Expression of PIN1, PIN5, PIN6, and PIN8 during leaf development
Veins form sequentially during Arabidopsis leaf development: the formation of the midvein is followed by the formation of the first loops of veins (“first loops”), which in turn is followed by the formation of second loops and minor veins [4, 5, 12, 13] (Fig. 1a-c).
Two distinct auxin-transport pathways have overlapping functions in control of Arabidopsis vein-network geometry [33]. One pathway—mediated by the PM-localized PIN1 protein—transports auxin intercellularly [27, 35]; the other pathway—mediated by the ER-localized PIN5, PIN6, and PIN8 proteins—transports auxin intracellularly [27–34].
Consistent with their role in control of vein network geometry [4, 33, 51, 52], PIN1 (AT1G73590), PIN6 (AT1G77110), and PIN8 (AT5G15100) are expressed in developing veins, although with different dynamics: expression of PIN1 and PIN6 is initiated in broad domains of leaf inner cells, domains that over time become restricted to single files of vascular precursor cells [33, 38–40] (Fig. 1d-i); by contrast, PIN8 expression is restricted from early on to single files of leaf vascular cells [33] (Fig. 1j-l). It remains unclear, however, whether these different dynamics of PIN expression comprise onset of PIN expression at different stages of leaf development.
To address this question, we compared expression of PIN1, PIN6, and PIN8 in first leaves 2, 3 and 4 days after germination (DAG). To visualize PIN expression, we used functional translational fusions (PIN promoter driving expression of the respective PIN:reporter fusion protein) [33, 37, 53] or transcriptional fusions (PIN promoter driving expression of a reporter protein) [33] (Additional file 1: Table S1); whenever we used transcriptional fusions, their expression matched that of the respective, functional translational fusions [33] (Additional file 2: Figure S1), suggesting that those PIN promoters contain all the regulatory elements required for functional expression of the respective genes.
While expression of a PIN1::PIN1:GFP translational fusion (PIN1 promoter driving expression of PIN1:GFP fusion protein) and of a PIN6::YFPnuc transcriptional fusion (PIN6 promoter driving expression of a nuclear yellow fluorescent protein) was already visible 2 DAG (Fig. 1d, g), expression of PIN8::YFPnuc was first detected 3 DAG (Fig. 1j, k), suggesting that PIN8 expression is initiated after the onset of expression of both PIN1 and PIN6.
PIN5 (AT5G16530) is expressed in veins of mature leaves [28, 54], but its expression during leaf development is unknown. Transcriptional and translational fusions of PIN5 are expressed in similar domains [28, 34, 54], suggesting that the PIN5 promoter contains all the regulatory elements required for PIN5 expression. Thus, to visualize PIN5 expression during leaf development, we imaged PIN5::YFPnuc expression in first leaves 2, 3, 4, 5 and 5.5 DAG.
Expression of PIN5::YFPnuc was first detected in the midvein of 4-DAG leaves (Fig. 1m-o); at 5 DAG, PIN5::YFPnuc was additionally expressed in first loops (Fig. 1p), and at 5.5 DAG PIN5::YFPnuc was additionally expressed in second loops and minor veins (Fig. 1q). Thus our results suggest that PIN5 expression is initiated after PIN8 expression (Fig. 1k, n, o) and that, as PIN8, PIN5 is expressed from early on in single files of leaf vascular cells (Fig. 1k, l, o-q).
Expression of PIN5 in single files of leaf vascular cells—suggested by PIN5::YFPnuc expression—was supported by expression of two functional (Additional file 1: Table S1) [34] PIN5::PIN5:GFP translational fusions (Fig. 1r).
Expression of PIN1, PIN5, PIN6, and PIN8 in leaf vascular cells
Because PIN1, PIN5, PIN6, and PIN8 are all expressed in developing veins (Fig. 1), we asked whether these genes were expressed in the same vascular cells. To address this question, we imaged pairwise combinations of fluorescent reporters of PIN1, PIN5, PIN6, and PIN8 in midvein cells of 4-DAG first leaves—where these genes are expressed (Fig. 1f, i, l, o)—and quantified reporter coexpression.
In none of the 20 analyzed leaves coexpressing PIN5::YFPnuc and PIN6::CFPnuc (PIN6 promoter driving expression of a nuclear cyan fluorescent protein) were cells expressing PIN5::YFPnuc ever on the same plane as cells expressing PIN6::CFPnuc: cells expressing PIN5::YFPnuc were located ventrally, while cells expressing PIN6::CFPnuc were located dorsally (Fig. 2a-c). Likewise, in none of the 20 analyzed leaves coexpressing PIN8::YFPnuc and PIN6::CFPnuc were cells expressing PIN8::YFPnuc ever on the same plane as cells expressing PIN6::CFPnuc: cells expressing PIN8::YFPnuc were located ventrally, while cells expressing PIN6::CFPnuc were located dorsally (Fig. 2d-f). And although cells expressing PIN5::YFPnuc or PIN8::PIN8:GFPMGS were both on the same ventral plane (Fig. 2g-i), only fewer than 3 % of the cells expressing either reporter expressed both (Fig. 2s).
Approximately 95 % of PIN5::YFPnuc-expressing cells expressed PIN1::PIN1:GFP, but only ~25 % of the PIN1::PIN1:GFP-expressing cells that were on the same ventral plane as cells expressing PIN5::YFPnuc expressed this reporter (Fig. 2j-l, s). Likewise, ~90 % of PIN8::YFPnuc-expressing cells expressed PIN1::PIN1:GFP, but only ~25 % of the PIN1::PIN1:GFP-expressing cells that were on the same ventral plane as cells expressing PIN8::YFPnuc expressed this reporter (Fig. 2m-o, s). Finally, consistent with previous observations [33], ~95 % of PIN6::YFPnuc-expressing cells expressed PIN1::PIN1:GFP, and ~75 % of the PIN1::PIN1:GFP-expressing cells that were on the same dorsal plane as cells expressing PIN6::YFPnuc expressed this reporter (Fig. 2p-s).
Thus our results suggest that PIN5, PIN6, and PIN8 are expressed in mutually exclusive domains of leaf vascular cells, and that the PIN1 cellular-expression domain overlaps with—but extends beyond—the ER-PIN cellular-expression domain.
Unique and redundant functions of PIN1, PIN5, PIN6, and PIN8 in control of vein network topology
PIN1, PIN5, PIN6, and PIN8 control vein network geometry [4, 33, 51, 52]; we asked what their functions are in control of vein network topology.
To characterize vein network topology, we derived (see
Methods and Additional file 3: Figure S2 for details) and used three descriptors based on numerical graph invariants: a cardinality index, a continuity index, and a connectivity index.
The cardinality index is a proxy for the number of “veins” (i.e. stretches of vascular elements that contact other stretches of vascular elements at least at one of their two ends) in a network (Fig. 3a).
The continuity index quantifies how close a vein network is to a network with the same number of veins but in which at least one end of each “vein fragment” (i.e. a stretch of vascular elements that are free of contact with other stretches of vascular elements) contacts a vein. The continuity index ranges from 0—for a network of sole vein fragments—to 1—for a network without vein fragments (Fig. 3a).
The connectivity index quantifies how close a vein network is to a network with the same number of veins but in which both ends of each vein or vein fragment contact other veins. The connectivity index ranges from 0—for a network of “open” veins (i.e. veins that contact vein fragments or other veins only at one end)—to 1—for a network of “closed” veins (i.e. veins that contact vein fragments or other veins at both ends) (Fig. 3a).
Although the number of veins in a leaf is variable and it is unpredictable whether a developing vein will remain open at maturity [6, 7, 10, 12–14], the cardinality and connectivity indices of vein networks in different populations of WT leaves grown in identical conditions were reproducible (Figs. 3, 6, 7 and 8). This observation suggests that while the outcome of vein formation events is unpredictable for single veins, it is predictable—within the limits of statistical variation—for networks of veins. Thus—as for non-stereotyped animal-networks (reviewed in [55])—topology descriptors such as the cardinality and connectivity indices can be compared statistically across genotypes and conditions to identify reproducible patterns and their controls.
The continuity index of vein networks in different populations of WT leaves grown in identical conditions was also reproducible (Additional file 4: Figure S3)—a finding consistent with the stringent requirement for continuity of tissue systems with transport function, such as vein networks, and with the successful use of vein fragmentation as diagnostic criterion for the identification of mutants in genetic screens [18, 56, 57].
The continuity index of none of the mutants or transgenics in our study was different from that of WT (Additional file 4: Figure S3), suggesting that PIN1, PIN5, PIN6, and PIN8 have no function in control of vein continuity or their functions in this process are redundant.
Consistent with previous observations [33], the vein network topology of pin5, pin6, or pin8 was no different from that of WT (Fig. 3b); by contrast, the cardinality and connectivity indices of pin1 vein networks were higher than those of WT vein networks (Fig. 3b), suggesting that PIN1 inhibits the formation of veins and their connection.
We next asked whether PIN5, PIN6, or PIN8 acted redundantly with PIN1 in inhibition of vein formation and connection. The vein network topology of neither pin1;pin5 (pin1;5 hereafter) nor pin1;8 differed from that of pin1 (Fig. 3b); however, the cardinality and connectivity indices of pin1;6 vein networks were higher than those of pin1 vein networks (Fig. 3b), suggesting that PIN6 acts redundantly with PIN1 in inhibition of vein formation and connection.
Next, we asked whether PIN5 or PIN8 acted redundantly with PIN6 in PIN1-dependent inhibition of vein formation and connection. The vein network topology of pin1;5;6 was no different from that of pin1;6 (Fig. 4), but the cardinality index of pin1;6;8 vein networks was higher than that of pin1;6 vein networks (Fig. 4), suggesting that PIN8 acts redundantly with PIN6 in PIN1-dependent inhibition of vein formation; by contrast, the connectivity index of pin1;6;8 vein networks was no different from that of pin1;6 vein networks (Fig. 4), suggesting that PIN8 has no function redundant to that of PIN6 in PIN1-dependent inhibition of vein connection. Because the vein network topology of neither pin6 nor pin8 differs from that of WT (Fig. 3b), but the cardinality index of pin6;8 vein networks is higher than that of WT (Fig. 3b), PIN6 and PIN8 also have redundant functions in inhibition of vein formation that are independent of PIN1. Thus the enhancement of pin1;6 cardinality defects by PIN8 could be interpreted as the result of the simultaneous loss of the PIN1-dependent pathway and of the parallel, PIN6/PIN8-dependent, PIN1-independent pathway—rather than evidence that PIN8 acts redundantly with PIN6 in PIN1-dependent inhibition of vein formation. However, we do not favor this interpretation because the cardinality defect of pin1;6;8 is much greater than the sum of the cardinality defects of pin1 and pin6;8.
We finally asked whether PIN5 acted redundantly with PIN6 and PIN8 in PIN1-dependent or PIN1-independent inhibition of vein formation. The vein network topology of pin1;5;6;8 was no different from that of pin1;6 (Fig. 4) and that of pin5;6;8 was no different from that of WT (Fig. 3b), suggesting that pin5 suppresses the effects of pin6 and pin8 on PIN1-independent inhibition of vein formation. In agreement with interpretations of similar genetic interactions in other organisms (e.g., [58–60]), the most parsimonious account for our observations is that PIN5 promotes vein formation; that PIN6 and PIN8 redundantly and completely inhibit PIN5-dependent promotion of vein formation; and that these functions of PIN5, PIN6, and PIN8 are independent of PIN1. Further, because expression of PIN5 and PIN8 is initiated at post-formative stages of vein development [33] (Fig. 1), these genes most likely control vein formation indirectly—for example, through feedback on vascular precursor cells located in more-immature parts of the leaf (e.g., [25]; reviewed in [61, 62]). Finally, because PIN5, PIN6, and PIN8 are expressed in non-overlapping sets of vascular cells (Fig. 2), the genetic interaction between these genes—as that between other genes expressed in mutually exclusive domains (e.g., [63–67] and references therein)—presumably reflects underlying cell-cell interactions.
Redundant functions of PIN1, PIN6, and PIN8 in control of auxin distribution in developing leaves
PIN1 inhibits vein formation, and PIN6 acts redundantly with PIN1 in inhibition of vein formation and with PIN8 in PIN1-independent inhibition of vein formation (Figs. 3 and 4). We asked whether such redundancy extended to control of auxin distribution in developing leaves, which is known to control vein formation [4, 5, 23, 33, 38, 39, 68]. To address this question, we imaged expression of the auxin reporter DR5rev::YFPnuc [33, 36, 69] in 4-DAG first leaves of WT, pin6;8, pin1, and pin1;6.
As previously reported [22, 33, 38, 68], in WT the DR5 promoter was strongly active in narrow domains that coincide with sites of vein formation (Fig. 5a). Consistent with previous observations [29–33], DR5rev::YFPnuc expression was weaker in pin6;8 than in WT, but domains of DR5rev::YFPnuc expression were equally narrow in pin6;8 and WT (Fig. 5a, b). Levels of DR5rev::YFPnuc expression were lower, and domains of DR5rev::YFPnuc expression were broader, in pin1 than in WT or pin6;8 (Fig. 5a-d); and DR5rev::YFPnuc expression levels were even lower, and DR5rev::YFPnuc expression domains even broader, in pin1;6 (Fig. 5c-f).
Thus our results suggest that the redundancy between PIN1, PIN6, and PIN8 that underlies control of vein formation extends to control of auxin distribution in developing leaves (see Conclusions).
Homologous and nonhomologous functions of PIN1 and PIN6 in vein network formation
PIN6 acts redundantly with PIN1 in control of vein network geometry [33] and topology (Fig. 3b); however, the redundancy between PIN1 and PIN6 is unequal: the geometry and topology of pin6 vein networks are no different from those of WT vein networks but those of pin1 vein networks are, suggesting that PIN1 can provide all—or nearly all—the functions of PIN6 in vein network formation and that, by contrast, PIN6 is unable to provide all the functions of PIN1 in this process. Such unequal redundancy could reflect nonhomologous functions of PIN1 and PIN6 in vein network formation—a possibility consistent with the different localization of PIN1 and PIN6: PIN1 is predominantly localized to the PM [35], while PIN6 is predominantly localized to the ER [33]. On the other hand—at least in other organisms—redundant, homologous functions can be provided by proteins that are localized to different cellular compartments (e.g., [70, 71] and references therein). Further, at least some of the functions of PIN1 in vein network formation depend on PIN1 expression in leaf epidermal cells [51, 72]—leaf epidermal cells that fail, by contrast, to express PIN6 [33] (Fig. 1). Thus the unequal redundancy of PIN1 and PIN6 in vein network formation could alternatively be accounted for by their different expression domains.
To test these possibilities, we used the promoter of the RIBOSOMAL PROTEIN S5A (RPS5A) gene (AT3G11940)—highly active in developing organs, including their epidermal cells [73]—to express PIN1 (RPS5A::PIN1) or PIN6 (RPS5A::PIN6) in the pin1 background, and compared phenotype features of RPS5A::PIN1;pin1 and RPS5A::PIN6;pin1 with those of pin1 and WT.
We first asked whether PIN6 could provide functions in control of vein network geometry homologous to those of PIN1. The geometry of ~15 % of the vein networks of RPS5A::PIN1 and RPS5A::PIN6, and—as previously reported [33]—of nearly 50 % of pin1 vein networks was abnormal (Fig. 6a-d). RPS5A::PIN1 shifted the spectrum of vein network geometries of pin1 toward the vein network geometry of WT but RPS5A::PIN6 failed to do so (Fig. 6a-d), suggesting that PIN6 is unable to provide functions in control of vein network geometry homologous to those of PIN1.
We next asked whether PIN6 could provide functions in control of vein network topology homologous to those of PIN1. The cardinality and connectivity indices of RPS5A::PIN1 vein networks were lower than those of WT vein networks (Fig. 6e), supporting that PIN1 inhibits vein formation and connection. The cardinality index of RPS5A::PIN6 vein networks was higher than that of WT vein networks (Fig. 6e), suggesting that ectopic expression of PIN6 in the epidermis promotes vein formation. As reported above (Fig. 3b), the cardinality and connectivity indices of pin1 vein networks were higher than those of WT vein networks (Fig. 6e). RPS5A::PIN1 shifted the cardinality index of pin1 vein networks toward that of WT vein networks but RPS5A::PIN6 failed to do so (Fig. 6e), suggesting that PIN6 is unable to provide functions in vein formation homologous to those of PIN1. By contrast, both RPS5A::PIN1 and RPS5A::PIN6 shifted the connectivity index of pin1 vein networks toward that of WT vein networks (Fig. 6e), suggesting that PIN6 can provide functions in vein connection homologous to those of PIN1. Interpretations of similar genetic interactions in other organisms (e.g., [74–76]) suggest that the suppression of vein connectedness defects of pin1 by RPS5A::PIN6 can be accounted for by at least two mechanisms. One possibility is that PIN6 acts downstream of PIN1 in the same pathway that controls vein connection; we do not favor this hypothesis, however, because it fails to predict the observed (Figs. 3 and 4) enhancement of vein connectedness defects of pin1 by pin6. Alternatively, vein connection may be unfavored at high auxin levels [77], which would be the result of at least two separate pathways: PIN1-mediated auxin transport toward sites of vein formation [38–41] (Fig. 5) and PIN6-mediated increase in auxin levels within developing vascular cells [31–33] (Fig. 5) (see Conclusions).
In addition to vein network formation, PIN6 acts redundantly with PIN1 in cotyledon patterning, and as in vein network formation, the redundancy between PIN1 and PIN6 in cotyledon patterning is unequal [33]. We thus asked whether PIN6 could provide functions in cotyledon patterning homologous to those of PIN1; our results (Additional file 5: Figure S4) suggest that it cannot.
Finally, RPS5A::PIN1 reverted the pin-shaped, sterile inflorescences of pin1 to WT-looking, fertile inflorescences but RPS5A::PIN6 failed to do so (Additional file 6: Figure S5), suggesting that PIN6 is unable to provide functions in inflorescence development homologous to those of PIN1.
In summary, PIN6 was unable to provide functions homologous to those of PIN1 in control of vein network geometry, vein formation, cotyledon patterning, and inflorescence development. Thus the unequal redundancy between PIN1 and PIN6 in these processes is unlikely to be the result of their different expression and might instead be accounted for by their nonhomologous functions—a conclusion consistent with the opposite effects of PIN1 and PIN6 on intercellular auxin transport [32]. By contrast, PIN6 was able to provide functions in vein connection homologous to those of PIN1, suggesting that PIN6 expression normally limits the ability of PIN6 to compensate for the effects of loss of PIN1 function in vein connection.
Homologous functions of PIN6 and PIN8 in PIN1-dependent vein-network formation
PIN8 acts redundantly with PIN6 in PIN1-dependent control of vein network geometry [33] and vein formation (Fig. 4); however, the redundancy between PIN6 and PIN8 in PIN1-dependent control of vein network formation is unequal: the geometry and cardinality index of pin1;8 vein networks are no different from those of pin1 vein networks, but those of pin1;6 vein networks are; thus PIN6 can provide all the functions of PIN8 in PIN1-dependent control of vein network geometry and vein formation, but PIN8 is unable to provide all the functions of PIN6 in these processes. Further, PIN8 seems to have no function in PIN1/PIN6-dependent vein connection. The unequal functions of PIN6 and PIN8 in vein network formation could be accounted for by the different expression of PIN6 and PIN8 during vein development [33] (Figs. 1 and 2), but it could also reflect nonhomologous functions of PIN6 and PIN8 in this process.
To test these possibilities, we expressed PIN6 or PIN8 by the promoter of the MONOPTEROS (MP) gene (AT1G19850) (MP::PIN6 or MP::PIN8)—highly active in developing veins [33]—in the pin1;6 background, and compared defects of MP::PIN6;pin1;6 and MP::PIN8;pin1;6 with those of pin1;6 and pin1.
We first asked whether PIN8 could provide functions in PIN1-dependent control of vein network geometry homologous to those of PIN6. As previously reported [33], the vein network geometry of MP::PIN6 and MP::PIN8 was no different from that of WT (Fig. 7a-e). By contrast, the geometry of nearly 60 % of pin1 vein networks was abnormal, and pin6 shifted the spectrum of vein network geometries of pin1 toward more severe phenotype classes [33] (Fig. 7a-e). The spectrum of vein network geometries of MP::PIN6;pin1;6 was no different from that of pin1 and that of MP::PIN8;pin1;6 was no different from that of MP::PIN6;pin1;6 (Fig. 7a-e), suggesting that PIN8 can provide functions in PIN1-dependent control of vein network geometry homologous to those of PIN6.
We next asked whether PIN8 could provide functions in PIN1-dependent control of vein network topology homologous to those of PIN6. Consistent with previous observations [33], MP::PIN6 and MP::PIN8 induced similar defects—as it frequently results from overexpression of genes with homologous functions (e.g., [78–80]): the cardinality and connectivity indices of both MP::PIN6 and MP::PIN8 vein networks were lower than those of WT vein networks (Fig. 7f), supporting that PIN6 inhibits vein formation and connection, and suggesting that PIN8 can inhibit vein connection in addition to vein formation. As reported above (Fig. 3b), the cardinality and connectivity indices of pin1 vein networks were higher than those of WT vein networks and those of pin1;6 vein networks were higher than those of pin1 vein networks (Fig. 7f). The vein network topology of MP::PIN6;pin1;6 was no different from that of pin1 and that of MP::PIN8;pin1;6 was no different from that of MP::PIN6;pin1;6 (Fig. 7f), suggesting that PIN8 can provide functions in PIN1-dependent control of vein network topology homologous to those of PIN6.
In addition to PIN1-dependent vein-network formation, PIN8 acts redundantly with PIN6 in PIN1-dependent cotyledon patterning, and as in PIN1-dependent vein network formation, the redundancy between PIN6 and PIN8 in PIN1-dependent cotyledon patterning is unequal [33]. We thus asked whether PIN8 could provide functions in PIN1-dependent cotyledon patterning homologous to those of PIN6; our results (Additional file 7: Figure S6) suggest that it can.
In summary, PIN8 was able to provide functions homologous to PIN6 in PIN1-dependent vein network formation and cotyledon patterning. Thus the unequal redundancy between PIN6 and PIN8 is unlikely the result of nonhomologous functions and might instead be accounted for by their different expression. Just as the ER-PIN genes PIN6 and PIN8 redundantly control PIN1-dependent vein network formation, the redundancy between the PM-PIN genes PIN1, PIN2, PIN3, PIN4, and PIN7 underlies— to varying extents—many other developmental processes (e.g., [37, 52, 81–84]). In the development of embryos and roots, PM-PIN genes compensate for loss of one another’s function by their ectopic expression in the domain of the gene whose function has been lost [82, 84]. For example, in pin7 embryos PIN4 becomes expressed at earlier stages of development and in the domain in which PIN7 is normally expressed, thereby compensating for loss of PIN7 function [84]. By contrast, in the pin1;6 background PIN8 expression remains restricted to post-formative stages of vein development [33], supporting that PIN8 controls vein network formation by feeding back on vascular precursor cells located in more-immature parts of the leaf.
Functions of PIN5 in PIN6/PIN8-dependent control of vein network topology
PIN6 has functions in control of vein network topology beyond control of PIN5 function (Fig. 4). We asked whether PIN5 could provide functions in control of vein network topology that are independent of control by PIN6 or PIN8.
To address this question, we used plants expressing PIN5 by the MP promoter (MP::PIN5) because the vein density of MP::PIN5 leaves is higher than that of WT leaves [33]. We reasoned that if PIN5 could provide functions that are independent of control by PIN6 or PIN8, at least some of the effects of MP::PIN5 on vein network topology should persist in the MP::PIN6 or MP::PIN8 backgrounds. By contrast, if all PIN5’s functions depended on control by PIN6 or PIN8, the effects of MP::PIN6 or MP::PIN8 on vein network topology should mask those of MP::PIN5.
Consistent with previous observations [33], the cardinality index of MP::PIN5 vein networks was higher than that of WT vein networks (Fig. 8), supporting that PIN5 promotes vein formation. As reported above (Fig. 7), the cardinality and connectivity indices of MP::PIN6 and MP::PIN8 vein networks were lower than those of WT vein networks (Fig. 8). Because the vein network topology of MP::PIN5;MP::PIN6 was no different from that of MP::PIN6 and that of MP::PIN5;MP::PIN8 was no different from that of MP::PIN8 (Fig. 8), we conclude that no function of PIN5 escapes control by PIN6 or PIN8.