Mapping by VESGEN of Leaf Venation Patterning in Arabidopsis thaliana with Bioinformatic Dimensions of Gene Expression

One fundamental requirement shared by higher terrestrial plants with humans and vertebrate animals is a complex, fractally branching vascular system (Guyton and Hall, 2006; Roth-Nebelsick et al., 2001). Such vasculatures respond with sensitivity to varying physiological needs by modulating the gene expression patterns of morphologically based vascular functions such as fluid, metabolic, and immune factor exchange. Indeed, the mid-Cretaceous phylogenetic explosion in terrestrial biodiversity has been attributed recently to the increased density and complexity of angiosperm leaf venation responding to elevated levels of CO₂ (Boyce et al., 2009; Brodribb and Feild, 2010; de Boer et al., 2012; Pennisi, 2010). Results of these studies suggest that greater venation within angiosperm leaves optimized the rates of vascular-dependent transpiration and photosynthesis, resulting in the sudden increase of other dependent taxa such as pollinators and early mammals.

The importance of leaf venation in angiosperm evolution and development has inspired decades of research and characterizations of leaf vein architecture, ranging from classic morphological studies to the application of modern molecular tools to understand the processes that contributes to vascular patterning (e.g., Candela et al., 1999; Hickey, 1979; Nicotra et al., 2011; Turner and Sieburth, 2003). The complexity of leaf venation can be seen in a maple leaf (Acer argutum, Ellis et al., 2009; Figure 1). Maples display the species-specific branching patterns characteristic of dicots, an angiosperm group that includes the model plant Arabidopsis thaliana (arabidopsis).

Figure 1.

As well characterized as leaf venation has been, there is a recognized need for improved, automated quantification of leaf venation patterning (Ellis et al., 2009; Roth-Nebelsick et al., 2001). It is this need that we address in the current study with the VESsel GENeration Analysis (VESGEN) software, where quantification and automation would enhance the success of the experiment. Moreover, as an insightful research discovery tool, VESGEN helps to define and understand the physiological and anatomical effects of changes in complex molecular signaling. Significant, novel discoveries made with VESGEN in normal and pathological microvascular remodeling for human and vertebrate development and disease include the multimodal morphological effects of the cytokine, vascular endothelial growth factor (VEGF) (Parsons-Wingerter et al., 2006a), and the surprisingly regenerative, early-stage progression of a blinding human retinal disease (Parsons-Wingerter et al., 2010).

Plants have a long history in spaceflight research (Ferl et al., 2002; Paul et al., 2013a; Wolverton and Kiss, 2009). In orbital studies to date, leaves have been characterized primarily for photosynthetic function and chemical composition, which can be enhanced or diminished in microgravity (Levine, 2010; Monje et al., 2005; Monje et al., 2006; Musgrave, 2007; Tripathy et al., 1996). An aspect of leaf morphology that has yet to be explored in spaceflight material is the pattern of venation in leaves as they develop on orbit. A developmental approach that has been explored in numerous ground studies is the comparison of how venation patterns change in leaves of developmental stages. Sometimes referred to as “juvenile”, the first few true leaves of arabidopsis are relatively small, have little or no serrations, lack abaxial trichomes, and possess simpler venation patterns than leaves that develop later. The arabidopsis juvenile phase can be further subdivided into the younger V1 and slightly older V2 phases (Chua et al., 2005; Clarke et al., 1999; Kang and Dengler, 2004; Kankel et al., 2003; Telfer et al., 1997; Willmann and Poethig, 2011).

As a typical example, several juvenile leaves emerge first in arabidopsis (Figure 2; Kang and Dengler, 2004). Once juvenile leaves are fully expanded, their characteristics persist throughout their existence, thereby providing a distinctive snapshot of the developmental changes associated with the juvenile-to-adult leaf progressive growth sequence termed “vegetative phase change” (Poethig, 1990; Pulido and Laufs, 2010; Scarpella and Helariutta, 2010; Wu et al., 2009). The more numerous, larger adult arabidopsis leaves emerge later, expand and mature with much more complex venation patterning. In addition to the development of trichomes on both abaxial and adaxial surfaces, they also generally have greater rates of photosynthesis (Bauer and Thoni, 1988; Huijser and Schmid, 2011; Velikova et al., 2008).

Figure 2.

The VESGEN software was developed by NASA as a complex plug-in to the image processing software ImageJ (National Institutes of Health, US) to analyze normal and pathological vascular remodeling and therapeutic interventions for vertebrate systems such as human vessels in clinical images. The software algorithms are based on physiological rules for vascular branching and blood flow (Parsons-Wingerter et al., 2010; Parsons-Wingerter and Reinecker, 2012; Vickerman et al., 2009). We are now adapting VESGEN to map and quantify dicot leaf venation patterning for ISS and terrestrial applications using the same physiological rules-based strategy. Branching rules for dicot leaf venation and vertebrate vessels are both similar and different (Guyton and Hall, 2006; Parsons-Wingerter et al., 2010; Vogel, 2012). For example, the fractal-based, branching vascular systems of both dicot leaves and vertebrates are regulated by homeobox gene, hormonal and cytokine signaling (Horowitz and Simons, 2009; Lamont and Childs, 2006; Scarpella and Helariutta, 2010), and vessel branching is essentially bifurcational (Bassingthwaighte et al., 1994; Ellis et al., 2009). Most vascular systems are composed of large structural branching vessels (dendrites or trees) that transport metabolic and immune substances to and from smaller capillary reticulate networks that directly support individual cells within an organ or tissue such as a leaf. Meaningful quantification of microvascular remodeling therefore requires measurements of site-specific changes within vascular trees and networks because the averaging of vessel diameter and other vascular parameters over many successively smaller, morphologically and functionally diverse branching generations does not yield particularly useful, informative results.

Unlike the vertebrate arteriovenous circulation that is driven by a cardiac pump, the plant vasculature consists of a single branching venous system for which the gravitational, osmotic and other governing forces driving vascular flow are not yet fully understood (Taiz and Zeiger, 2010). The dense, reticulated leaf venation patterns of dicots branch sequentially to terminate as an open veinlet at each stomatic pore, where the veinlet participates in photosynthesis by delivering water and removing glucose (de Boer et al., 2012). Dicots are now the dominant plants on Earth in both number and prevalence of species (Crepet and Niklas, 2009; Parsons-Wingerter et al., 2010; Pennisi, 2010). Clearly, dicot leaf venation is important for both terrestrial life and the successful adaptation of plants to future space exploration and colonization.

VESGEN therefore represents a coalescence of vascular pattern recognition analysis across biological kingdoms by utilizing both commonalities and differences in vascular patterning in biology, especially as applied to the space life sciences context. Numerous other computer softwares currently available for imaging pattern analysis in plant tissues and organs include root and cellular morphology, leaf area, and some vascular patterning features (Lobet, 2013-2014; Price et al., 2011; Rolland-Lagan et al., 2009). The unique, innovative contribution of VESGEN derives from our design strategy that uses vascular physiological branching rules to first determine (map) the vessel branching orders (generations), and then quantify site-specific changes in the vascular pattern in response to developmental, environmental, evolutionary, and pathological stimuli.

For the ISS-grown juvenile arabidopsis Leaf 2, patterning of the larger structural vein orders 1°-2° was essentially equivalent to the terrestrial control grown under normal gravity (Table 1, Figure 3). Vessel number density (N_v), an important vascular complexity parameter, was 1.24E-5 μm^-2 for the STS-130 leaf and 1.29E-5 μm^-2 for ground control. However, increased leaf maturity and greater photosynthetic functional capacity was indicated by the smaller reticulate vein orders ≥ 3° for the STS-130 leaf, in which N_v was 1.29E-5 μm^-2 compared to 7.7E-6 μm^-2 for the ground control. Whether such a result is representative of arabidopsis leaf development in microgravity environments would clearly require significant statistical replication. However, our results do indicate the suitability of the method for addressing such questions.

Vascular complexity increased during normal terrestrial maturation of adult arabidopsis Leaf 8 (Table 2, Figure 4) by fractal-based measures of venation geometry in association with progressively modulating expression of AtHB-8::GUS. By the fractal dimension (D_f), the space-filling capacity of the leaf vascular pattern increased from 1.38 at Day 2 to 1.47 at Day 8. The numbers (N_r) of reticulate veins increased from 153 to 255, respectively, although N_r of structural veins decreased from 68 to 51. The trends of our results are in good agreement with qualitative and quantitative results reported by Kang and Dengler (2004) that included overall measurements of vessel area and counts of vessel branch points and endpoints. However, their vessel quantification did not distinguish between structural and reticulate vascular patterning. Vessel number density (N_v) for structural vessels decreased from 1.53E-4 μm^-2 at Day 2 to 2.3E-6 μm^-2 at Day 8, and for reticulate vessels, from 1.15E-4 μm^-2 to 3.9E-6 μm^-2. Results for N_v are not in good agreement with those of Kang and Dengler. The Day 8 image, which was scanned from the journal pdf file, is probably of decreased image resolution compared to the original analyzed by Kang and Dengler, which presumably was of sufficient microscopic resolution to resolve all vascular structures. Nonetheless, our results demonstrate the successful semi-automated VESGEN analysis of venation in the mature adult leaf.

At Day 2, the average intensity of our bioinformatic histogram analysis (256 gray levels) for AtHB-8::GUS co-localization with grouped structural orders (1°-2°) was 86 ± 52 (mean ± SD), and 40 ± 20 with grouped reticulate orders (≥ 3°). Results are confirmed qualitatively by visual inspection of altered expression patterns of AtHB-8::GUS (Figure 4). The greater intensity of AtHB-8::GUS in structural veins could result from either increased density of expression or from the two-dimensional image rendering of larger vessels (or both). We could not map the co-localized expression of AtHB-8::GUS into the veins at Day 8 because the journal image of AtHB-8::GUS localization did not quite overlap with the image of venation patterning. However, to illustrate a second type of useful bioinformatic analysis, the basal location of AtHB-8::GUS expression was successfully estimated and mapped as highly restricted by Day 8 to 29% of the total vascular area, compared to 100% at Day 2.

From the above results, a VESGEN ensemble that combines dimensional parameters of vascular geometry with bioinformatic dimensions of co-localized gene, protein, and other molecular expression can be formulated as:Equation 1:f,g={[f1,i,…,f2,i,…,…],[g1,i,…,g2,i,…,…]}f, g=\left\{\left[f_{1, i, \ldots}, f_{2, i, \ldots}, \ldots\right],\left[g_{1, i, \ldots}, g_{2, i, \ldots}, \ldots\right]\right\}

where the vector functions f and g consist of dynamic (i.e., time-dependent), two-dimensionally co-localized information. The set f contains parameters of vascular geometry such as D_f and N_v. The vector function g comprises the bioinformatic set of corresponding molecular expression patterns co-localized with the vascular geometry. Results for the adult arabidopsis Leaf 8 at t = Day 2 of development can now be expressed as:Equation 2:f,g={[Df,Nv1−2,Nv≥3],[ AtHB- 8:: GUS 1−2, AtHB- 8:: GUS ≥3]}f, g=\left\{\left[D_{f}, N_{v 1-2}, N_{v \geq 3}\right],\left[\text { AtHB- } 8:: \text { GUS }_{1-2}, \text { AtHB- } 8:: \text { GUS }_{\geq 3}\right]\right\}

in which f₁ = D_f; f_2,1 = N_v1-2, the vessel number density in structural veins of orders 1°-2°, and f_2,2 = N_{v ≥3}, vessel number density in reticulate veins of orders ≥3°. Many other parameters of vascular morphological complexity such as vessel diameter, length, and densities of branch points and end points can be included in f (for example, see McKay et al., 2008). For co-localized expression of molecular marker AtHB-8::GUS in structural and reticulate veins of orders 1°-2° and ≥3°, g_1,1 = AtHB-8::GUS_1-2 and g_1,2 = AtHB-8:: GUS_≥3, respectively. Vast numbers of other bioinformatic molecular expression patterns can be combined with vascular geometric information within the VESGEN systems ensemble.

As a methodological feasibility study for space and terrestrial applications, we investigated the physiological rules and methods for mapping and quantification of dicot leaf venation by VESGEN software. Our goals were to: (1) begin applying vascular branching rules of dicot leaf venation to the VESGEN analysis using Abrabidopsis as a first model, (2) assess the similarities and differences of these branching rules compared to vertebrate vascular branching, and (3) perform a semi-automated analysis of several representative juvenile and adult arabidopsis leaves to begin testing and applying the rules. Differences in vascular patterning of a juvenile arabidopsis leaf between spaceflight and ground control (Table 1, Figure 3) and terrestrial maturation in an adult arabidopsis leaf (Table 2, Figure 4) were mapped and quantified. As a further feasibility study, we examined the expression of AtHB8::GUS within the developing adult leaf. Localization of AtHB8::GUS was mapped by VESGEN onto the spatial geometry of hierarchical vascular branching for Day 2 (Table 2, Figure 4).

Previous studies of genetically controlled modifications of leaf venation in development and evolution report basic vascular measures such as overall densities of vessel length and branch points (Boyce et al., 2009; Brodribb and Feild, 2010; Candela et al., 1999; Kang and Dengler, 2004; Kang et al., 2007) and occasionally, vessel diameter and the fractal dimension (Roth-Nebelsick et al., 2001). VESGEN results for hierarchic, site-specific changes within branching vascular trees in humans and vertebrates demonstrate that remodeling vascular patterns serve as informative read-outs that necessarily integrate the interactive signaling of complex signal transduction pathways (Chen et al., 2013; Liu et al., 2009; McKay et al., 2008; Parsons-Wingerter et al., 2006a; Parsons-Wingerter et al., 2000a; Parsons-Wingerter et al., 2000b; Parsons-Wingerter et al., 2006b; Vickerman et al., 2009; Zamanian-Daryoush et al., 2013). Our goal of fully automating the VESGEN software mapping capabilities for dicot leaf venation patterning includes not only morphological parameters of Euclidean spatial dimensions (the f set, Results), but also the associated bioinformatic dimensions of localized gene, protein, and other types of molecular expression that are essentially infinite in combinatorial number (g set).

We chose arabidopsis for our first feasibility study of mapping leaf venation patterning because of its central importance in terrestrial and space plant research. Convenient features of this model organism includes its small size (20-25 cm), short growing period (several weeks), small genome (approximately 25,000 genes), and especially, its distinction as the only plant among the five major genetic model organisms that is readily susceptible to laboratory engineering of targeted single-point mutations (Figure 2; Muller and Grossniklaus, 2010; Taiz and Zeiger, 2010).

Kang and Dengler’s comprehensive results (2004) on the scaling of expanding leaf venation in arabidopsis offer important insights. Critical branching complexity parameters, such as densities of vessel branch points and endpoints, appear largely preserved throughout the enormous rescaling of the growing laminar area of Adult Leaf 8 by approximately 100-fold (i.e., vascular density is highly constant and rescaling is highly invariant). Presumably this density is highly favorable for vascular support of photosynthesis at the stomatic pores. (A similar analysis was not provided for the developing Juvenile Leaf 2 for which branch points were measured, but leaf laminar areas are lacking.) Fractal patterns in biology (Parsons-Wingerter et al., 1998) and mathematics (Mandelbrot, 1983) frequently remain invariant and rescale by preserving their geometric pattern, a mathematical property termed self-similarity. In contrast, Kang and Dengler’s results indicate that the venation of the developing adult leaf is not rescaling as a self-similar structure. If this were true, the maturing branching patterns would scale proportionally the distance between the endveinlets, rather than highly preserving their absolute geometric distance or size. Rather, local endveinlet density associated with the stomatic pores appears to be the primary determinant of vascular pattern. In contrast, recent measurements (Dhondt et al., 2012) suggest that venation in the developing arabidopsis Leaf 3 rescales with greater self-similarity. Important questions remain to be resolved about the rescaling of developing and mutated venation patterning in juvenile and especially, adult leaves.

The excellent, comprehensive Plant Image Analysis website (Lobet, 2013-2014) offers 101 softwares that analyze numerous aspects of plant imaging such as cellular phenomena, phylogeny searching, and leaf and root morphology. Several of these computer programs analyze characteristics of leaf venation patterning such as leaf laminar area and areolar extraction. The software capability of quantifying overall vessel density in arabidopsis leaves (Rolland-Lagan et al., 2009; Price et al., 2011) are included in the Plant Image Analysis website, and were also reported in our own first, preliminary VESGEN study (Parsons-Wingerter and Vickerman, 2011) that further quantified large (structural) and small (reticulate) vein density. For our new, more advanced VESGEN analysis, up to seven specific branching orders of leaf venation were mapped and quantified (Tables 1-2, Figures 3-4). Results from branching rules for dicot leaf venation were compared to those from vertebrate vascular branching (Figure 3). In addition, the geometric orders of leaf vascular branching were associated for the first time with co-localized mappings of the growth co-factor AtHB-8 as a bioinformatic (information) dimension (Results, Figure 4).

VESGEN and other software analyzing plant morphology can be considered by researchers studying plant development on the ISS with microscopes supported by NASA, the European Space Agency, and the Japanese Aerospace Exploration Agency. For example, the NASA Light Microscopy Module (LMM) that has previously supported colloidal experiments in fluid physics is now being developed for NASA Life Sciences and Space Biology applications (Sicker and Meyer, 2013). Feasibility experiments (STS-134 and 135 payloads) were performed with the LMM to demonstrate imaging of the nematode, fruit fly and other organisms at various conditions of magnification, fluorescence, and live video streaming. Such recent and ongoing experiments will reduce the integration, operations, and experiment development costs of our proposed VESGEN experiments.

The complexity of adult leaf venation patterning in dicot angiosperms, a defining feature of their optimized photosynthetic efficiency, is of enormous functional importance. Perhaps in part because of this system’s complexity, venation patterning in adult dicot leaves are understudied on Earth and therefore on the ISS (Ellis et al., 2009; Kang and Dengler, 2004; Roth-Nebelsick et al., 2001). Automating the hierarchical mapping of branching venation pattern would contribute important new capabilities for analyzing leaf adaptations on Earth and in space in applications ranging from fundamental botany to ecology, environmental engineering, and agriculture.