A Developmental Framework for Graft Formation and Vascular Reconnection in Arabidopsis thaliana

Summary

Background

Plant grafting is a biologically important phenomenon involving the physical joining of two plants to generate a chimeric organism. It is widely practiced in horticulture and used in science to study the long distance movement of molecules. Despite its widespread use, the mechanism of graft formation and vascular reconnection is not well understood. Here, we study the dynamics and mechanisms of vascular regeneration in Arabidopsis thaliana during graft formation when the vascular strands are severed and reconnected.

Results

We demonstrate a temporal separation between tissue attachment, phloem connection, root growth and xylem connection. By analysing cell division patterns and hormone responses at the graft junction, we found that tissues initially show an asymmetry in cell division, cell differentiation and gene expression, and through contact with the opposing tissue, lose this asymmetry and reform the vascular connection. In addition, we identified genes involved in vascular reconnection at the graft junction, and demonstrate that these auxin response genes are required below the graft junction.

Conclusions

We propose an inter-tissue communication process that occurs at the graft junction and promotes vascular connection by tissue-specific auxin responses involving ABERRANT LATERAL ROOT FORMATION 4 (ALF4). Our study has implications for phenomena where forming vascular connections are important including graft formation, parasitic plant infection and wound healing.

Introduction

The majority of plants possess the ability to adhere tissues and reconnect their vasculature after severing of the vascular strands by wounding. The ability to heal the vascular tissue is particularly important, as this tissue transports water, nutrients and signalling molecules throughout the plant [1]. It is also horticulturally relevant, as plant grafting involves the severing and rejoining of vascular strands from different plant species or varieties to introduce resistance to abiotic and biotic stresses, to propagate plants, or to change plant size [2]. Parasitic plants also in a sense “graft”, as they join elements of their vasculature to the host vasculature upon infection [3]. Despite this biological and horticultural relevance, the mechanism of vascular tissue regeneration remains poorly understood. An efficient method to graft the model plant Arabidopsis thaliana has been developed [4] and used to study the long-distance transport of molecules including proteins, hormones and small RNAs [4–6]. Previous work describes a hierarchy of events that occurs at the graft junction of Arabidopsis and other plants. After cutting, ruptured cells collapse to form a necrotic layer at the graft junction and cells from opposing tissues, termed the scion and rootstock, adhere to each other. Through cell proliferation above and below the graft junction, a mass of pluripotent cells, termed callus is formed. Lastly, it is thought that these callus cells differentiate into vascular tissue to reconnect the phloem and xylem across the graft junction [7–9]. This step is important for graft formation, as tissue attachment and callus formation occur in the absence of compatible grafting, but only in compatible grafts does the vasculature connect [9, 10].

A common theme to plant wound responses is the involvement of plant hormones, which are critical regulators of growth and development. Injury by wounding prompts organs to divide and differentiate, though this phenomenon may be restricted to a specific subset of cells [11]. Part of the wound response is mediated by the plant hormone cytokinin and by the WOUND INDUCED DEDIFFERENTIATION 1 (WIND1) pathway [12]. WIND1 is strongly upregulated upon wounding, and overexpression of this gene results in excess callus formation [12]. In cut Arabidopsis inflorescence stems, the plant hormone auxin promotes the division of pith cells and wound-healing [13]. Ethylene and jasmonic acid are also involved in the wound-healing response and promote the expression of the RAP2.6L and ANAC071 transcription factors around a cut site [13].

In particular, the plant hormone auxin plays a pivotal role in vascular development [14, 15]. Classical experiments demonstrated that the patterns of auxin flow through a tissue determines the sites of vein formation [16]. Similarly, when auxin is added to callus, it promotes the formation of xylem and phloem [17]. In Arabidopsis, normal vein patterning depends on polar auxin transport and can be modified by auxin transport inhibitors such as 1-N-naphthylphthalamic acid (NPA) or mutations in genes coding for auxin transport proteins [15, 18, 19]. Auxin signals by binding the TRANSPORT INHIBITOR RESPONSE 1/AUXIN SIGNALING F-BOX (TIR1/AFB) receptors that target the AUXIN/INDOLE-3-ACETIC ACID (Aux/IAA) proteins for degradation [20, 21]. These proteins are negative regulators of many Auxin Response Factors (ARFs), so that the presence of auxin ultimately promotes ARF activity and therefore auxin-regulated transcription [22]. Many mutant Arabidopsis lines that block auxin signalling have been identified. The AUXIN-RESISTANT 1 (AXR1) gene is required for normal TIR1 function, and when mutated, changes the stabilisation dynamics of the AUX/IAA proteins [23]. Mutations in the TIR1/AFB binding DII domain of AUX/IAAs renders these proteins insensitive to auxin, and can therefore keep ARFs and auxin signalling repressed. Mutants in several auxin signalling genes including MONOPTEROS (MP/ARF5), BODENLOS (BDL/IAA12) and CULLIN 1 (CUL1/AXR6) perturb vascular patterning [24–27]. Mutating ARF6 and ARF8 reduces cell division in the pith cells upon cutting [28]. Alternatively, increasing auxin in plants by exogenous applications promotes the formation of callus from xylem pole pericycle cells, the cells that give rise to lateral roots [29, 30]. One gene required for lateral root and callus formation in plant tissue culture is ALF4 (AT5G11030) [30–32]. ALF4 mutants are resistant to auxin [31], and ALF4 is hypothesised to act downstream of auxin to maintain the xylem pole pericycle cells in a mitotically competent state [32]. The ALF4 gene is expressed throughout the plant and the protein is nuclear localised but contains no similarities to proteins in other families, and its precise role is unknown [31, 32].

The research reported here characterises vascular reconnection during graft formation in Arabidopsis thaliana. We use a well-established hypocotyl grafting method to describe the dynamics of tissue adhesion, phloem connection, xylem formation and root growth. We analyse cell division and gene expression changes at the graft junction, identifying a process whereby contact with the opposing tissue reduces gene expression asymmetries between the rootstock and scion that arise at the graft junction. Furthermore, we identify several genes involved in vascular connection and demonstrate that these genes are required specifically in the rootstock near the graft junction. We propose that these genes act as part of a mechanism that senses the opposing tissue by perceiving transported auxin, thereby promoting wound healing and vascular formation.

Results

Tissue attachment, phloem reconnection, root growth and xylem reconnection are temporally separated

Hallmarks of successful graft formation are the attachment of scion to rootstock (Figure S1), and the resumption of root growth. We used previously described Arabidopsis grafting protocols [4, 33] and found that scion and rootstock attached within two days after grafting (DAG), as assayed by adherence of scion and rootstock when plants were lifted (Materials and Methods, Figure 1B). Grafting initially arrested root growth, but five DAG, the majority of roots resumed growth (Figure 1B). We then asked whether the vasculature connected prior to or after the resumption of root growth. Previous analyses have used the water-soluble dye carboxyfluorescein diacetate (CFDA) to monitor movement through the phloem and xylem [34, 35]. CFDA is a non-fluorescent compound until taken up by a cell whereupon the acetate group is cleaved off, creating a fluorescent molecule. We reasoned that we could use CFDA to monitor phloem and xylem connectivity at the graft junction. CFDA application on the cotyledons would be expected to allow detection of phloem connectivity, as phloem transport can occur from shoot to root, whereas application to the roots would be expected to allow detection of xylem connectivity, since xylem transports from root to shoot (Figure 1A)[1]. Consistent with these expectations, we observed fluorescence in the hypocotyl phloem poles after CFDA was applied to the cotyledons, and observed fluorescence in the tissues surrounding the hypocotyl xylem after CFDA was applied to the roots of ungrafted plants (Figure S1). Upon treatment of the scion with CFDA, we observed fluorescence in the rootstock of grafted individuals three DAG, and by four days, nearly all individuals fluoresced (Figure 1C). CFDA treatment to the rootstock in grafted individuals produced fluorescence in the scions six DAG, and by seven days, the scions from the majority of individuals fluoresced (Figure 1D, Figure S1). As a second test of vascular reconnection, we grafted scions carrying the pSUC2::GFP transgene [36], which express free GFP in the phloem companion cells, to non-transgenic rootstocks (Figure S1), an assay previously used to monitor phloem connectivity [37]. We monitored individuals daily, and observed that three DAG the root vasculature showed GFP fluorescence similar to ungrafted pSUC2::GFP plants (Figure 1C, Figure S1), consistent with previously published reports [37]. We also exposed grafted individuals to a low humidity environment to test water transport across the graft junction. Newly grafted individuals wilted under these conditions. However, when transferred to low humidity seven DAG, a substantial number of individuals could take up sufficient water from the medium to remain turgid (Figure 1E). Our results are thus consistent with a scenario in which attachment between the rootstock and scion occurs first, followed by phloem regeneration at about three days, root growth at approximately five days, and xylem regeneration at around seven DAG.

Figure 1. Phloem connection, root growth and xylem connection are temporally separated.

A) Cartoons showing the transport assays used. pSUC2::GFP Col-0 scions were grafted to Col-0 rootstocks, and fluorescence monitored in the roots. Alternatively, CFDA was applied to scion or rootstock, and fluorescence monitored in the vascular tissue of the rootstock or scion.

B) Plant attachment precedes root growth in grafted Col/Col plants (scion/rootstock notation) 0 to 10 DAG. Mean shown from 5 experiments with 20–36 plants per time point per experiment (+/− SEM).

C) Phloem reconnection occurs 3–4 DAG, as monitored by fluorescence appearance in the rootstock after CFDA application to the cotyledons of Col/Col plants. Alternatively, fluorescence was monitored in Col rootstocks grafted to pSUC2::GFP scions. Mean shown from 3–5 experiments with 11–24 plants per time point per experiment (+/− SEM).

D) Xylem reconnection occurs 6–7 DAG, as monitored by fluorescence appearance in the scions after CFDA application to the rootstocks of Col/Col or ungrafted plants. Mean shown from 3 experiments with 11–37 plants per time point per experiment (+/− SEM).

E) Hydraulic connectivity is restored 7–8 DAG, as monitored by placing Col/Col or ungrafted plants in a low humidity environment and monitoring scion wilting after 24 hours. Mean shown from 3 experiments with 11–37 plants per time point per experiment (+/− SEM).

Contact between scion and rootstock resolves a tissue asymmetry in cell differentiation, cell division and gene expression

To characterise further vascular formation and cell differentiation at the graft junction, we visualised xylem by clearing the tissue using a previously described method [38]. New xylem vessels formed 4–5 DAG above the graft junction, whereas xylem formed below the graft junction 5–6 DAG (Figure 2A,B). Since xylem is composed of dead cells, it appears that xylem precursor cells differentiated and underwent programmed cell death to form new xylem vessels. New xylem first differentiated above the graft junction and one to two days later differentiated below the graft junction to reconnect the xylem (Figure 2A,B). Even in grafts where the old xylem vessels appeared aligned, the new xylem often took an indirect route to connect below the cut site (Figure 2A), possibly avoiding damaged xylem elements. Xylem formation appeared driven by an apically-derived process, consistent with our observation that xylem formation occurred in cut shoots but not cut roots (Figure 2C, Figure S2). To test whether a similar phenomenon occurred with the phloem, we grafted pSUC2::GFP scions to two segments of Col-0 to form a three segment or interstock graft (Figure S1). The top junction connected before the lower junction (Figure 2D), consistent with a process that begins in apical tissues also driving phloem reconnection.

Figure 2. Cells differentiate and expand upon graft formation.

A–C) New xylem vessels (spiral structures denoted by white triangles) form above and below the graft junction (A,B) and in cut shoots (C) as observed with DIC optics of cleared whole mount hypocotyls. For (B), n=10–27 plants per time point. AR – adventitious root.

D) In interstock grafts, the phloem reconnects in the top segment first, evident by the presence of GFP signal from the pSUC2::GFP scion in the middle section at 3 DAG. At 4 DAG, the signal continues to the rootstock. Insert shows the fluorescence at the root tip.

E) Vascular tissue expands across the graft junction, as observed in a longitudinal-section though a pUBQ10::PM-tdTomato scion grafted to a p35S::YFP-ER rootstock. Inserts above and to the left show orthogonal views of the centre panel (dotted lines). Arrow denotes expanded tissue. Bottom left number represents the number of individuals that showed vascular invasion.

F) Epidermal cells expand at both the scion (p35S::GFP-LTi, p35S:: H2B-RFP) and rootstock (p35S::mCherry-LTi, p35S::H2B-YFP) halves of the graft junction. White triangles denote cell expansion 0 to 3 DAG.

A–F) Scale bar is 50μm.

To understand better cell growth during vascular connection, we grafted Arabidopsis expressing different fluorescent reporters regulated by the constitutive promoters CaMV 35S and Ubiquitin10 [39, 40] on a microscope coverslip (Figure S1) and monitored fluorescence over 7 days. Contact between the scion and rootstock promoted vascular formation, but suppressed the cell expansion and cell division that occurred at the cut surface of ungrafted shoots (Figure S2). In some individuals, vascular cells expanded and proliferated across the graft junction into the adjoining tissue (Figure 2E, Figure S2). Within 24 hours of grafting, epidermal cells above and below the graft junction expanded to fill the graft junction and the gaps formed due to cutting and apparent cell lysis (Figure 2F). This epidermal expansion continued for three DAG (Figure 2F). To assay cell division, we monitored the endodermis and observed cell divisions occurred above the graft junction three DAG (Figure 3A), which correlated with the expression of a marker involved in Casparian strip formation in the endodermis, CASP1 [41] and later, the presence of lignin, consistent with the reformation of the Casparian strip (Figure 3B, Figure S3). We observed pCASP1::NLS-GFP expression at the graft junction four DAG, and this expression was substantially higher than in cut and ungrafted controls (Figure 3C, Figure S3). CASP1 expression appeared below the graft junction one to two days later than above the graft junction (Figure 3D, Figure S3), indicating an asymmetry in cell differentiation at the graft junction between the scion and rootstock. To assess cell division patterns during graft formation in more detail, we performed in situ hybridisations to detect mRNA expression of Histone H4, a marker of S phase, on plants two or three DAG. We observed a higher level of Histone H4 expression in the vascular tissue of scions than in rootstocks two DAG (Figure 3E,F). At three DAG, Histone H4 expression was present in the rootstock vascular tissue at levels similar to those in the scion (Figure 3E,F). Histone H4 expression was detected close to the graft junction and occasionally was present in the outer cell layers where new vasculature may be forming (Figure 3F), but not in the hypocotyl of uncut control individuals (Figure S3). Image analysis of the Histone H4 in situ hybridisations in the vascular region revealed a statistically significant difference of the mitotic index between the rootstock and the scion two DAG, but no difference was found three DAG (Figure 3E).

Figure 3. Cell proliferation and gene expression dynamics at the graft junction.

A) Transverse-sections 50μm above or 50μm below the graft junction reveal that endodermal cells (white triangles) divide above but not below the graft junction at 3 DAG in pUBQ10::PM-tdTomato expressing plants.

B) Autofluorescence of lignin at the graft junction suggests that the lignin-containing Casparian strip network is reformed across the graft junction (dashed line) in many plants at 7 DAG (left panel), but not all (right panel). Bottom left number represents the number of grafts observed with this phenotype.

C) A marker of Casparian strip formation, pCASP1::NLS-GFP, is upregulated around the cut site (dashed lines) in grafted plants, but not in ungrafted plants.

D) pCASP1::NLS-GFP plants were grafted to themselves or to wild-type plants, and fluorescence intensity of the GFP signal was quantified daily for the same plants. The number of plants above the threshold was plotted to compare pCASP1 activation in the hypocotyl scion versus activation in the rootstock. n = 14–21 per treatment, and representative individuals are shown in Figure 3C and Figure S3.

E) Mitotic index in the vascular tissue of plants 2 DAG (n = 10) and 3 DAG (n = 15). Whereas a significant difference in the mitotic index was observed 2 DAG between the scion and rootstock (p<0.001; Welch Two Sample t-test), no difference was found 3 DAG.

F) In situ hybridisation showing Histone H4 expression in two grafted plants at 2 DAG and two grafted plants at 3 DAG. Cell divisions (arrowheads) occurred in the vascular tissue close the graft junction (dashed line).

G) The wound-responsive marker pWIND1::GFP is upregulated above and below the graft junction upon grafting, and diminishes by 10 DAG. Scale bars 50μm (A–G).

To see if markers associated with wounding would be upregulated during grafting, we tested the wound-induced gene, WIND1 [12], and found that pWIND1::GFP was highly upregulated three DAG in the vasculature and epidermis of the scion’s hypocotyl, but not in the rootstock (Figure 3G, Figure S3). Expression of pWIND1::GFP was at a similar level below and above the graft junction six DAG and decreased by ten DAG (Figure 3G). Cut but ungrafted shoots showed a similar wound-response in the scion, but we did not observe a WIND1 response in cut rootstocks (Figure S3), indicating the response in grafted rootstocks was promoted by the presence of the scion. Ungrafted controls did not show an upregulation of WIND1 (Figure S3). These results point towards coordinated gene expression, cell division and expansion driven by contact between the scion and rootstock.

Grafting activates auxin and cytokinin responses in the vascular cambium and pericycle

Auxin and cytokinin are two key hormones implicated in vascular differentiation [42, 43], so we sought to understand their roles in vascular reconnection. We used the auxin-responsive promoter DR5 [44], and as a positive control, observed activation at the graft junction upon exogenous auxin treatment after six hours (Figure S5). We did not observe a strong increase in auxin response one or two DAG (Figure S5), but at three DAG, there was a response (Figure S4). The delayed hormone response may be due to sensitivity of the reporters, or it may be a response driven by a wound and vascular connection pathway rather than the expected accumulation of auxin at the cut surface due to basipetal transport. The pDR5rev::GFP-ER [45] response peaked above and below the graft junction at five days (Figure 4A, S4) and diminished by 10 days (Figure 4A, S4). To identify which cells respond, we hand sectioned above and below the graft junction (Figure 4B), and observed auxin response in the pericycle cells on both sides of the graft junction (Figure 4C). Notably, the response below the graft junction was strongest in the pericycle cells adjacent to the xylem (the xylem pole pericycle cells) and was often asymmetric (Figure 4C). Ungrafted controls did not show this strong response (Figure S4). Cut but ungrafted roots had no response, and cut shoots had a strong auxin response throughout the tissue (Figure S4). Auxin response in the rootstock was specific to grafting, and could result from auxin transported from the scion.

Figure 4. Auxin and cytokinin response are spatially and temporally controlled during vasculature connection.

A–D) The auxin-responsive pDR5::GFP-ER gene increases expression at the graft junction upon graft formation and diminishes by 10 DAG (A,B). In transverse-sections 50μm above the graft junction (C), pDR5::GFP-ER expresses highly in the pericycle, whereas 50μm below the graft junction (D), pDR5::GFP-ER expresses highly in the xylem pole pericycle cells. Hypocotyls express pUBQ10::PM-tdTomato (C–D) to outline cell membranes.

E–H) The cytokinin-response pARR5::GFP gene increases expression at the graft junction upon graft formation and diminishes by 13 DAG (E,F). In transverse-sections 50μm above and below the graft junction (G,H), pARR5::GFP expresses highly in the pericycle and cambial cells. Hypocotyls express pUBQ10::PM-tdTomato (G,H) or are counterstained with FM4-64 (F).

I–K) The cytokinin responsive pTCSn::GFP-ER gene increase expression upon graft formation both above and below the graft junction (I). In transverse-sections 50μm above and below the graft junction (J,K), expression is high in the pericycle and vascular cambium. Hypocotyls express pUBQ10::PM-tdTomato.

A–K) Dashed lines represent where the hypocotyl was cut, and solid lines where transverse-sections were made. Asterisks indicate pericycle cells, and arrows point to the orientation of the xylem. Scale bar is 25μm.

We also tested the cytokinin-responsive promoters ARR5 and TCSn [46–48], and as a control, observed activation at the graft junction upon exogenous cytokinin treatment after six hours (Figure S4). In grafted individuals, pARR5::GFP showed a response four DAG and peaked at six DAG (Figure 4E–F, S4), whereas pTCSn::GFP-ER showed a response at five DAG and peaked at seven DAG (Figure 4I, S4). With both reporters, the response occurred above and below the junction (Figure 4F, 4I), and was present in the pericycle cells and vascular cambium (Figure 4G–H, 4J–K). Ungrafted plants had a slight response in the phloem poles, and cut shoots showed no response (Figure S4). Cut roots showed strong pARR5::GFP and pTCSn::GFP expression throughout the tissue (Figure S4). Cytokinin response in the scion was specific to grafting, and could result from cytokinin transported from the rootstock.

A subset of auxin response genes are important for phloem connection

Auxin is a well-known inducer of vascular differentiation [16, 17]. To test whether auxin levels or auxin signalling are involved in vasculature reconnection, we took rootstocks with mutations or transgenes that perturb auxin levels, auxin signalling or auxin response and grafted these to wild-type scions expressing pSUC2::GFP. We reasoned that if auxin or auxin response were required for phloem connection, it would alter the ability of plants to reconnect and alter the time after grafting when GFP fluorescence occurs in the rootstock. Of the 32 genotypes analysed, four strongly affected phloem reconnection: alf4, axr1, iaa18 and the tir1 afb2 afb3 triple mutant (Table 1). iaa28 had an intermediate effect with a 46% reconnection rate at four DAG. These mutations in the rootstock did not abolish graft formation, but instead delayed phloem connection by approximately two fold and diminished the amount of GFP in the root tips compared to wild-type rootstocks (Figure 5A, Figure S5). This effect was not due to a weak or malformed root, as interstock grafts with a wild-type scion, wild-type rootstock and 1 mm segment of alf4 or axr1 hypocotyl also delayed phloem reconnection (Figure 5B). This result suggests that ALF4 and AXR1 are required locally below the graft junction to promote phloem reconnection. Previous reports found axr1 to have subtle differences in mature vasculature [49]. We did not detect any vascular abnormalities in young alf4 hypocotyls (Figure S5). Ungrafted axr1 and alf4 plants were not impaired in CFDA transport (Figure 5C, data not shown), and alf4 and axr1 had rates of attachment after grafting similar to those of wild-type plants (Figure S5). To test whether these genes could also be involved in xylem connection, we performed CFDA xylem and wilting assays on alf4 grafts. The onset of CFDA transport from rootstock to scion was extended in alf4 compared to wild-type plants (8 days versus 6 days), whereas the wilting assay gave similar values for alf4 and wild-type plants (Figure 5C, Figure S5). It appears that alf4 inhibits xylem connection, but not as strongly as it affects phloem connection.

Table 1. Genotypes tested in the rootstock for phloem connection.

pSUC2::GFP Col-0 scions were grafted to rootstocks of varying genotypes. Percent (%) fluorescence in the rootstocks of 20–28 plants measured 4 DAG. Asterisks indicate that the average was calculated from at least two biological replicates.

	Root genotype	% Fluor		Root genotype	% Fluor		Root genotype	% Fluor
	Col-0*	90		Col-0*	90		C24	75
Auxin	alf4-1*	24	Auxin	tir1-1 afb2-3	75	Cytokinin	ipt-161	54
	alf4-063	14		tir1 afb2 afb3	29
	arf2-8	83		tir1 afb1 afb2/+ afb3	71		Col-0*	90
	arf6-2	92		tir3-101	79	Cytokinin	ahk2, ahk3	71
	arf8-3	88		p35S::YUC1	81		ahp6-3	92
	arf8 35S::ARF 17	98		p35S::iaaL	96		amp1-1	96
	arf7-1 arf19-1*	92					arr1,12	96
	aux1-7	88		Ler*	77		ckx3, ckx5	88
	axr1-12*	12	Auxin	iaa3 (shy2-2)	67		cre1, ahk3	83
	axr4-2	83		iaa3 (shy2- 31)	100		ipt1,3,5,7*	83
	iaa1 (axr5-1)	71		iaa18-1*	6		p35S::CKX1	96
	iaa7 (axr2-1)	100					p35S::CKX3	96
	iaa12 (bdl-2)	58		Ws*	92		p35S::LOG7	88
	iaa14 (slr1)	79	Auxin	iaa28-1*	46
	iaa17 (axr3-1)	67					Col-0*	90
	iaa18 (crane-2)	13		Col-0*	90	Other	scr-3	83
	iaa19 (msg2-1)	83	Ethylene	ctr1-1	54		shr-2	80
	pin3,4	88		ein2-1	88		WIND1-SRDX	88
	pin4,7	96		eto2	83		Capsella rubella	63
	pin3,4,7	88		etr1-1	83		Cardamine hirsuta	70
	tir1-1	88					Olimarabidopsis pumila	96
	tir1-1 afb2-5	91					Thellungiella salsuginea	75

Figure 5. Auxin response genes, including ALF4, are involved specifically in the hypocotyl rootstock for vascular connection.

A) Mutations in auxin response genes delay the movement of GFP from wild-type pSUC2::GFP scions into grafted rootstocks. Genotype of the rootstock indicated on the panel. n=24 plants per graft combination.

B) Interstock grafts with pSUC2::GFP scions delay GFP movement to the rootstock by one day compared to two segment grafts. A 1mm segment of alf4-1 or axr1-12 mutant tissue in the hypocotyl is sufficient to further delay phloem connection between Col scions and Col rootstocks. Scion/hypocotyl/rootstock genotype indicated. n=24 plants per graft combination.

C) alf4-1 affects the transport of CFDA from the rootstock to the scion. Mean shown from 3 experiments with 15–25 plants per time point per experiment (+/− SEM).

D) Root growth in Col rootstocks is not affected by the presence of an axr1-12 or alf4-1 scion. Mean shown from 2–8 experiments with 20–25 plants per graft combination per experiment (+/− SEM).

E–F) ALF4 and AXR1 are required in the rootstock for efficient movement of CFDA across the graft junction, whereas ALF4 and AXR1 are not required in the scion. A mutant axr1-12 scion can partially rescue the requirement for AXR1 in the root (F). Mean shown from 24 plants per graft combination per time point (E) or from 2 experiments with 20–25 plants per graft combination per time point (F).

There is accumulating evidence that cytokinin also plays an important role in vascular differentiation [43]. Thus, we tested mutants or transgenic plants perturbed in cytokinin levels or signalling to test whether cytokinin response contributes to vasculature reconnection. Of the 11 genotypes affected in cytokinin levels or signalling, only the ipt-161 genotype that overexpresses a bacterial cytokinin-biosynthesis gene (ipt) that causes excess cytokinin production had an effect, which may have been partially due to the C24 background that does not graft as well as Col-0 (Table 1). Strong cytokinin signalling mutants such as wol and arr1, arr10, arr12 triple mutants (arr1,10,12) lack phloem in the primary root [43], so we performed interstock grafts with a segment of mutant hypocotyl between a wild-type scion and rootstock. wol and arr1,10,12 did not substantially delay phloem connection (Figure S5). Furthermore, we tested arr1,10,12 and a genotype that causes excess cytokinin production, p35S::LOG7, in the scion and saw no effect on reconnection (Figure S5), suggesting that these cytokinin response mutants or transgenic plants do not affect phloem formation across the graft junction. Finally, we tested the role of ethylene in vascular reconnection as ethylene signalling is important for vascular cell division and wound healing [13, 50]. We observed no changes in phloem connection rates in mutants deficient in ethylene response, but the ctr1 mutant, enhanced in ethylene signalling, had an intermediate delay in phloem reconnection (Table 1).

ALF4 and AXR1 act below the graft junction to promote graft formation

Our results indicate that grafting initially produces an asymmetry in gene expression and cell division. We sought to test whether this was also the true for genes involved in vascular connection. We grafted alf4 or axr1 scions to wild-type rootstocks – the opposite order to the previous experiments – and the root growth observed was similar to self-grafted wild-type plants (Figure 5D). We also measured CFDA movement through the phloem and found that alf4 or axr1 scions grafted to wild-type rootstocks had normal CFDA transport dynamics (Figure 5E,F), demonstrating that ALF4 and AXR1 are specifically required in the rootstock and not in the scion for normal phloem reconnection after grafting. Surprisingly, when axr1 rootstocks were grafted to axr1 scions, grafts behaved similarly to wild-type controls (Figure 5F), indicating that an axr1 scion can rescue an axr1 rootstock. To understand further how alf4 perturbs graft formation, we observed xylem differentiation at the graft junction in mutant plants. alf4 grafts had an increase in xylem formation above the graft junction seven DAG compared to wild-type plants, whereas xylem formation below the graft junction was delayed compared to wild-type plants (Figure 6A,B). These observations suggested a block at the graft junction of an apically-derived signal such as auxin, so we tested auxin response in alf4 rootstocks. The auxin responsive pDR5rev::GFP-ER reporter showed a slight decrease in alf4 scions, but was strongly reduced in the vasculature and pericycle of alf4 rootstocks (Figure 6C–E, Figure S6) consistent with an absence of auxin response in this tissue. Together, these results indicate that ALF4 is required in a tissue-specific manner to initiate an auxin response in the rootstock.

Figure 6. ALF4 is involved in auxin-dependent vasculature connection, and a model and time course of graft formation in the Arabidopsis hypocotyl.

A) New xylem vessels (spiral structures denoted by white triangles) form above and below the graft junction in wild-type grafts, but formation is enhanced above the graft junction and delayed below the graft junction in alf4/alf4 grafts. Cleared whole mount hypocotyls shown.

B) New xylem vessel formation measured in a population of cleared whole mount hypocotyls at various time points after grafting. n=13–27 plants per time point.

C–D) ALF4 is required in the rootstock to activate vascular auxin response. Col-0 (C) or alf4 scions (D) in the pDR5::GFP-ER background were grafted to pUBQ10::PM-tdTomato rootstocks, or grafted reciprocally. Scale bar is 25μm.

E) Vascular and epidermal auxin response is decreased 20% in alf4-1 compared to Col when scions are grafted to wild-type (WT) rootstocks, whereas it is decreased 49% in alf4-1 compared to Col when rootstocks are grafted to WT scions (*p<0.01; student’s two-sample equal variance t-Test, with a two-tailed distribution). n = 18–20 plants per treatment (+/− SEM).

F) A time course of hormone response, cell differentiation and physiological processes associated with hypocotyl graft formation in Arabidopsis seedlings.

G) A hypothetical model of graft formation: Auxin produced in the scion moves across the graft junction and triggers an auxin response that involves the auxin receptors (TIR/AFBs) and AXR1. These activate an AUX/IAA-ARF response that may involve IAA18 targets to reconnect the phloem. The precise function of ALF4 is unknown, but this protein may assist in the transport and perception of auxin. Notably, AXR1 and ALF4 are not required in the scion, indicating the scion has a different mechanism for phloem reconnection.

Discussion

Grafting is of horticultural and scientific importance, and has garnered recent interest with the identification of chloroplast and nuclear genomes apparently moving across the graft junction [51, 52]. We present a series of tools and assays to measure graft success and genetically dissect the mechanism of vascular reconnection (Figure 6F). We find that phloem connects across the graft junction before the xylem, which is consistent with phloem forming before xylem in the primary vascular development of young organs [42]. Our time course analysis is consistent with the finding that phloem connects at three DAG in young Arabidopsis hypocotyls [37], though Yin et al (2012) also describe rootstock-to-scion dye movement and xylem alignment at three DAG. These results can be reconciled by the proposal that grafted plants with their xylem vessels aligned transported dye from rootstock to scion up to three DAG (24% of plants in our assay at three DAG; Figure 1D) but these plants wilted in a low humidity environment, indicating the xylem connection was not comparable to wild-type plants. After three days, transport ceased in most individuals (Figure 1D), possibly due to callus formation blocking xylem alignment (Figure S6). By seven DAG, new xylem vessels restored connectivity in the majority of plants (Figure 1D–E, Figure 2A). Thus, we believe that functional xylem reconnection requires new xylem formation and occurs after phloem reconnection.

Arabidopsis grafting is also a useful tool to study cell differentiation, since phloem is formed when procambium cells differentiate and enucleate to form phloem sieve elements [53]. Similarly, xylem is formed when procambium cells differentiate and undergo programmed cell death to become xylem vessels [54]. Our data are consistent with procambium cells or undifferentiated callus dividing and differentiating to phloem or xylem vessels. The time differences in connectivity may be due to phloem and xylem differentiation being initiated at different times, or simply that xylem differentiation requires more time due to the increased number of steps such as secondary wall thickening and cell death [54]. Phloem sieve elements have plasmodesmatal connections with each other [53] and since functional phloem formed by three DAG, we conclude that plasmodesmata formed across the graft junction by this time, consistent with previous findings that plasmodesmata form across a graft junction [55].

Our genetic analysis found no strong requirements for genes involved in cytokinin response or ethylene response (Table 1). Mutants with high cytokinin levels (ipt-161) or elevated ethylene signalling (ctr1) affected phloem formation, but not as strongly as the auxin signalling mutants. This is different from wounding in mature inflorescence stems, where ethylene response is required for healing and pith division after cutting [13]. It could be that ethylene response is specifically required in non-vascular cells, or that different tissues have different hormone requirements for healing. Distant relatives to Arabidopsis such as Thellungiella salsuginea efficiently grafted to Arabidopsis scions, and perturbing the Arabidopsis cortex (scr-3 mutant), endodermis (shr-2 mutant) or wound response (chimeric repressor pWIND1::WIND1-SRDX) in the rootstock did not affect phloem connection in Arabidopsis grafts (Table 1). A subset of auxin signalling mutants, including alf4, axr1, iaa18 and the tir1 afb2 afb3 triple mutant, affected vascular reconnection, though none completely abolished it. These data suggest that cytokinins and ethylene make only a minor contribution to phloem reconnection, while auxin is the primary driver of vasculature reconnection, and that phloem regeneration is an extremely robust process.

Our analyses demonstrate that the rootstock and scion show an asymmetry since after cutting, wound response, cell division, xylem differentiation and CASP1 expression were activated above the graft junction. These observations are consistent with previous transcriptional analyses using inflorescence stems that demonstrate differential expression of RAP2.6L and ANAC071 transcription factors, ARFs, and cell wall formation genes above compared to below a cut site [13, 28, 56]. However, the responses we observed lose asymmetry over time so that after several days, endodermal specification, cell division and wound response were similar on both sides of the graft junction. We propose that an apically-derived substance, most likely auxin and/or sucrose, drives reconnection and promotes endodermal differentiation, xylem differentiation, cell division and wound response in the rootstock. Although we did not observe WIND1 response in cut but ungrafted rootstocks (Figure S2), a previous publication describes strong WIND1 upregulation in excised root hypocotyls placed on sucrose-containing media [12], suggesting that nutrient transport to the rootstock may be an important factor in wound healing and graft formation.

Auxin is important for the formation of vasculature and the differentiation of xylem and phloem in plants [15, 42, 53]. Our results are consistent with these observations as both xylem and phloem formation are apparently driven by an apically-derived substance, most likely auxin and/or sucrose. Surprisingly, altering levels of endogenous or exogenous auxin by grafting with p35S::iaaL, p35S::YUC1 or by grafting on synthetic auxin-containing media (NAA) did not substantially affect phloem reconnection (Table 1; Figure S5). We propose that increasing or decreasing auxin amounts is not rate-limiting for phloem reconnection in young Arabidopsis hypocotyls. Instead, auxin response plays a more important role, and only when the response is decreased may increasing auxin levels overcome this deficiency. This hypothesis is consistent with our observation that mutations in the auxin receptors and genes associated with auxin response perturbed vascular reconnection. In particular, grafting an axr1 scion rescued the phloem connectivity defect in an axr1 rootstock but did not improve phloem reconnection to a wild-type rootstock (Figure 5F). axr1 mutants have elevated levels of auxin [57, 58] which could explain this rescue, although we were unable to rescue the axr1 rootstock phenotype by increasing endogenous or exogenous auxin levels (Figure S5), possibly because of incorrect auxin levels or because auxin was not present in sufficient amounts in the vasculature.

In addition to an asymmetry in gene expression and cell division, our study demonstrates an asymmetry in the genetic requirements for grafting since AXR1 and ALF4 are only important below the graft junction for phloem reconnection. We hypothesise that the alf4 rootstock has problems importing and/or perceiving auxin since a block in auxin transport at the graft junction could cause auxin accumulation in the scion, increasing xylem differentiation and it could delay phloem reconnection and growth of the rootstock vascular bundle (Figure S6). Grafting on media containing auxin transport inhibitors (NPA) or grafting with mutants of the auxin efflux transporters (PIN proteins) that are normally expressed in the hypocotyl did not affect phloem reconnection [59] (Table 1, Figure S5). NPA treatment does not reduce vascular strand formation [18, 19] and there is functional redundancy among auxin transporters including the PIN proteins [60], so these results do not rule out the importance of auxin transport. In a previous study, alf4 did not affect the pDR5::GFP-mediated auxin response but instead acted downstream to block pericycle cell divisions [61]. We see a strong decrease in pDR5::GFP vascular signal in the alf4 rootstock, suggesting that this tissue has an absence of auxin and/or auxin response. Thus, we propose that ALF4 acts in the rootstock to promote transport and perception of scion-derived auxin, which drives vascular formation from the rootstock to the scion. In the absence of ALF4, vascular reconnection occurs, albeit more slowly than in wild type. This delay could represent an alternative reconnection pathway or one that is rootstock-independent. ALF4 is required for xylem pole pericycle cell division [31, 32] and these cells show a strong ALF4-dependent auxin response, indicating that the xylem pole pericycle cells may be key to driving vascular reconnection.

The dominant gain-of-function mutations iaa18 and iaa28 also affected graft formation (Table 1). IAA18 is expressed in the phloem and xylem whereas IAA28 is expressed in the xylem [62], consistent with the hypothesis that altering auxin response in a subset of vascular cells may be key to promoting or blocking vascular reconnection and graft formation. Although alf4 is perturbed in tissue culture-mediated callus formation [30], we found no evidence that it is perturbed in wound-induced callus formation (Figure S6). Instead, alf4 grafts produced large amounts of callus at the graft junction that was not found in wild-type grafts (Figure S6). This callus could result from a perturbed auxin transport and auxin response that we believe is related to the delay in reconnection. Together, our data indicate that the rootstock is not a passive partner in graft formation and it uses a different genetic pathway to drive connection to the scion. It reinforces the importance of auxin in vascular formation including a new role for ALF4, and provides information and tools that may help to improve graft formation in the future.

Experimental Procedure

Plant Material

Arabidopsis thaliana wild type, accession Columbia was used throughout, unless otherwise indicated. All the lines used in this study have been previously published, and information on them is present in supplemental experimental procedures. alf4-1 plants were from heterozygous or homozygous parents. Homozygous seeds were obtained by grafting a wild type rootstock to an alf4-1 scion, and self-pollinating by hand. Progeny were verified by primers that identified the alf4-1 deletion (5′-CGATGAATACTTAGCTGTTTCGTG-3′ and 5′-TCCAAAGCTACGTTCTCATACAAA-3′). These plants were phenotypically indistinguishable from alf4-1 mutants derived from a heterozygous parent.

Microscope imaging and image analysis

Fluorescent images of graft junctions were taken on a Zeiss LSM-700 or LSM-780 laser scanning confocal microscope using a 20X water-dipping objective, or a 10X dry objective. Fluorescent images of interstock grafts were taken on a Zeiss V12 dissecting microscope fitted with a Hamamatsu EM-CCD camera. A Zeiss V12 dissecting microscope was used to image CFDA fluorescence, and GFP fluorescence from pSUC2::GFP scions to wild-type roots. Colour images were taken on a Zeiss V20 dissecting microscope. Images were processed using FIJI software. Brightness and image intensity was adjusted for controls and samples equally. For longitudinal images, z-stack projections are shown and made using the average intensity function. For the pDR5::GFP-ER longitudinal images, z-stack projections were made from the stacks containing the vascular tissue.

Hormone response assays

For hormone response assays, Arabidopsis plants were grafted on a coverslip (Figure S1). 1.2μM benzyl adenine (BA) or 10μM 1-naphthaleneacetic acid (NAA) were dissolved in 0.1% dimethyl sulfoxide (DMSO) and added to 1/2 Murashige and Skoog (MS) medium + 1% bacto agar (pH5.7) and left to solidify. 48 hours after grafting, plates were opened and 2mm square blocks of agar containing BA, NAA or DMSO only (no hormone control) were placed next to the graft junction. Plates were sealed, and graft junctions were imaged at 0 and 6 hours after hormone treatment.

For phloem connectivity assays, Arabidopsis plants were germinated on 1/2 MS media and transferred to media containing hormones four days after germination, and three days before grafting. Plants were grafted on damp 3mm Chr Whatman filter paper containing 0.06μM BA, 0.5μM NAA, 5μM NPA or DMSO, and left to recover for ten days on the hormone-containing Whatman paper.

Attachment assays

After grafting, Arabidopsis plants were picked up using forceps at the root/hypocotyl junction. If the scion remained attached during the manipulation, then the graft was considered attached.

Microscope sample preparation for cross-sections and whole mounts

Grafted hypocotyls were placed in a molten 3% agar solution, and left to solidify. A double-sided razor blade was used to cut through the hypocotyl to make transverse or longitudinal hand sections. Sections were placed on a glass coverslip, and 1% molten agar placed around the sample to glue it to the coverslip. For toluidine blue stained sections, a solution of 0.1% toluidine blue (Sigma Aldrich) was used to stain the sample. Water was then added immediately, and the sample imaged using a 20X objective. For visualising xylem and casparian strips, samples were cleared using a previously described protocol [38] and mounted in 50% glycerol for visualisation with a Zeiss Axioimager.M2 microscope with DIC optics (xylem) or a Zeiss 780 confocal microscope with a 488 laser to detect autofluorescence from lignin [63].

Arabidopsis grafting

Ethanol sterilised Arabidopsis seeds were germinated on 1/2 MS + 1% bacto agar (pH5.7; no sucrose) and grown on vertically-mounted Petri dishes under short day conditions (8 hours of 80–100μmoles light) at 20°C. Grafting was performed in a laminar flow hood with a portable dissecting microscope. 7 day-old seedlings were transferred to 9cm Petri dishes that contained one layer of 2.5×4cm sterilised Hybond N membrane (GE Healthcare) on top of two layers of 3mm Chr Whatman paper (8cm sterilised disks; Scientific Laboratory Supplies). The paper was kept moist using sterile distilled water. A transverse cut was made through the hypocotyl close to the shoot using a vascular dissecting knife (Ultra Fine Micro Knife; Fine Science Tools). In addition, one cotyledon was removed to assist in aligning the grafted pieces. In the case of self-grafts, an additional 1mm segment was cut from the hypocotyl and discarded. Grafts were assembled by butt alignment of the two cut halves with no supporting collar. After grafting, Petri dishes were sealed with parafilm, mounted vertically under short day conditions and monitored for 10 days at 20°C.

For grafting on a microscope coverslip, a 10cm square Petri dish was modified by removing a 2.5 × 4cm section of plastic from the back, and gluing a microscope coverslip in its place (see Figure S1). The Petri dish was sterlised with ethanol and a 2.5×4 cm rectangle of Hybond N membrane was placed on top, just overlapping part of the microscope cover slip. Three 3 × 8cm strips of Whatman paper placed at the sides and base of the Petri dish. The Whatman paper and Hybond N were moistened with sterile water. 7 days after germination, Arabidopsis seedlings were placed on the coverslip, so that the roots were on the Hybond N, and hypocotyls were on the coverslip. Grafting proceeded as above, and plates were sealed with Parafilm and graft junctions imaged through the coverslip with a 10X dry objective on a Zeiss 700 or 780 confocal microscope (see Figure S1).

Xylem assays

CFDA assays - Carboxyfluorescein diacetate (CFDA; Cambridge Bioscience) was dissolved in DMSO (1mg per 10μl DMSO), and then 4.6μl added to 1ml of 1/2 MS + Gamborg’s B5 vitamins (Duchefa Biochemie) + 0.8% bacto agar (pH 6.7) to make a final concentration of 1 mM. 500μl of solution was pipetted in a line into an empty 9cm Petri dish, and allowed to solidify. Grafted Arabidopsis or ungrafted controls were cut off 2–3mm below the root/hypocotyl junction and the cut surface was placed in the solidified CFDA-agar solution. 20 minutes after placing the cut Arabidopsis plants in the CFDA-agar solution, fluorescence in the cotyledon vasculature was observed under a Zeiss V12 dissecting scope with a YFP filter and plants scored as either fluorescing or not. 20 minutes was chosen as a time point as it was the time that allowed the vast majority (>95%) of the ungrafted control plants to take up the dye, yet minimize the time which fluorescent dye might pass across the graft junction through methods other than through a continuous vascular connection, such as by apoplastic movement. It also reduced the influence of radial movement of CFDA from the vasculature.

Wilting assays - after imaging the above assay, the Petri dish was left at room temperature with the lid on (but no other form of sealing). Plants were then checked for wilting. Plants were scored as wilted if, after 24 hours of placing in the CFDA-agar solution, they showed strong collapsing of the cotyledon leaf surface. After CFDA application and wilt scoring, grafts were discarded, and new plants used for the next time point.

Phloem assays

CFDA assays - CFDA was dissolved in DMSO (1mg per 10μl DMSO), and then 4.6μl added to 1ml of 1/2 MS + Gamborg’s B5 vitamins + 0.8% bacto agar (pH 6.7) to make a final concentration of 1 mM. Cotyledons of grafted plants or ungrafted controls were lightly damaged by grasping the leaf with narrow forceps. A drop (~1μl) of CFDA solution was pipetted onto the macerated point while the agar was still liquid. After 1 hour, fluorescence in the root vasculature was observed under a Zeiss V12 dissecting microscope with a YFP filter and plants scored as either fluorescing or not. 1 hour was chosen as a time point as it was the time that allowed the vast majority (>95%) of the ungrafted control plants to take up the dye, yet minimize the time which fluorescent dye might pass across the graft junction through methods other than continuous vascular, such as apoplastic movement. It also reduced the influence of radial movement of CFDA from the vasculature. After CFDA application and scoring, grafts were discarded, and new plants used for the next time point.

pSUC2::GFP assays – pSUC2::GFP Col-0 scions were grafted to wild type or mutant roots, and roots were observed with a Zeiss V12 dissecting microscope 0 to 10 DAG with a GFP filter. The number of plants with fluorescent roots was counted daily. The same plants were observed during the course of the 10-day assay.

H4 in situ hybridisation

In situ hybridisation with HISTONE H4 RNA antisense probe was performed on seedlings two and three DAG in accordance with standard protocols [64]. The quantification of the HIS4 expression and the calculation of the mitotic index were done as described [65] with the following modifications: All images were acquired using a Zeiss AxioCamMR3 camera with the Zeiss AxioVision Image software. Images for image analysis were captured with a 20x lens in brightfield with fully opened diaphragm at 1388 × 1040 picture size. The threshold was determined by the mean intensity of unstained cell populations in the scion meristem region or young leaves of the same section minus four standard deviations. These tissues were chosen for background determination because they contained a continuous array of small cells, similar to the vascular region of graft junctions, in contrast to the surrounding large vacuolised cells of the graft. The mitotic index was determined as the ratio between the total number of thresholded pixels to a defined total number of 60.000 pixels, which represents the area of the vascular tissue of the graft analysed for the scion and rootstock. Up to 5 sections per graft showing the vasculature were analysed, and the mean intensity was calculated. The calculation of the mitotic index was done using the ImageJ software and Excel. Statistical analysis was performed in R. Data were tested for normality using a Shapiro-Wilk test and means were compared pairwise using Welch’s t test.