In response to local infection with necrotizing pathogen, plants display a whole plant enhanced immunity to secondary challenge (systemic acquired resistance or SAR) 1. A characteristic set of genes termed pathogenesis-related (PR) is induced both locally and in distal tissues in parallel with SAR induction 2. Accumulation of salicylic acid (SA) has been shown to be necessary for induction of SAR 3.

The Arabidopsis ndr1-1 mutant showed drastically impaired induction of SAR and PR-1 gene expression in response to infection with Pseudomonas syringae bacteria carrying the avirulence (avr) gene avrRpt24. This mutant is also incapable of showing a hypersensitive response (HR) to these bacteria 56 except at very high levels of inoculum 7. The HR is a macroscopic leaf collapse. The HR occurs in response to infection with avirulent bacteria at levels of inoculum that cause a large percentage of leaf cells to undergo programmed cell death (PCD) 8. In contrast to results with avrRpt2, P. syringae bacteria carrying the avirulence genes avrB, avrRpm1, avrPphB or avrRps4 elicit an exaggerated HR relative to that on the wild-type Columbia parent 57. In a study in which bacteria carrying avrB were chosen as a representative member of this class, in further contrast to results with avrRpt2, these bacteria were capable of eliciting SAR and PR-1 induction on ndr1-1 mutant plants 4.

However, induction of SA accumulation by ndr1-1 mutant plants was impaired at early time points to a similar extent in response to bacteria carrying either avr gene 4. These results suggested that at least two pathways control SAR and PR-1 gene expression. One pathway is directly dependent on SA accumulation. The second pathway is not directly dependent on SA levels. The correlation between ability to elicit the HR and to elicit SAR and PR-1 induction led to the model that the second pathway was caused by signals associated with hypersensitive cell death. Despite the severe impairment in SA accumulation within the first 12 hours post-inoculation of ndr1 plants, bacteria carrying avrB could nonetheless induce PR gene expression and SAR because of the HR-associated pathway. Bacteria carrying avrRpt2 could not because both pathways were impaired by the ndr1 mutation. This study could not address the question as to whether the HR-associated pathway alone was sufficient to induce SAR and PR-1 because SA accumulation did occur late in the response.

Induction of SAR has been reported to be dependent upon the NPR1 gene 91011. In these studies, neither SA/SA analogs nor avirulent bacteria were able to induce SAR on npr1 mutant plants. The biological inducer of SAR tested in these experiments was P. syringae bacteria carrying avrRpt2. In contrast to ndr1 plants, npr1 plants do show the HR in response to P. syringae carrying avrRpt2 (data not shown). These results could be taken to suggest that the HR-associated pathway is not sufficient to induce SAR in the absence of the NPR1-dependent, SA-associated pathway (but see below for why this conclusion is not correct).

In contrast to SAR induction, induction of PR-1 by avirulent bacteria has been shown to occur in npr1 mutants 11. It is, however, reduced in extent and delayed. If npr1 mutants are indeed fully blocked in SAR induction, then the pathways leading to PR-1 induction can not be the same as those leading to SAR. However, we have argued on the basis of results with ndr1 mutant plants that HR "triggering" in response to the avrRpt2/RPS2 combination is weaker than that in response to avrB/RPM14. Weaker "triggering" results in a lower probability of individual cells undergoing PCD early in the HR, prior to the onset of potentiation caused by increased SA levels 412. Signaling elicited by avrRpt2 clearly differs in many respects from that directed by the other avr genes. Use of bacteria carrying avrB to elicit SAR would allow a more rigorous test of whether the NPR1-dependent pathway is essential for SAR induction.

To clarify the signaling circuitry governing SAR and PR gene induction, we have constructed a ndr1/npr1 double mutant. This line was compared with wild-type Columbia Arabidopsis and both single mutant lines for the ability to display biological induction of SAR and PR gene expression. We conclude that SAR, PR-1 induction and PR-5 induction are regulated in parallel by the two pathways we have previously delineated 4, but the effects of these pathways are additive rather than obligatorily synergistic. Moreover, additional pathways also contribute additively to PR-5 induction.

PR genes are thought to be tightly associated with SAR because they are co-regulated with SAR 1314 and because many PR gene products have been shown to have antimicrobial activity 2. The complete impairment of npr1 plants in SAR induction 9 as contrasted with a partial impairment in PR-1 induction 11 in response to P. syringae carrying avrRpt2 challenged the hypothesis of co-regulation. However, avrRpt2 was shown to differ markedly from the other characterized P. syringae avr genes in the responses it elicited on ndr1 mutant plants 457. The apparent challenge to the hypothesis of co-regulation could have been because of the choice of an atypical avr gene. Therefore, SAR and PR gene expression of npr1 mutant plants were assessed in response to a second avr gene (avrB). Markedly different results were obtained. Mutant npr1 plants were not fully impaired in induction of SAR, PR-1 gene expression or PR-5 gene expression in response to P. syringae carrying avrB. These results are consistent with the hypothesis of co-regulation. The apparent NPR1-dependence of SAR induction in response to bacteria carrying avrRpt2 is likely quantitative rather than qualitative.

A potential alternative explanation for these results might be residual function of mutant gene products in the plant lines used. The ndr1-1 mutant is an unambiguous null mutant in that a deletion removes most of the coding region and the proximal promoter region 6. However, none of the characterized npr1 mutants have a large deletion 1516. As such, attempts to define allelic series have relied upon the level of residual function associated with these mutants in assays of gene expression or fungal disease resistance 1516. However, the complexity of signaling leading to these outputs complicates analyses of these phenotypes. A "null" mutant might still show gene expression or disease resistance caused by pathways that do not require the corresponding gene product.

What is required is an assay where the output is due solely to an NPR1-dependent pathway. The experiment presented in Figure 2 provided such data. The combination of npr1-2 with ndr1-1 in a double mutant prevented the avrRpt2-elicited HR under the conditions used. Under these conditions, DC3000•avrRpt2 failed to elicit any PR-1 gene expression (the level was statistically indistinguishable from that seen with inoculation of a MgCl₂ blank). Impairment of SA accumulation in the double mutant (Zhang, Gutsche and Shapiro, manuscript in preparation) resembled that in the ndr1 single mutant 4 at early time points prior to the onset of NPR1-dependent negative feedback on SA accumulation 1011. Nonetheless, SA accumulation likely occurred by the time points taken in SAR and PR gene expression experiments reported herein.

This assay depends solely on NPR1 function. The complete lack of PR-1 induction supports our contention that npr1-2 is indeed a null mutant or perhaps a severe hypomorph. This contention is further supported by published data showing that the npr1-2 allele failed to interact with putative target TGA transcription factor genes in a two-hybrid assay 17 or in an in vivo, reporter gene-based assay 18. The treatment of npr1-2 seedlings with exogenous SA failed to induce increases in PR-1 gene expression under some experimental conditions 19 albeit not under all conditions 15. The potential alternative explanation that results presented herein were caused by residual gene product function in the mutant lines used is thus not correct.

The idea that SAR and PR gene expression are co-regulated has also been challenged by experiments with the eds5 and sid2 mutants 20. These mutants have been implicated in SA biosynthesis. They showed impairments in PR-1 induction in response to bacteria that were very similar to those displayed by ndr1 plants but marginal if any impairment in PR-2 or PR-5 induction. Direct comparison with results presented here required addressing an issue with the potential to affect interpretation of that study. In that study, PR gene expression was quantitated from uninfected parts of leaves that had been infected with 1 × 10⁸ mL^-1 DC3000•avrRpt2. This level of inoculum was sufficiently high to cause "false positive" spurious HR-like responses to DC3000•empty vector 5. Indeed, when sid2 mutant plants were inoculated with 2 × 10⁷ mL^-1 DC3000•avrRpt2, no HR was seen (data not shown).

The high levels of inoculum used in the cited study can only be used without causing "false positive" HRs if a bacterial strain that is not a pathogen of Arabidopsis is used to deliver the avr gene product 57. When sid2 plants were inoculated with 1 × 10⁸ mL^-1 of a nonpathogen strain (P. syringae pv. glycinea Race 5) carrying avrRpt2, HRs were seen (data not shown). The identical results with the two bacterial strain backgrounds imply that the cited results with sid2 mutant plants were obtained under conditions in which high levels of bacterial inoculum overcame some effects of the mutation. The ndr1-1 mutation can also be overcome at high levels of inoculum. However, the level at which ndr1-1 phenotypes are overcome is higher than that for sid2 phenotypes in that 1 × 10⁹ mL^-1 of a nonpathogen strain carrying avrRpt2 is required to yield reproducible HRs on ndr1-1 plants 7.

Clearly, the interpretation of results with PR-5 induction depends critically on the experimental design. The relatively unimpaired PR-2 and PR-5 induction in sid2 plants might imply that the HR-associated signals were quantitatively more important than the SA-associated signals. The same two pathways invoked to explain SAR and PR-1 induction could potentially account for PR-2 and PR-5 induction if this assumption were correct.

The results presented herein, however, argue strongly that PR-5 induction is indeed regulated differently than SAR or PR-1 induction. The two pathways identified as making the major contributions to PR-1 and SAR induction were shown to contribute to PR-5 induction. However, significant PR-5 induction occurred in the double mutant in response to DC3000•avrRpt2 (Figure 3). DC3000•avrRpt2 elicited HRs on Columbia or npr1-2 plants but not on ndr1-1 plants or double mutant plants (data not shown). As such, these results can not be explained by HR-associated signals or SA-associated, NPR1-dependent signals. There must be additional pathways contributing to PR-5 induction. This conclusion is consistent with the original interpretation of PR gene expression data in eds5 and sid2 mutants that PR-1 and PR-5 are regulated differently 20.

The data presented in Figures 2 and 3 suggest that there must be at least two additional pathways contributing to PR-5 induction. DC3000•empty vector did not elicit significant PR-1 induction on the double mutant at either time point (Student's t test, P > 0.05). However, the analogous experiment with PR-5 induction gave the opposite results. Moreover, at both time points, the PR-5 induction elicited by infection with DC3000•empty vector was to a similar extent regardless of allele status at the NDR1 or NPR1 loci (no statistically significant differences, Student's t test, P > 0.05). These bacteria elicited PR-5 gene expression in spite of the absence of a functional avr gene, the absence of PCD, and mutations in both the NPR1 and NDR1 genes. As even E. coli has been shown to induce plant defense-related gene expression 21, it should not be surprising that bacteria can elicit changes in plant gene expression via many different signaling pathways. Indeed, conserved features of bacteria can act as elicitors of gene expression through pathways distinct from those associated with gene-for-gene disease resistance 22.

However, an additional pathway leading to PR-5 gene expression that stems from avrRpt2 function must also exist. DC3000•avrRpt2 elicited no statistically significant induction of PR-1 by the double mutant at either time point (Student's t test, P > 0.05). However, clear induction of PR-5 was seen in the analogous experiment. The level of PR-5 induction was higher than in response to DC3000•empty vector (Student's t test, P < 0.05). As the double mutant was susceptible to disease caused by this bacterial strain (data not shown), perhaps this residual gene induction was due to the virulence function of avrRpt223. Mutant npr1 plants have previously been shown to be impaired in induction of PR-1 but not PR-5 in response to virulent bacterial pathogen 19.

These details are presented in a model of the signaling network for induction of SAR, PR-1 gene expression and PR-5 gene expression by avirulent P. syringae bacteria (Figure 4). We have used a representation scheme proposed recently for plant signal transduction networks that is based on Boolean formalisms used in engineering and computer science 24. Either of two "generators" shown can elicit signaling. One generator is infection of plants carrying a functional RPS2 allele with bacteria carrying avrRpt2. The second generator is infection of plants carrying a functional RPM1 allele with bacteria carrying avrB. An "or" gate is used to show that either way of eliciting signaling resulted in accumulation of SA. The NDR1 allele status was represented with a "switch". The ndr1-1 allele was pictured as analogous to an electrical switch where the contact is in an open position, impairing early SA accumulation.

The HR progresses more slowly in response to avrRpt2 than in response to avrB42526. Evidence has been presented, however, that differences between HRs are not merely differences in kinetics 27. As such, the avrRpt2-elicited HR and the avrB-elicited HR were pictured as two separate components. When assayed under standard conditions, ndr1-1 mutant plants did not show the avrRpt2-elicited HR. An "and" gate was used to show that both infection with bacteria carrying avrRpt2 and early accumulation of SA were needed to give the characteristic, "slow HR". By contrast, ndr1-1 mutant plants did show the HR to bacteria carrying avrB. A straight line that is not connected via a gate to SA accumulation thus leads to "Fast HR-associated signals".

SAR is gated by an "or" gate with three possible inputs. As DC3000•avrB elicited SAR on the ndr1-1/npr1-2 double mutant, neither of these gene products was essential for SAR induction. The early SA accumulation was not likely to have been essential either, because that was markedly impaired by the ndr1-1 mutation. However, the characteristic avrB-elicited, "fast HR" did occur in the double mutant. As such, a straight line deriving from "Fast HR-associated signals" is one of the inputs to this "or" gate. Another input is an "and" gate. The evidence requiring this gate is that SAR elicited by DC3000•avrRpt2 was prevented by either a npr1 mutation 9 or a ndr1 mutation 4. We conclude that the characteristic avrRpt2-elicited "slow HR" can only contribute to SAR in combination with NPR1-dependent, early SA-elicited signaling, necessitating the "and" gate. The final input is SA-elicited, NPR1-dependent signaling acting alone 10.

PR-1 induction is also gated by an "or" gate with three alternative inputs. The "Fast HR" alone (this study), the "slow HR" alone 11, or SA-elicited, NPR1-dependent signaling 91011 were each sufficient by themselves to give PR-1 induction. Induction of PR-5 gene expression can also be via any of these three inputs acting independently. However, we (this study) and others 1920 have presented evidence that other pathways can also contribute to PR-5 induction. As such, PR-5 is gated by an "or" gate. One input is a branch point from the "or" gate leading to PR-1 induction. The second input is pictured as a question mark. Evidence has been presented herein that this second input represents at least two pathways. However, the details of these pathways are not known and are thus not pictured.

No attempt has been made to represent the kinetics of signaling events in Figure 4. Boolean formalism can incorporate kinetic information through use of gating delays. Differential equations-based kinetic modeling is an alternative means of capturing the rich dynamics of signaling networks 2829. As time-resolved data sets become available and feedback loops are delineated, it may be necessary to use these methods to understand system dynamics and control 30. However, the authors agree with Genoud et. al. (2001) that the scheme used in that work and in Figure 4 of this paper has advantages for representing the topology of signaling networks. This formalism allows clear differentiation of additive from obligatorily synergistic signaling interactions through use of "or" and "and" gates, respectively.

In conclusion, the additive effects of two pathways explain most if not all induction of SAR and PR-1 gene expression by the avirulent bacteria examined. One of these pathways stems from signals associated with hypersensitive cell death and is NPR1-independent. The other pathway stems from SA accumulation and signals through NPR1. By contrast, PR-5 induction by avirulent bacteria involves at least two additional pathways that also act in an additive fashion.

CZ isolated the ndr1-1/npr1-2 line used, established homogeneity of both mutations, carried out the PR gene expression studies, and contributed to preparation of the manuscript. AS conceived of the study, performed the genetic cross to make the ndr1-1/npr1-2 line, carried out the SAR experiments, developed the Boolean model of the signaling pathway, and contributed to preparation of the manuscript. Both authors read and approved the final manuscript.