarb. Noble rhubarb is

commonly named yellow tower in Chinese because the yellowish

leaves are reﬂexed and form a compact tower-like structure

5 0 25 0 (mya)

R. officinale

R. franzenbachii

R. alexandrae pop3

R. palmatum

R. nobile pop2

R. nobile pop1

R. forestii

R. alexandrae pop2

R. alexandrae pop1

R. rhabarbarum

R. kialense

R. delavayi

R. tanguticum

R. pumilum

R. moorcroftianum

R. rhomboideum

R. acuminatum

Oxyria sinensis

Oxyria digyna

Rumex acetosa

Rumex hastatulus

Rumex palustris

0.77

0.84

0.73

0.89

0.82

0.90

Fig. 1 Glasshouse plant morphology and its evolution in Rheum.aRepresentatives of glasshouse plants. A–C: Rheum nobile,R. alexandrae var. 1 and R.

alexandrae var. 2 (Polygonaceae). D, E: Ajuga lupulina var. lupulina and A. lupulina var. major (Lamiaceae). F–G: Saussurea obvallata,S. involucrata and S.

velutina (Asteraceae). bPhylogeny of Rheum estimated from 132 orthologs using coalescent method with local posterior probabilities shown in node. The

glasshouse lineages are labeled in red. Node bars represent 95% highest posterior densities (HPD) of divergence time estimated using MCMCTree.

ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-05044-1

2COMMUNICATIONS BIOLOGY | (2023) 6:706 | https://doi.org/10.1038/s42003-023-05044-1 | www.nature.com/commsbio

Content courtesy of Springer Nature, terms of use apply. Rights reserved

covering the sensitive reproductive organs (Fig. 1a). Of the

60 species described in Rheum, two glasshouse plant species, R.

nobile and R. alexandrae, have been recorded7. To track the origin

and evolution of the glasshouse syndrome in Rheum, six newly

generated transcriptomes together with 16 publicly available tran-

scriptomes covering four of the six sections of Rheum were used in

phylogenomic reconstruction (Supplementary Data 1). Previous

phylogenetic analyses30 using a few cpDNA fragments failed to

resolve the deep phylogenetic relations within Rheum. Our phy-

logenomic analysis with 132 genes generated a high-quality phy-

logeny of Rheum (Fig. 1b). R. nobile was sister to the rest of Rheum

sampled, while the accessions of R. alexandrae were polyphyletic

and recovered in two different clades that contain non-glasshouse

species (Fig. 1b). R. alexandrae is widely distributed in the

Hengduan mountain region, and only a single accession has been

included in previous molecular phylogenies30,31. Given the

remarkable morphological divergence (Fig. 1a: B, C), and the

polyphylogeny of R. alexandrae revealed here, a third, previously

unrecognized origin of glasshouse syndrome is plausible. Further

analysis including more populations and closely related species

is needed to clarify the circumscription of R. alexandrae.In

addition, the densitree of Rheum generated from 187 single-copy

genes showed reticulated relationships among some lineages

(Supplementary Fig. 1). Nonetheless, at least three independent

origins of the glasshouse syndrome in Rheum are inferred based on

the sampling and phylogeny presented here.

The noble rhubarb glasshouse is highly effective in blocking

UV but transmits visible and infrared light. The glasshouse

structure is composed of specialized cauline leaves which are

termed as bracts (used to represent the yellowish glasshouse leaves

hereafter) (Fig. 2a). By directly measuring the reﬂectance and

transmittance spectra of fresh tissues in the wild using a spectro-

photometer equipped with an integrating sphere (Supplementary

Fig. 2), we demonstrated that the semi-translucent bracts of noble

rhubarb can block >95% UV radiation (250–400 nm) while

transmitting 60–80% visible and infrared light (>400 nm) (Fig. 2b,

Supplementary Figs. 3, 4). In contrast, the green leaves effectively

block both UV light and most visible light (Fig. 2b, Supplementary

Fig. 4). T-test show that the light transmission rate in visible range

(400–750 nm) is signiﬁcantly higher in bracts than in leaves

(p< 0.01). Our results revealed that the bract is more effective as a

light ﬁlter than previously assumed8,10, probably because we used

the fresh tissue for measurements and we used scattered light that

more closely resembles natural conditions. However, our limited

sampling limits any insight into variation among individual plants.

Further measurements in population level are needed to explicitly

characterize the natural variation in glasshouse morphology. In

addition, both bracts and leaves reﬂected a negligible proportion of

UV radiation (Supplementary Fig. 4), indicating that the UV light

was largely absorbed by the tissue rather than being reﬂected. UV

radiation is intense at high elevations and can cause pollen mal-

formation and reproductive sterility32. It is plausible that this

spectral characteristic of the bracts may generate a favorable

microclimate for the reproductive organs to develop.

In addition to UV radiation, we further evaluated this

microclimate hypothesis by measuring changes of temperature in

the ambient environment versus inside the leafy glasshouse at a

48 h interval during the growth season in the ﬁeld. During the

daytime (08:00–20:00), the internal temperature was constantly

higher than the external temperature and reached as high as 25 °C

around noon, while the ambient temperature was lower than 15 °C

at the same time (Fig. 2c). An optimal ﬂower temperature has been

shown to be crucial for plant reproduction33, as temperature

mediates ﬂower development, pollen viability and pollen tube

growth, and inﬂuences pollinating insect activity. Our previous

control experiments found that the reproductive ﬁtness of noble

rhubarb signiﬁcantly decreased when the bracts were removed13.

These observations indicate that the glasshouse morphology is an

important adaptive trait for alpine environments.

It has been widely acknowledged that ﬂavonoids can function

as photoprotection molecules for land plants because of their

high efﬁciency in scavenging UV-induced reactive oxygen

species. To explore how ﬂavonoids function in noble rhubarb

in resistance to UV radiation, we examined ﬂavonoid content

in ﬁve different organs (bract, leaf, stem, fruit and root).

Using high performance liquid chromatography (HPLC), we

isolated ﬁve UV-absorbing compounds from noble rhubarb

organs (Fig. 2d, Supplementary Fig. 5). On the basis of the mass

spectra and co-chromatography with an authentic standard, the

compound was identiﬁed as quercetin 3-O-rutinoside (C1),

quercetin 3-O-glucoside (C2), quercetin-3-O-galactoside (C3),

quercetin 3-O-arabinopyranoside (C4,) and quercetin 3-O-[6´´-

(3-hydroxy-3-methylglutaroyl)-glucoside] (C5) (Supplementary

Figs. 6–10). All these UV-absorbing substances were quercetin-

based glycosides, a subclade of ﬂavonols that have been

considered as photoprotective compounds for plants because

of their UV-absorbing characteristics34,35. It is notable that, in

addition to bracts, the leaves also accumulate large amounts

of ﬂavonols, while the other organs (stem, fruit and root)

contain only traces of them (Fig. 2d, e). Leaves and bracts

contain comparative levels of ﬂavonoids as a whole, while bracts

are specialized by losing quercetin 3-O-glucoside and quercetin

3-O-arabinopyranoside (Fig. 2d, e).

Following a phylogenetic approach, we identiﬁed putative

noble rhubarb genes that encode the enzymes in the ﬂavonoid

biosynthesis pathway and characterized their expression proﬁle

using transcriptome data. Notably, chalcone synthase (CHS),

which encodes the enzyme that directs phenylpropanoid meta-

bolic ﬂux to the ﬂavonoid pathway, is expanded to 5 members

(Supplementary Fig. 11). Some of these CHS copies display tissue-

speciﬁc expression patterns. Consistent with our observation that

quercetin-derived ﬂavonols are the major UV-absorbing sub-

stances accumulated in bracts and leaves, at least one copy of

the key gene ﬂavonol synthase (FLS, Rn_tig2787.230) is highly

expressed in bracts and leaves but is scarcely expressed elsewhere

(Supplementary Fig. 11). In addition, the ﬂavone synthase gene

(FNS, Rn_tig2803.307) which competes with ﬂavanone-3-

hydroxylase (F3H) for the substrate to produce ﬂavones, is not

expressed in mature plants (Supplementary Fig. 11).

Sequencing, assembly and annotation of the noble rhubarb

genome. We generated 116.35 Gb PacBio long reads (78× coverage)

and 155.4 Gb (105× coverage) cleaned Illumina short reads for

noble rhubarb (Supplementary Data 2). Long reads were used for de

novo assembly, which was corrected and polished with short reads

(the full assembly pipeline is shown in Supplementary Fig. 12),

resulting in a 1.36 Gb draft genome sequence with a contig N50 of

9.8 Mb (Fig. 3a). The draft genome sequence comprises 245 contigs

and accounts for 94% of the genome size of about 1.48Gb as esti-

mated by k-mer distribution with high frequency k-mer (>1k)

excluded (Supplementary Fig. 13). BUSCO assessment recovered

94.5% complete genes from the assembly (BUSCO v5.1.2 with

eudicotyledons_odb10 database, Supplementary Fig. 14), and the

Merqury36 kmer plot (Supplementary Fig. 15) shows a main single

peak which corresponds to a haploid assembly. These results attest

to the completeness of the noble rhubarb genome sequence

(Table 1, Supplementary Data 2). Using a combination of ab initio

and hint-based methods (Supplementary Fig. 16), 58,950 high

conﬁdence gene models were predicted.

COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-05044-1 ARTICLE

COMMUNICATIONS BIOLOGY | (2023) 6:706 | https://doi.org/10.1038/s42003-023-05044-1 | www.nature.com/commsbio 3

Content courtesy of Springer Nature, terms of use apply. Rights reserved

Ks plots of homologous colinear gene pairs revealed a main peak

at 0.73 in R. nobile (Fig. 3b, Supplementary Fig. 17a). To

characterize this potential genome polyploidization event, we

further performed Ks calculations for Fagopyrum tataricum

paralogous gene pairs and R. nobile-F. tataricum orthologous gene

pairs. A similar Ks peak was also observed for F. tataricum paralogs

and the mean Ks value for the R. nobile-F. tataricum orthologs

is 0.48 (Fig. 3b). The divergence time for R. nobile-F. tataricum is

around 45 mya37, thusthe age of the polyploidization (Ks =0.73) is

68.43 mya (T =Ks/2μ). Therefore, the polyploidy event is shared

by both R. nobile and F. tataricum. F. tataricum and R. nobile are

both from the buckwheat family Polygonaceae, which have an

ancient genome polyploidization based on both Ks and gene tree

approaches (Supplementary Fig. 17b). We then examined the

syntenic depth of genic blocks between R. nobile/F. tataricum and

beet (Beta vulgaris) that lacks genome polyploidization after the

core-eudicots shared gamma event38. The syntenic pattern 1-1

(20%), 1-2 (23%), 1-3 (20%) or 1-4 (12%) was observed in

B.vulgaris-R.nobile comparison, similar to the ratio in B.vulgaris-F.

tataricum comparison (Supplementary Fig. 18). Thus, WGT and

24:00 06:00 12:00 18:00 24:00 06:00 12:00 18:00 24:00

Temperture (oC)

Time

glasshouse

ambient

6.98 1.86

59.90

68.31 70.85 72.49

8.40

5.19 4.56

19.94

6.03

52.59

250

Transmittance (%)

Wavelength (nm)

bract

leaf

350 450*** 550*** 650*** 750***

12 14 16 18 20 22 24 26 28 30 32

Bract

Leaf

Stem

Fruit

Root

Retention Time (min)

Absorbance at 350 nm

s1 2 3 4 5

Flavonol glycosides

Content (mg/g)

1 2 3 4 5 1 2 3 4 5 1 2 3 4 5

bract

leaf

stem

1 cm

Fig. 2 The translucent leaves in the noble rhubarb function as glasshouse. a A mature noble rhubarb plant in the Hengduan Mountains (N28.53°,

E99.95°; Elev. 4500 m) with insert showing a morphological comparison between a photosynthetic basal leaf and a translucent leaf. bSpectral

characteristics of bracts and leaves in light transmittance, showing the different patterns of light transmittance by bracts and leaves. Signiﬁcance was

statistically tested at six wavelengths (250–750 nm) using Ttest (n=3 biologically independent samples; ***p< 0.001). cTemperature change inside and

outside of the inﬂorescence glasshouse in a 48 h interval at mid-August in the eastern Himalayas (N29.40°, E94.90°; Elev. 4200 m). dThe HPLC

spectrum of UV-absorbing substances extracted from ﬁve tissue types; s: the ﬂavonol standard isovitexin; 1–5: the compounds isolated from tissues. eThe

contents of the ﬁve ﬂavonol species isolated from noble rhubarb tissues. Fruits and roots only contained traces of the ﬂavonols and are not shown. In

boxplots, the center, lower, and upper lines depict the median, 25th, and 75th percentile respectively, and whiskers represent maxima and minima.

ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-05044-1

4COMMUNICATIONS BIOLOGY | (2023) 6:706 | https://doi.org/10.1038/s42003-023-05044-1 | www.nature.com/commsbio

Content courtesy of Springer Nature, terms of use apply. Rights reserved

WGD cannot be distinguished based on this data. Nonetheless,

these data indicated that both R. nobile and F. tataricum genomes

have both undergone substantial genome fragmentation after the

early Polygonaceae polyploidization event. This is further sup-

ported by the overall genome synteny between B. vulgaris, R. nobile

and F. tataricum (Supplementary Fig. 19).

In context of the evolution of the Caryophyllales (Supplementary

Data 3), we examined gene family expansion/contraction in noble

rhubarb. We identiﬁed 244 gene families that are signiﬁcantly

expanded in noble rhubarb (Fig. 3c). We then performed Gene

Ontology (GO) enrichment analysis to explore functions associated

with these expanded gene families. The expanded gene families are

enriched signiﬁcantly in 46 GO terms including response to cold,

DNA repair, and regulation of protein stability (Fig. 3d). Among

these, several GO terms are of particular interest regarding alpine

adaptation (Supplementary Data 4). The ﬁrst is “response to cold”

Drosera spatulata

Aldrovanda vesiculosa

Fagopyrum tataricum

Rheum nobile

Spinacia oleracea

Chenopodium quinoa

Beta vulgaris

Amaranthus hypochondriacus

Simmondsia chinensis

100 80 60 40 20 0 (Mya)

+72 -5

+50 -37

+90 -160

+244 -24

+49

-89

+391

-18

+49

-89

+44

-48

+42 -14

Terpenoid backbone biosynthesis

Phenylpropanoid biosynthesis

Protein processing in endoplasmic reticulum

alpha−Linolenic acid metabolism

Endocytosis

Ether lipid metabolism

Biosynthesis of amino acids

Arachidonic acid metabolism

Carbon metabolism

Cyanoamino acid metabolism

Linoleic acid metabolism

Cysteine and methionine metabolism

Pentose and glucuronate interconversions

Carbon fixation in photosynthetic organisms

Pentose phosphate pathway

Ribosome biogenesis in eukaryotes

Nucleotide excision repair

DNA replication

Mismatch repair

Homologous recombination

0.0 0.1 0.2 0.3

genes

0.1

0.2

q value

inflorescence

meristem gr

wth

organrganganga ic cyic cycyic cyclic compound

metetabetabolic li pprocepss

nitrogen cn ccn compouuuundnn

metabolic

processs

organic suic sccbstanbbstbs ce

metabbbolic procecccss

riboflaaavin nnn

biosybiosybiosyyynnnthetnthenthenh ic prrocess

egulation of systemi

uired resistanc

lipid

hetheterereeteree ocyclocyclcleeeeee

meetababee bolic ic iprprocepr sssssss

regulation of membrane

lipid distribution

−5

505−

semantic similarity

−50

−40

−30

−20

−10

regul

ation o

stem

eve

opment

xylem and pholem

patteppp rn formation

response to cold

cellullar ar r nir trogen

compococo uuuund mu etabolism

protepppp in stabiliaty

DNA reeeepair

rich factor

R. nobile - R. nobile

F. tartaricum - F. tartaricum

R. nobile - F. tartaricum

polyploidization

speciation

2.00

1.75

1.50

1.25

1.00

0.75

0.50

0.25

0.00

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

The frequency of gene pairs

III

I: gene density

II: LTR-gypsy density

III: LTR-copia density

IV: CpG islands

Fig. 3 Genome architecture and gene family expansion and contraction. a Multi-dimensional display of genomic components of the noble rhubarb

genome. The density was calculated per 100 Kb. bFrequency distributions of synonymous substitution rates (Ks) between homologous gene pairs in

syntenic blocks of R. nobile-R. nobile,F. tataricum-F. tataricum and R. nobile-F. tataricum. Both species are members of the family Polygonaceae. cNumber of

signiﬁcantly expanded (+) and contracted (-) gene families in nine Caryophyllales species with high-quality genome sequences available. dREVIGO

clusters of overrepresented GO terms for signiﬁcantly expanded gene families in the noble rhubarb genome. Each bubble represents a summarized GO

term from the full GO list by reducing functional redundancies, and their closeness on the plot reﬂects their closeness in the GO graph, i.e. the semantic

similarity. eKEGG enrichment of metabolic genes that were signiﬁcantly expanded in the noble rhubarb genome.

COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-05044-1 ARTICLE

COMMUNICATIONS BIOLOGY | (2023) 6:706 | https://doi.org/10.1038/s42003-023-05044-1 | www.nature.com/commsbio 5

Content courtesy of Springer Nature, terms of use apply. Rights reserved

(GO: 0009409, p=4.4e−9) which is self-explanatory. The second

is “regulation of membrane lipid composition”(GO: 0010876,

p=3.6e−15). Other signiﬁcantly overrepresented GO categories

included inﬂorescence meristem growth (GO: 0010081, p=5e−10;

GO: 0099009, p=3.1e−4), regulation of protein stability (GO:

0031647, p=3.6e−2) and response to temperature stimulus (GO:

0009266, p=1.8e−4). Kyoto Encyclopedia of Genes and Genomes

(KEGG39) enrichment analyses revealed signiﬁcant overrepresen-

tation of 18 pathways (Qvalue < 0.20 after multiple test correction,

Fig. 3e). The top four pathways enriched speciﬁcally in noble

rhubarb are all related to DNA repair (Fig. 3e). Other enriched

KEGG terms included linoleic acid metabolism, ether lipid

metabolism, and alpha-linolenic acid metabolism (Fig. 3e), all of

which are related to the biosynthesis of unsaturated fatty acids that

can increase cell membrane ﬂuidity in response to cold stress40.

The noble rhubarb genome is rich in LTRs. A much higher

proportion of repetitive sequences was observed in noble rhubarb

(77.46%) than in F. tataricum (Tartary buckwheat, 50.96%; Sup-

plementary Fig. 20). Long terminal repeats (LTRs) are the most

abundant elements and account for 42.91%of rhubarb genome. To

investigate the evolutionary dynamics of LTRs in rhubarb, we re-

assembled and annotated the genome of Rumex hastatulus41 which

is in the sister genus to Rheum.TheRu. hastatulus genome has also

signiﬁcantly expanded compared to F. tataricum, and repetitive

sequences make up 83.29%. Similarly, LTRs contribute most to the

genome expansion, accounting for 63.05% of Ru. hastatulus gen-

ome. Following the same procedure, 10619 and 18863 intact LTRs

were identiﬁed from R. nobile and Ru. hastatulus, respectively,

while only 775 were detected in Tartary buckwheat (Fig. 4a).

Analyzing the history of LTR insertions revealed distinct LTR

evolutionary dynamics among them. LTRs have proliferated in

R. nobile and Ru. hastatulus (Fig. 4a), however, the proliferations in

R. nobile peaked at ~1-2 Mya, while proliferation in Ru. hastatulus

was dominated by more recent LTR insertions within the past 0.8

Mya (Fig. 4a, b). In addition, Gypsy LTR is the main type in R.

nobile while Copia LTR dominates in Ru. hastatulus (Fig. 4a).

Further inspection of LTR proliferation identiﬁed stronger

association between LTRs and gene coding loci in R. nobile than

in Ru. hastatulus, with 37.8% of the total intact LTRs found

inserted near (<5 kb) the coding sequences in R. nobile, while the

proportion was 19.5% in Ru. hastatulus (Fig. 4c). A closer look at

the insertions indicated that the insertions were enriched at the

ﬂanking sequences of genes (<1 kb) in the R. nobile but not in Ru.

hastatulus (Fig. 4d). The enrichments observed in glasshouse

plants could be the results of insertion preference of TEs, as was

observed in rice42.

To test this hypothesis, we then analyzed the solo LTRs, which

are the remnants of intact LTRs by LTR removal mediated by

interelement recombination. Large amounts of solo LTRs (>100 k)

were identiﬁed in both species (Supplementary Fig. 21), indicating

high frequency and efﬁciency of LTR removal in addition to active

LTR ampliﬁcation. These solo LTRs were distributed evenly on

chromosomes in terms of their association with genes (Supple-

mentary Fig. 22), suggesting that the insertion of LTRs was

random, without preference to genic region in both species.

To test the functional consequences of LTR insertion on genome-

wide gene expression proﬁle, we focused on the gene pairs that are

derived from recent gene duplications (Ks <0.5) (Supplementary

Fig. 16) and analyzed the gene expression divergence and tissue

speciﬁcity. The pairs that had LTR insertions in <1 kb ﬂanking

region had signiﬁcantly higher levels of divergence in expression

(Fig. 4e), although not in tissue speciﬁcity (Fig. 4f).

Transcriptomic proﬁle of transition from green leaf to glass-

house bract in noble rhubarb. A main goal of this research is to

decipher the genetic basis of the glasshouse morphology. We

focus on the differentiation of the glasshouse bracts from normal

leaves using a comparative transcriptomic approach. There are

developmental intermediates between green leaves and yellowish

bracts (Fig. 5a, b), and the glasshouse-like phenotype starts from

where there are ﬂowers and is gradually enhanced upward. We

deﬁned the transition from green leaves to glasshouse bracts into

three zones: zone 1) from the base of the plant to the base of the

raceme, where inﬂorescences start to develop (Fig. 5b, top); zone

2: from the base of the raceme to where the glasshouse bracts

develop; leaves in this zone are green at the base and yellow at the

apex (Fig. 5b, middle); and zone 3: the upper section of the plant,

where bracts are uniformly yellow (Fig. 5b, bottom). Based on our

deﬁnition of the transition, we generated transcriptomes from all

transition points (S0,1,2,3). The overall expression proﬁles of the

transitions were distinct between each pair, with S2 being close to

S3 (Supplementary Fig. 23). Using the basal green leaves (S0) as a

comparison, 1849, 2102, and 2635 signiﬁcantly differentially

expressed genes (DEGs, P< 0.05, |log

(fold change)| > 1) at the

transition point S1, S2 and S3 were identiﬁed, among which 559,

685, and 1045 were upregulated and 1290, 1417, and 1590 were

downregulated (Fig. 5c).

To identify the genetic modules that are associated with the

morphological transition from green leaves to glasshouse bracts,

we performed GO/KEGG enrichment analysis for the DEGs

identiﬁed at transition zones. We ﬁrst explored the up-regulated

DEGs (Fig. 6) and identiﬁed nine signiﬁcantly enriched KEGG

pathways (Fig. 6a). Phenylpropanoid biosynthesis was shared by

all three zones, and starch and sucrose metabolism and stilbenoid

biosynthesis were shared by S2 and S3. The other pathways

largely related to specialized metabolites, such as glucosinolate,

cutin and ﬂavonoids, were unique in S3 (Fig. 6a).

To further examine which biological processes these DEGs

were involved in, we performed GO enrichment analysis. In total,

163 signiﬁcantly overrepresented GO terms (adjusted p≤0.05)

were identiﬁed in glasshouse bracts as a whole (Supplementary

Data 5), and 41 of them were shared by all zones. The

enriched GO terms were further grouped into nine network

modules (node ≥3) (Fig. 6b) by similarity matrices calculated

using the Jaccard correlation coefﬁcient43. The most notable

module was cell wall biogenesis, which included 25 GO terms

Table 1 Statistics for the genome assembly of noble rhubarb.

Assembly

Assembled size 1.36 Gbp

GC content 39%

Number of contig 245

N50 of contig 9.8 Mbp

Longest contig 23.56 Mbp

BUSCOs C:96.1% (S:87.3%, D:7.2%), F:1.6%,

M:3.9%

Merqury QV & Error rate 46.3 & 2.33471e−05

Merqury k-mer completeness 99.24%

LAI (LTR Assembly Index) 9.90

Annotation

Number of protein coding genes 58,950

Number of tRNA 3310

Number of sRNA 4798

Content of repeats 1,002.07 Mbp (73.54%)

-LTR 552.44 Mbp (40.54%)

-DNA transposons 102.44 Mbp (7.52%)

-LINEs 49.85 Mbp (3.66%)

-SINEs 1.53 Mbp (0.11%)

-Unclassised Repeats 279.86 Mbp (20.54%)

ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-05044-1

6COMMUNICATIONS BIOLOGY | (2023) 6:706 | https://doi.org/10.1038/s42003-023-05044-1 | www.nature.com/commsbio

Content courtesy of Springer Nature, terms of use apply. Rights reserved

such as phenylpropanoid metabolic process (GO:0009698), cell

wall macromolecule biosynthetic process (GO:0044038), and cell

wall biogenesis (GO:0042546). Secondary cell wall biogenesis is

the core component of this module as lignin biosynthesis

(GO:0009809), xylan biosynthesis (GO:0045492) and glucuro-

noxylan biosynthesis (GO:0010417) provide polymers for sec-

ondary cell wall deposition. This module was shared by all

glasshouse zones, representing a common feature of glasshouse

bracts as a whole. Nonetheless, as expected from the view of

morphological transition, several modules were speciﬁctoS2

and/or S3, such as the metabolism of aspartate family amino

acids, and amino acids transport (Fig. 6b).

Following the same procedure, we then examined the down-

regulated DEGs (Fig. 6c, d). Overall, 12 KEGG pathways were

enriched in glasshouses bracts, and two of them (photosynthesis and

fatty acid biosynthesis) were shared by all zones(Fig. 6c). In contrast

to the up-regulated DEGs which were mainly enriched in specialized

metabolism pathways, the down-regulated DEGs were mainly

enriched in primary metabolism pathways including fatty acids,

amino acids and propanoate metabolism, to name a few (Fig. 6c).

The suppression of these pathways in glasshouse bracts, together

with S3-speciﬁc pathways, such as carbon ﬁxation/metabolism

and glyoxylate metabolism, were likely to be a consequence of

down-regulation of photosynthesis. Notably, diterpenoid biosynth-

esis, a specialized metabolism pathway, was signiﬁcantly enriched

among down regulated genes in S2 and S3 (Fig. 6c). GO enrichment

analysis identiﬁed 60 GO terms for the down-regulated DEGs

(Supplementary Data 5), and these GO terms were further grouped

into six network modules (node ≥3) (Fig. 6d). Four of them were

unique to S2 and/or S3, and no module as a whole was shared by all

zones (Fig. 6d).

Discussion

Genomic modiﬁcation associated with adaptation to high eleva-

tions has been well documented in animals15 but has been less

explored in plants. The rhubarb genus has experienced a rapid

evolutionary radiation in Himalayan alpine regions over the past

10 Mya30,31, with noble rhubarb inhabiting up to 5000 m in

elevation. The rapid radiation of Rheum may have been accom-

panied by hybridization and gene ﬂow between species, resulting

in reticulated evolutionary relationships as captured in our phy-

logenomic analysis. Our comparative genomic analysis provided

several insights into this alpine adaptation. Firstly, genes involved

in DNA repair, such as homologous recombination and mis-

match repair, signiﬁcantly expanded. As noble rhubarb grows at

Rhfr

Rhal pop 2

Rhof

Rhal pop3

Rhrho

Rhmo

Rhpa

Ruac

Rhfo

Rhta

Rhpu

Oxsp1

Rhno pop2

Rhde

Rhki

Rupa

Rhac

Rhal pop 1

Rhrh

Rhno pop1

Oxsp2

Ruha

50 40 30 20 10 0 (mya)

10.35 (5.18-16.69)

Gypsy

Copia

number of intact LTRs (x102)

LTR insertion time (mya)

percentage (%)

coding region

downstream

upstream

Rhno Ruha

count of insertions (x100)

-5 -4 -3 -2 -1 0 01 2 34 5

gene

Rhno

Ruha

distance (kb)

0.00

0.25

0.50

0.75

expression divergence

tissue-specificity

51234

** ns

Fig. 4 Evolutionary dynamics of LTRs. a Histogram of intact LTR insertions in R. nobile (top), Ru. hastalutus (middle) and F. tataricum (bottom). bTime-

calibrated phylogeny of Rheum and its close relatives Rumex and Oxyria with the proliferation of LTRs labeled with red shapes. cProportion of intact LTR

insertions within the coding region of annotated genes and downstream (5 kb) or upstream (5 kb) of annotated genes in R. nobile (Rhno) and Ru. hastalutus

(Ruha). dLTR insertions were enriched in the ﬂanking regions of coding sequences (1 kb) in Rhno but not in Ruha. e,fViolin plots of the divergences (log

transformed) between gene pairs at expression and tissue speciﬁcity levels (Wilcoxon rank sum test). Boxplots represent the median, 25th, and 75th

percentile, respectively.

COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-05044-1 ARTICLE

COMMUNICATIONS BIOLOGY | (2023) 6:706 | https://doi.org/10.1038/s42003-023-05044-1 | www.nature.com/commsbio 7

Content courtesy of Springer Nature, terms of use apply. Rights reserved

S1/0 S2/0 S3/0 S2/1 S3/1 S3/2

down

counts (x100)

Z2Z1

Fig. 5 Morphological transition. a The leaves were classiﬁed into three zones based on their morphology. bThe phenotype of representative leaves in each

zone. cSigniﬁcantly differentially expressed genes (P< 0.05, |log

(fold-change)| > 1) from different comparisons between leaf types.

r w

r wr w

r r rt

r r rtr r rt

w r

w rw r

r r r rt

r r r rtr r r rt

r r

r rr r

wawa

rrt

rrtrrt

vwa

vwavwa

rrt

rrtrrt

pur r n

pur rn

purr n

pur r n

pur r npur r n

rrt

rrtrrt

rrt

rrtrrt

r r nv v r

r r nv v rr r nv v r

r r

r rr r

photoperiodism

photoperiodismphotoperiodism

positiv ryonic development

positiv ryonic developmentpositiv ryonic development

positive regulation of reproductive process

positive regulation of reproductive processpositive regulation of reproductive process

photosynthesis

photosynthesisphotosynthesis

phytoale

phytoalephytoale

metal ion transport

metal ion transportmetal ion transport

nitrate transport

nitrate transportnitrate transport

cellular response to external stimulus

cellular response to external stimuluscellular response to external stimulus

photosystem II assem ly

photosystem II assem lyphotosystem II assem ly

toto

cell junction assem ly

cell junction assem lycell junction assem ly

response to extracellular stimulus

response to extracellular stimulusresponse to extracellular stimulus

anion transport

anion transportanion transport

response to nitrate

response to nitrateresponse to nitrate

inorganic cation tr rane transport

inorganic cation tr rane transportinorganic cation tr rane transport

rate adhesion

rate adhesionrate adhesion

diter

diterditer

inflorescence development

inflorescence developmentinflorescence development

hor

horhor

cellular response to ethylene stimulus

cellular response to ethylene stimuluscellular response to ethylene stimulus

photosynthesis, light reaction

photosynthesis, light reactionphotosynthesis, light reaction

gener

genergener

electron transport chain

electron transport chainelectron transport chain

photosynthetic electron transport in photosystem I

photosynthetic electron transport in photosystem Iphotosynthetic electron transport in photosystem I

response to chitin

response to chitinresponse to chitin

disacchar

disacchardisacchar

vegetative to reproductive phase transition of meristem

vegetative to reproductive phase transition of meristemvegetative to reproductive phase transition of meristem

NADH dehydrogenase complex assem ly

NADH dehydrogenase complex assem lyNADH dehydrogenase complex assem ly

photosynthesis, light harvesting

photosynthesis, light harvestingphotosynthesis, light harvesting

protein tetramerization

protein tetramerizationprotein tetramerization

shor ay photoperiodism, flowering

shor ay photoperiodism, floweringshor ay photoperiodism, flowering

cellular response to nitrogen levels

cellular response to nitrogen levelscellular response to nitrogen levels

cellular response to nitrogen starvation

cellular response to nitrogen starvationcellular response to nitrogen starvation

oxylic acid tr rane transport

oxylic acid tr rane transportoxylic acid tr rane transport

jasmonic acid mediated signaling pathway

jasmonic acid mediated signaling pathwayjasmonic acid mediated signaling pathway

cellular hor

cellular horcellular hor

cellular response to jasmonic acid stimulus

cellular response to jasmonic acid stimuluscellular response to jasmonic acid stimulus

frfr

cellular response to water deprivation

cellular response to water deprivationcellular response to water deprivation

Linoleic acid metabolism

Tryptophan metabolism

Flavonoid biosynthesis

Cutin, suberine and wax biosynthesis

Glucosinolate biosynthesis

Stilbenoid, diarylheptanoid

and gingerol biosynthesis

Starch and sucrose metabolism

Phenylpropanoid biosynthesis

S1 S2 S3

GeneRatio

0.10

0.03

0.01

p.adjust S1 S2 S3

GeneRatio

0.10

0.04

0.02

p.adjust

Glycine, serine and threonine metabolism

Carbon metabolism

Glyoxylate and dicarboxylate metabolism

Carbon fixation

Tryptophan metabolism

Diterpenoid biosynthesis

Pyruvate metabolism

Propanoate metabolism

Fatty acid metabolism

Fatty acid biosynthesis

Photosynthesis

Fig. 6 Genetic modules underlying the morphological transition of glasshouse leaves. a Enriched KEGG pathways for up-regulated DEGs at three

transition points. bNetwork modules of enriched GO terms for up-regulated DEGs. GO terms are linked by similarity matrices calculated using the Jaccard

correlation coefﬁcient (JC), and each module is labeled with representative terms selected using REViGO (--Medium, --Arabidopsis, --SimRel) (http://

revigo.irb.hr/ accessed at 14-12-2021). cEnriched KEGG pathways for down-regulated DEGs. dNetwork modules of enriched GO terms for down-regulated

DEGs. The modules are identiﬁed in same way as described in b.

ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-05044-1

8COMMUNICATIONS BIOLOGY | (2023) 6:706 | https://doi.org/10.1038/s42003-023-05044-1 | www.nature.com/commsbio

Content courtesy of Springer Nature, terms of use apply. Rights reserved

high elevations, intense UV radiation is a major challenge. UV-B

irradiation can cause direct DNA damage44 and enhanced DNA

repair systems are essential for the survival of noble rhubarb in

the alpine environment. For example, homologous recombination

is widely used by cells to accurately repair harmful breaks that

occur on both strands of DNA45. Secondly, genes involved in

lipid metabolism and regulation, speciﬁcally the unsaturated fatty

acids, signiﬁcantly expanded. As a major component of cell

membranes, lipids have a signiﬁcant role in response to cold

stress, both as a mechanical defense through leaf surface pro-

tection and plasma membrane remodeling, and as signal trans-

duction molecules40. For example, by increasing the level of

unsaturated lipids, the plasma membrane can maintain its ﬂuidity

and stabilization under low temperature, allowing cells to

mechanically adapt to cold46,47.

In addition, LTR proliferation has bloated the noble rhubarb

genome and promoted expression divergences among duplicated

genes, which might contribute to the evolutionary adaptation of

noble rhubarb to alpine environments. LTRs are the most impor-

tant drivers of plant genome evolution48 and their proliferation has

been demonstrated to be associated with evolutionary innovations

in diverse lineages of organisms, including plants49,50 and

animals51. In particular, they can be co-opted to play key orga-

nismal functions by providing genes with promoters and

enhancers52,53, by rewiring regulatory networks54, and by assisting

the evolution of entirely new genes48,55.

Transposable elements have been found to be more active

when organisms are under stress, such as in face of changing

environments56. The uptick of LTR activity in glasshouse plants (1-

2 Mya) might be activated by environmental stress caused by

Tertiary orogeny, as it coincided with the latest phase of the rapid

uplift of the Tibetan Plateau (1.6–3.6 Mya)57. Following the initial

insertion, the high frequency and efﬁciency of LTR removal as

evidenced by the large numbers of solo LTRs has affected the

distribution of LTR. During the removal process, the enrichment of

LTR insertions in regulatory regions (<1 kb ﬂanking) and the

altered gene expression suggest functional role of LTR insertions

and selective retention. These results indicate that the expansions of

LTRs in noble rhubarb genome may have rewired the genome-wide

gene regulation network. The activity of LTR-retrotransposons

fueled by Himalayan geological events may have served as an

engine of evolution for plants to adapt to alpine environments.

The specialization of glasshouse bracts to serve as UV ﬁlters and

warmth-trapping structures may involve complex genetic and

physiological alterations. The bracts of R. nobile have neither

palisade nor spongy parenchyma in their mesophyll but are char-

acterized by highly vacuolated epidermal and hypodermal layers9,

while that of R. alexandrae, another glasshouse plant, have partially

differentiated mesophyll cells with intercellular spaces analogous to

spongy parenchyma10. Both have fewer and malformed chlor-

oplasts compared with normal leaves9,10. In general, the anatomical

structure of the bracts is similar to that of young leaves at a pre-

expansion stage, as undifferentiated mesophyll is a general feature

of immature leaves58.

Our comparative transcriptomic analyses on the differentiation

of glasshouse bracts provided further insights into the develop-

ment of glasshouse morphology. Firstly, photosynthesis was

down regulated in glasshouse bracts indicating the suppression of

photosynthesis in glasshouse bracts. This is evident as glasshouse

bracts are visibly light yellow, likely due to the absence of the

green chlorophyll pigments and fewer and malformed chlor-

oplasts compared to normal leaves10. Secondly, secondary cell

wall biogenesis appears to be upregulated in glasshouse

bracts. Secondary cell walls are typically deposited in specialized

cells, such as tracheary elements, ﬁbers and other scler-

enchymatous cells, and are in general a small component in leaf

tissues59. The function of secondary cell walls and cuticle upre-

gulation may relate to the optical property of the tissue. Thirdly,

terpenoid metabolism, including diterpenoid metabolic process

(GO:0016101) and isoprenoids metabolic process (GO:0006720),

was down regulated in glasshouse bracts, probably because of the

suppression of photosynthesis.

Finally, the bracts accumulate high levels of ﬂavonoids, which

constitute the key feature of its UV ﬁltering function16. Our

metabolite proﬁling showed that normal leaves and glasshouse

bracts accumulate comparable level of ﬂavonoids in the noble

rhubarb, but the latter are differentiated by losing two of the ﬁve

quercetin-based glycosides and by hyper-accumulating hyperin

(quercetin-3-O-galactoside). Flavonoids are well known as pho-

toprotection molecules because of their high efﬁciency in

scavenging UV-induced reactive oxygen species35 and have been

proposed as one of the key metabolic innovations promoting

plant terrestrialization60. Flavonoid metabolism is thus an adap-

tive trait for land plants61, and is particularly essential for alpine

plants exposed to excess UV radiation. In addition, non-leaf tis-

sues in noble rhubarb contain only traces of ﬂavonoids, sug-

gesting tissue-speciﬁc genetic regulation. The presence of

glasshouse bracts equipped with UV-absorbing ﬂavonoids and

low in chlorophyll that covers the inﬂuence thus provides several

advantages: 1) they selectively block short-wavelength light to

protect reproductive organs from UV damage; 2) they effectively

transmit infrared light which has high thermal efﬁciency, thereby

trapping warmth and heating the ﬂowers; and given the beneﬁts

mentioned above, 3) they create an optimal thermal microclimate

for reproduction, and promote pollination by attracting pollina-

tors, as has been reported in R. alexandrae12.

Glasshouse morphology is a charismatic plant adaptation in the

alpine biome. Although the basic morphological and ecological

features of the syndrome have been described9,10,12,13,30, to what

extent these phenotypes are derived through similar genetic

mechanisms are not known. We provided a phylogenetic frame-

work to describe the multiple origins of glasshouse morphology in

Rheum. The reported unique genomic architecture of R. nobile will

facilitate further genomic comparisons between glasshouse species

and non-glasshouse species. The transcriptomic atlas described

here in context of the morphological transition of glasshouse leaves

will enable additional comparative analyses to identify the genetic

basis of the convergent morphogenesis that has given rise to

independent origins of the glasshouse syndrome.

Methods

Phylogenetic placement of glasshouse trait. To track the origin and evolution of

glasshouse morphology in Rheum L., we sampled 17 accessions representing four of

the six sections of Rheum7, together with ﬁve species from Polygonaceae as out-

groups to build phylogenetic tree. Because of the extensive morphological diver-

gences found in Rheum, multiple accessions for the glasshouse plants R. nobile and

R. alexandrae were included in the analysis. Previous phylogenetic inferences30,62

with general gene markers, such as ITS and ETS, showed limited power to

reconstruct the evolution history of Rheum, and we thus employed the phyloge-

nomic approach63 to infer the phylogeny of Rheum. In total, 22 transcriptomes

either from public repositories or generated in this study (Supplementary Data 1)

were used in the phylogenomic analysis. Fresh plant tissues were collected in the

ﬁeld, immediately frozen in liquid nitrogen, and were stored in −80 °C until RNA

extraction. The voucher specimen was deposited in the Herbarium of Kunming

Institute of Botany (KUN). Total RNA was isolated from fresh tissues using the

PureLink Plant RNA reagent (Life Technologies) and further puriﬁed using TRIzol

reagent (Invitrogen). Quality and quantity were examined using a Bioanalyzer 2100

(Agilent Technolo-gies, CA). cDNA libraries with insert sizes 300 bp were con-

structed using the TruSeq Kit (Il-lumina) and then sequenced as 2 × 150bp reads

on the Illumina HiSeq 1500 platform (Illumina Inc., CA, USA) at BGI (Shenzhen,

China). The information of raw reads acquired for each sample is in Supple-

mentary Data 1. Multiple accessions for the glasshouse plant species R. nobile and

R. alexandrae were included. In addition, Fagopyrum tataricum which is in the

same family as Rheum was used as outgroup.

RNA-seq reads, either generated in this study or obtained from public

repository (Supplementary Data 1), were ﬁrst subjected to quality control using

COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-023-05044-1 ARTICLE

COMMUNICATIONS BIOLOGY | (2023) 6:706 | https://doi.org/10.1038/s42003-023-05044-1 | www.nature.com/commsbio 9

Content courtesy of Springer Nature, terms of use apply. Rights reserved

FastQC v0.11.8 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and

trimming using Trimmomatic v0.39 (HEADCROP:10 LEADING:3 TRAILING:3

SLIDINGWINDOW:4:15 MINLEN:36)64. The clean reads were assembled de novo

using Trinity v2.3.265 with default settings. CDS sequences were identiﬁed and

translated using TransDecoder v5.5.066 with homology based ORF (open reading

frame) retention criteria (--retain_pfam_hits, -retain_blastp_hits). The homology

information was obtained via BLASTp against the plant protein database

alluniRefprexp07041667 and pfam searches with Pfam 33.168. All translated amino

acid sequences were then reduced using CD-HIT-V4.6.1 (-c 0.99 -n 5 -M 60000)69.

These assembled transcriptomes, together with the CDS sequences from the

annotated genome of F. tataricum70 were used for phylogenomic reconstruction.

Orthology clustering was performed using Orthograph v0.771 with the high-

quality orthologous gene clusters of noncore Caryophyllales72 as reference. Clusters

with at least 70% species occupancy were further used for gene tree inference. Each

cluster was aligned using MAFFT v7.40773 with default settings, and the alignment

was trimmed with Trimal v1.2 (-gt 0.5 -st 0.1)74. Phylogenetic trees were estimated

using FastTree v2.1.775. Long branches [absolute length (LABS) > 2 or relative length

(LREL) > 10)] that were likely introduced by transcriptome assembly artifacts and/or

distantly related homologs63 were removed. The cleaned clusters were realigned and

trimmed following the same procedure. Alignments with species occupancy > 80%

and alignment length > 300 were used for further analysis, resulting in 2554 clean

orthologous groups of which 187 groups contain only single-copy gene. Phylogenies

were reinferred using RAxML-NG v0.7.0b (--model GTR +G --tree_pars 10)76. The

whole process was pipelined using python scripts adapted from63. In addition, we

tested the robustness of the phylogeny using different subsets of genes. First, subset of

single copy orthologs (one-to-one orthologs) from the 2554 orthologous groups were

chosen, which resulted in 187 genes; second, further ﬁltering by average bootstrap

values for each gene tree (bs > 90) which resulted in 132 genes; third, based on the

ﬁltered gene set, both concatenated- and coalescent-based approach were used to

infer the ﬁnal species tree. Species tree was then constructed using ASTRAL-Pro

v5.7.777 based on the single-copy 132 gene trees. Uncertainty for the species tree was

estimated using local posterior probability (localPP)78. Densitree of Rheum was

generated from 187 single-copy genes.

To infer the species divergence time, we used the Bayesian clock dating79. The

topology of the ASTRAL tree and the concatenation of the 132 single-copy gene

sequences were used. Two fossils were employed as minimum-age calibrations,

including Polygonocarpum johnsonii (>66 mya) which calibrates the crown node of

Polygonaceae80,81 and Aldrovanda intermediata (>41.2 mya) which calibrates the

crown node of Aldrovanda +Dionaea80,82. The calculations were performed using

MCMCTree module (clock =2sampfreq =1000 nsample =50000000)

as implemented in PAML v4.9e83.

Ecological experiments and spectroscopic characterization. To determine the

effect of the bracts on the light spectrum reaching the reproductive organs, we

measured the reﬂectance and transmittance spectrum of bracts and leaves using a

spectrophotometer. We followed previous studies84,85 to set up the equipments

(Supplementary Fig. 2). Both transmittance and reﬂectance spectra were measured

using an IdeaOptics PG2000-Pro spectrophotometer (200–1100 nm; IdeaOptics,

Shanghai, China) equipped with DH-2000 deuterium-halogen lamp (IdeaOptics).

This light source integrates the properties of deuterium lamps and halogen lamps

and emits stable UV light (180–400 nm) and visible light (400–750 nm),

which coveres the effective light radiation reaching the plants in the wild. The

lamps were warmed up for 30 min to obtain stable light emission before measuring.

The reﬂectance spectrum was calibrated with respect to a white Lambertian

reﬂectance standard (STD-WS, IdeaOptics).

Three mature plants with fully developed bracts were randomly selected for

measuring in the ﬁeld (N29.40°-E94.90°, Alt. 4200 m). Bracts from the upper,

middle and lower part of each plant were measured separately (Supplementary

Figs. 3, 4). For transmittance measurements, the sample was illuminated from

outside the sphere directly at an area with 1 mm diameter using optical ﬁber, and

the scattered light was uniformed and detected as described above. For reﬂectance

measurements, the light from the light source was coupled into a 600 μm core-size

optical ﬁber, collimated and illuminated the sample at an 8° angle from within the

IS30 integrated sphere (IdeaOptics). The scattered light from the sample was then

uniformly diffused within the integrated sphere and collected in another 600 μm

core-size optical ﬁber coupled to the spectrophotometer in 90° angle

(Supplementary Fig. 2, inset). To account for the different patterns of light

transmittance observed in bracts and leaves, we performed statistical tests. The

transmission rate at six wavelengths (250, 350, 450, 550, 650, 750 nm) which cover

the UV-visible range were compared using Ttest with three biological replicates.

The ambient temperature and temperature inside the glasshouse-morphology

(ca. 100 cm above the ground) were synchronously recorded using a two-channel

thermocouple data logger (HOBO-U23-003, Onset, USA) equipped with two alloy

needle-type sensor probes (Onset, USA). Temperatures were recorded every

2 minutes in a 48-hour interval at the mid-August in east Himalayas (N29.40°-

E94.90°, Alt. 4200).

Flavonoids extraction and quantiﬁcation. To proﬁle the ﬂavonoid content in

R. noble, the ﬂavonoids from different organs (bracts, leaves, stems, fruits, roots

and seeds) were extracted using Methanol-Formic Acid solution (75%, 0.1%, both

v/v). For each tissue type, three biological replicates, each represented by three

technical replicates from the same individual, were used in the measurements. Full

steps for the sample preparation are available in protocols.io (https://protocols.io/

view/characterization-of-ﬂavonoids-bkekktcw). Apigenin, Quercetin (Aladdin,

China) and Isovitexin (Macklin, China) were used as reference standards and were

dissolved in 75% MeOH with 0.1% formic acid. The crude extracts were ﬁltered by

0.22 μm PVDF membrane and 10 μL was sampled and separated on an Athena C18

HPLC column (120 Å, 4.6 ×150 mm, 3μm; Anpel Technologies, Shanghai, China)

connected to the Access Max HPLC system (Thermo Fisher Scientiﬁc). Mobile

phases of 0.5% (v/v) formic acid in deionized water (solvent A) and 0.5% formic

acid in acetonitrile (solvent B) were used at a ﬂow rate of 600 μL/min. The gradient

program started at 10% B which was increased linearly to 20% in 6 min and to 30%

in 20 min; then eluent B was held constant for three minutes at 90% and linearly

decreased to 10% in two minutes and was held constant for one minute. Finally, the

column was equilibrated for 10 minutes at the initial solvent composition.

Mass spectrometry was performed using a 6546 Q-TOF LC-MS-MS system

from Agilent Technologies (Santa Clara, CA, USA) coupled with an electrospray

ionization (ESI) interface. The parameters were optimized as follows: ESI voltage

−4,000 V, nebulizer gas 60, auxiliary gas 50, curtain gas 35, turbo gas temperature

500 °C, declustering potential −60 V, and focusing potential −350 V. The samples

were analyzed with an information-dependent acquisition (IDA) method, which

can automatically select candidate ions for the MS-MS analysis. The TOF mass

range was set from m/z 50 to 800, and the mass range for product ion scan was m/z

50–800. The collision energy was set to 10 eV to observe the pseudo-molecular [M-

H]-ion and the losses of substituent groups, and 20–40 eV to obtain informati on

about the basic skeletons. The mass analyzer was calibrated using Taurocholic acid

(2 ng/μL) by direct injection at a ﬂow rate of 5 μL/min. Under the negative ESI

mode, molecules with phenol hydroxyl could produce strong and stable [M-H]-,

which could be helpful for identiﬁcation86. The ESI-MSnspectrum for each isolated

compound (Supplementary Figs. 5–10) was searched against the NIST/EPA/NIH

mass spectral library as supplemented in Xcalibur v4.3 (Thermo Scientiﬁc, MA,

USA), and was also compared with that of reference compounds generated under

the same LC-MS conditions. The most abundant ﬂavonoid molecule (peak 3,

Supplementary Fig. 5) produced a deprotonated ion at m/z477 in the ESI-MS1

spectrum (Supplementary Fig. 6). In the ESI-MS2spectrum, two high intensity

fragments at m/z301 [M-H-162]–and 300 [M-H-162] –• were observed. Ions at

m/z301 suggested the loss of a hexose unit. Ions at m

The genome of the glasshouse plant noble rhubarb (Rheum nobile) provides a window into alpine adaptation

Abstract and Figures