Anatomy of the Tetrastigma host roots

Tetrastigma host roots have the typical anatomy of an eudicot liana root with secondary growth (Fig. 2A; Supplementary Data Fig. S1). The Tetrastigma host root of Rhizanthes lowii has mostly tetrarch primary xylem without pith. The secondary xylem is paratracheal and composed of mostly solitary, wide vessels, closely associated with much narrower vessels and xylem parenchyma (Fig. 2B). Perforations of the wide vessels are usually scalariform. Thin-walled tyloses are present especially towards the centre of the root (Fig. 2A). The radial xylem rays are multiseriate, 4–10 cell layers wide, and are heavily loaded with starch. The xylem rays of the Tetrastigma hosts of Rafflesia and Sapria possess mucilage cells (Fig. 2B). The rays do not always reach the centre of the root but are always dilatated at the phloem. Tangential bands of parenchyma cells traverse the axial vascular system, sometimes connecting neighbouring rays, especially in the Rafflesia host (Fig. 2B). The cambial zone is a well-defined area several cell layers wide. The position of the obliterated primary phloem is identifiable by a band of primary phloem fibres (Supplementary Data Fig. S1A, B). Tangential groups of thick-walled, lignified secondary phloem fibres alternate with unlignified sieve tubes and parenchyma cells in the secondary phloem of the Rafflesia and Sapria hosts (Fig. 2C; Supplementary Data Fig. S1B, C). Occasional islands of sclerenchyma cells are observed in the periderm (Supplementary Data Fig. S1A). The phelloderm and the dilatated phloem rays harbour numerous raphide-containing mucilage idioblasts, tanniniferous cells and druse-containing cells (Supplementary Data Fig. S1D). Rhytidome is present in the host of Rafflesia but it is not apparent in the roots of the Rhizanthes and Sapria hosts, where the phellogen appears to originate from cells in the outer cortex.

Open in a separate window

Fig. 2.

Micrographs of the Rafflesiaceae host infection, stained with ruthenium red and toluidine blue (similarly in all histological sections unless otherwise noted). Endophyte cells are always marked with dotted yellow lines. (A) Transverse section of the root of the Tetrastigma species host of Rhizanthes lowii. (B) Xylem of the Tetrastigma host of Rafflesia cantleyi, depicting vessel dimorphism with large vessels (red asterisks) and small vessels (green asterisks), and parenchyma septa (dotted orange lines) connecting adjacent rays. (C) Secondary phloem of the Tetrastigma host of Rafflesia cantleyi with dilatating parenchyma rays and sclerenchyma septa (dotted orange lines). (D) Rhizanthes lowii endophytes in the host phloem; abundant starch grains are present in the host rays (PAS/toluidine blue contrast staining). (E) Rhizanthes endophyte cells with large nuclei, single nucleoli and proteinaceous inclusions; one of the cells is binucleate (yellow arrow). (F) Rafflesia cantleyi endophyte inside the host xylem; the cell walls of the parasite stain distinctly red. (G) Bright field image of unstained Tetrastigma root with Rhizanthes endophytic strands. (H) Fluorescent micrograph of the same area as (G), treated with aniline blue to stain the callose deposits in the sieve elements of the phloem (arrows); callose deposition is not detected from around the endophyte. (I) Rhizanthes endophyte strand in the tightly packed with starch xylem parenchyma rays of the host (PAS staining). (J) Difference in the nuclear size of an endophyte cell of Rhizanthes lowii (yellow arrow) and the phloem parenchyma cells of its Tetrastigma host (black arrows). (K) Difference in the nuclear size of an endophyte cell of Sapria himalayana (yellow arrow) and cambium cells of its Tetrastigma host (black arrows). (L) Multiseriate endophytic strand of Rhizanthes in the xylem ray of the host. (M) A cluster of endophyte cells of Rhizanthes preceding protocorm formation in the phloem of the host. (N) Rhizanthes endophyte among the small vessels of the host. (O) Rhizanthes endophyte strand at the border of a xylem parenchyma ray and the axial vessel elements. (P) Rhizanthes endophyte traversing the cambium (Ca) of the host. (Q) Rhizanthes endophyte in the phloem of the host. (R) Uninfected vessel elements from the Tetrastigma host of Rhizanthes; compare relative to (N), (O) and (S). (S). Rhizanthes ring-like structures in transverse section, surrounding large vessels of the Tetrastigma host. Scale bars: A (2 mm); B, C, R, S (500 μm); D, F, G, H, I, L, M, N, O, P, Q (100 μm); E, J, K (20 μm).

Cytology of the endophyte

Based on our previous studies of early parasite bud development (Nikolov et al., 2014), we were able to confirm and expand on earlier observations of cell characteristics, which describe the endophyte of Rafflesiaceae as rows of cells with nuclei larger in size in comparison with the nuclei of the host (Schaar, 1898; Brown, 1912). The endophyte cells exhibit thin cell walls and dense cytoplasm, and are quite inconspicuous and well integrated in the root structure of Tetrastigma (Fig. 2D). The endophyte cells are most easily recognized by the large size of their ovoid nuclei, which are comparable among genera and of average diameter 18 μm (s.d. ± 3 μm) for Rhizanthes, 18·5 μm (s.d. ± 2 μm) for Rafflesia and 16·5 μm (s.d. ± 3·5 μm) for Sapria (Fig. 2J, K). These dimensions greatly exceed the size of the host's ellipsoidal nuclei (cell nuclei sizes of all measured Rafflesiaceae hosts are comparable, having an average length of 6·25 ± 0·7 μm, n = 20). Such a difference in size translates into an order of magnitude difference of the volumes of the host and the parasite nuclei. The endophytes grow vegetatively as uniseriate strands interspersed among the host tissue (Fig. 2D, E) and, during transition to flowering – as small clusters or multiseriate strands (Fig. 2P, Q). The endophyte cells towards the periphery of the host xylem tend to be anisotropically elongated but are more isodiametric in the host phloem. The cell walls stain distinctly red with ruthenium red (Fig. 2F). However, we did not detect significant callose depositions by the host around the endophyte cells after aniline blue staining (Fig. 2G, H). The cell walls are uniformly thin and do not exhibit the invaginations or sculptural elaborations increasing the surface area of transfer cells found in other parasites (e.g. Cytinus, De Vega et al., 2007). Endophyte cells have dense cytoplasm, and often one to several small vacuoles, the largest reaching the size of the nucleus. Some cells accumulate large proteinaceous inclusions, but starch accumulation has not been observed (Fig. 2E, I). The nucleus of each endophyte cell is located more or less centrally, in contrast to the parietal nuclei of the host. In addition, binucleate endophyte cells are occasionally present (Fig. 2E). We could not identify morphological characters that distinguish the three genera, which is not surprising given the extreme reduction of the vegetative body.

Distribution of the endophyte within the host

Based on the parasite distribution in consecutive transverse sections, Rafflesiaceae endophytes appear to form uniseriate strands that are oriented in an approximately radial direction within the root (Fig. 2G, H). No differentiation of cells is observed in different regions of the host – the cytology of endophyte cells in a strand is quite uniform throughout. Mitotic figures were not observed in the endophyte. However, the orientations of the cell walls of neighbouring endophyte cells in a single transverse section, and the presence of the same radially oriented strand in only a few adjacent consecutive transverse sections, suggest that they grow primarily by anticlinal divisions (i.e. division plates form perpendicular to the axis of the strand; Fig. 2P, Q). However, periclinal divisions (i.e. division plates forming parallel to the axis of the strand) may also occur to generate the small clumps of parasite cells occasionally observed toward the host bark that may give rise to protocorms and subsequently flowering shoots (Figs 2L, M and ​and4C).4C). Branched strands in transverse sections and independent strands that appear to coalesce into a single strand over a series of adjacent transverse sections were rarely observed. All of these patterns seem to suggest that the body of the parasite is largely fragmented, due either to multiple infections or to subsequent cell separation after infection by a single seed. The endophyte appears equally distributed throughout the xylem (both vessel elements and rays) and phloem, often traversing the cambium (Fig. 2N–Q). The parasite tends to occur at the border of rays and tracheary elements (Fig. 2O). The endophyte appears capable of growing intrusively, as shown by ring-like formations surrounding developed vessel elements in transverse sections, where the endophyte has inserted itself between mature host tissues (Fig. 2R, S). The intercalation of endophyte strands between host sclerenchyma fibres supports this conclusion (Supplementary Data Fig. S1E).

Open in a separate window

Fig. 4.

Development of Rafflesiaceae. All, except (A), depict Rhizanthes lowii. (A) Section of a seed of Rafflesia patma, showing a young sporophyte (proembryo) composed of two tiers of cells above a single cell (modified from Ernst and Schmid, 1913). (B) Tetrastigma host root external appearance and transverse section with a uniseriate filament, the predominant vegetative form. (C) Initiation of protocorm formation by periclinal divisions to give rise to a multiseriate filament. (D) The protocorm is transformed into a cormus by the separation of a sheet of tissue (dotted yellow line) at the front of the growing tear-shaped body. (E) Young floral shoot emerging from the host with completely differentiated floral organs. (F) Anthetic plant. Scale bars: B, C (100 μm); D (500 μm).

Comparison of Rafflesiaceae s.s. with other endophytic angiosperms

In contrast to the largely uniseriate and undifferentiated body of Rafflesiaceae, the endophytic system of the Mediterranean Cytinus (Cytinaceae), which parasitize Cistaceae, forms clumps of cells in the host phloem and towards the periphery of the host xylem. These clumps expand laterally to produce continuous sheaths of parasitic tissue at later stages (De Vega et al., 2007). Secondary growth of the host buries these sheaths inside the host wood and new sheaths are later formed toward the periphery. These successive layers collectively communicate through radially oriented strands that traverse the xylem, called sinkers. The mature endophytes have well-developed vascular systems of both phloem and xylem. Pilostyles (Apodanthaceae), which parasitizes Fabaceae, is similar in developing both cortical strands and radial sinkers, but does not seem to produce uninterrupted sheaths of parasitic tissue. Instead, the cortical strands form an anastomozing cortical complex (Solms-Laubach, 1874b). Three different cell types have been observed in the cortical complex, including putative sieve elements (Kuijt and Olson, 1985; Solms-Laubach, 1874b). Mitrastemon (Mitrastemonaceae), which parasitizes Fagaceae, forms multiseriate, spindle-shaped, vascularized strands in the host cambium and phloem, which anastomose outside of the host cambial layer and occasionally form sinkers into the xylem (Watanabe, 1936, 1937). Finally, the Santalalean endophytes Arceuthobium douglasii Engelm., Tristerix aphyllus Tiegh. ex Barlow & Wiens and Viscum minimum Harv. also form multiseriate strands (and radial sinkers and longitudinal strands in Arceuthobium), which have differentiated phloem and xylem elements (Mauseth, 1990; Lye, 2006; Mauseth and Rezaei, 2013). This is equally true across other parasitic angiosperms that produce endophytic tissue as part of a haustorium, such as parasitic Orobanchaceae (Joel, 2013).

Despite the extreme modifications of these endophytic parasites, all of them possess some histological complexity and are composed of several cell types in their more advanced vegetative stages. This is not the case in Rafflesiaceae s.s., which have only one cell type in the endophyte and are not vascularized but persist as uniseriate filaments throughout their vegetative stage. The endophyte of at least some other parasitic angiosperms is not comparable with that in Rafflesiaceae because it is not proembryonic, but instead originates from a shoot. Therefore, Rafflesiaceae, which produce the world's largest flowers, also produce the most reduced endophyte. The endophytic system of all three genera in the family is very uniform and shows little variation that could be attributed to specific growth conditions. Thus, while it does not appear to be phylogenetically informative in the family, this cellular homogeneity does serve to distinguish Rafflesiaceae further from the other former members of Rafflesiales.

Nuclear size

The nuclear diameter of endophyte cells is comparable with the nuclear diameter of floral meristem cells of Rafflesiaceae (Nikolov et al., 2014). It exceeds the diameter of the nuclei of their hosts by at least 2-fold, making it a convenient landmark to differentiate host from parasite. Presumably, this large nuclear volume translates into a large genome size, but this remains unknown for Rafflesiaceae. The large size of the nuclei of Rafflesiaceae is comparable with that of Lilium longiflorum (Liliaceae), which possesses among the largest angiosperm genomes (approx. 90 Gb) (Price et al., 1973). Interestingly, other endophytic parasites, such as Cytinus hypocistis and V. minimum, also possess nuclei much larger than those of their respective hosts (De Vega et al., 2007; Mauseth and Rezaei, 2013). It is unclear whether this evolutionary convergence toward larger nuclear sizes in endophytes has an adaptive advantage, or if it represents a passive process as a consequence of the relatively unlimited host resources [e.g. the accumulation of repetitive elements or host-to-parasite gene transfer (Davis and Wurdack, 2004; Xi et al., 2013a, b)]. The large genome size may determine slow growth due to the increased time needed for genome replication. Slow growth of the endophyte is also suggested by the apparent absence of mitotic figures in our material.

Topology of the parasite endophyte strands within the host

There is little endophyte tissue relative to the host, i.e. the endophyte cells in one strand are more likely to be in direct contact with host cells than with other endophyte cells. As such, individual endophyte strands could be described as individuals, and their number per unit volume of host is high, especially in Rhizanthes (L. A. Nikolov, pers. obs). One way to explain this pattern is to invoke multiple infections of the host. Field observations note ants and rodents, which consume the fleshy pulp of the mature fruits, as potential agents of seed dispersal (Bouman and Meijer, 1994; Meijer, 1997; Bänziger, 2004; Pelser et al, 2013). It has been documented that the passage through the rodent's digestive tract does little damage to the seeds and may actually be required for germination because their exotegmen is lost in the process (Bänziger, 2004). Given that numerous seeds are ingested together, it is possible that many seeds may infect a single vine in one inoculation event. In this case, neighbouring endophytes might be closely related but not necessarily genetically identical to each other if a single fruit was consumed. Of course, endophytes residing in a single host could be distantly related in the case of multiple independent inoculations from different genotypes. This scenario is unlikely, however, due in part to the general scarcity of ripe fruits because successful fertilization events are apparently rare (Nais, 2001). Alternatively, neighbouring endophytes may be genetically identical clones that result from fragmentation of an initially single strand due to localized disruptions by host cells or localized endophyte cell death. This might represent a kind of asexual reproduction in Rafflesiaceae, but only within a single host because dispersal of the ‘progeny’ is not likely to occur. Population genetic studies will greatly help to clarify these scenarios.

Parasite–host interactions

The parasite cells do not accumulate starch; however, the host rays and phloem parenchyma in closest proximity to most endophytic strands are heavily loaded with starch and probably represent an abundant and important resource for the parasite. We did not detect significant depletion or excess of starch around individual endophytic strands. It appears that the host does not recognize the endophyte as a pathogen when it is present as a uniseriate filament. For instance, the host does not isolate the parasite with callose, which is a common response following pathogen invasion (Nürnberger and Lipka, 2005). This raises a question about the nature of the host–endophyte interaction. Although unable to photosynthesize and obtaining its resources from the host, it is interesting to consider the possibility that the Rafflesiaceae endophyte could be a commensal or mutually beneficial partner to Tetrastigma, for example by providing biochemical versatility or by increasing host immunity (e.g. Bonfante and Genre, 2010). Along these lines, the host appears to react only when the parasite grows significantly to form an incipient shoot. This suggests that there could be a mechanical stimulus to this reaction (and possibly chemical, if the surface properties of the endophyte change during protocorm expansion). The parasite also appears to be capable of modulating its development, as observed in the transformation of the leading apical boundary of the protocorm from semi-circular to crescent-shaped in response to the host phellogen blocking its growth. The functional significance of these processes may be 2-fold. On the one hand, isolating and effectively killing the parasite probably allows for control and reduction of the host investment in the parasite's growth. On the other hand, it may select for parasites that develop quickly before their growth is arrested by cork isolation. Under these circumstances, the proposed heterochronic shifts associated with a prolonged proembryonic stage and the direct advancement to flowering are not surprising. This developmental trajectory might be selected for as a result of the intimate dynamics between the host and the parasite that progressively reduced the vegetative stage of the parasite until it was primarily present as an endophytic strand eliciting no obvious defence response from the host. Thus, the loss of photosynthetic function as a result of parasitism might be only one explanation for the reduced vegetative morphology of Rafflesiaceae. Another intriguing possibility is the adaptive significance of the lack of vegetative elaboration as a way to decrease the duration of host control on the parasite.

We thank two anonymous reviewers for constructive comments. We thank S. Vessabutr and the staff at Queen Sirikit Botanic Garden for allowing access to Sapria, S.-Y. Wong and members of the Wong Lab for logistical support, M. Matthews for help with microtome sectioning, and K. Bomblies, D. Boufford, C. Extavour, W. Friedman, D. Haig, A. Knoll, S. Mathews and members of the Davis, Kramer and Friedman labs for valuable discussions. This research was supported by a National Science Foundation (NSF) AToL grant to C.C.D. (DEB-0622764) and NSF grant D.E.B.-1120243 to C.C.D. and E.M.K.