Search PMN  


PDF version
for printing

Peer Reviewed

© 2012 Plant Management Network.
Accepted for publication 5 November 2012. Published 26 November 2012.

Spatial Distribution of Crown Gall in a Commercial Nursery of Weeping Fig

W. W. Turechek, USDA-ARS U.S. Horticultural Research Laboratory, Fort Pierce, FL 34945

Corresponding author: W. W. Turechek.

Turechek, W.W. 2012. Spatial distribution of crown gall in a commercial nursery of weeping fig. Online. Plant Health Progress doi:10.1094/PHP-2012-1126-01-RS.


Agrobacterium larrymoorei causes tumors on weeping fig. The association between propagation and pathogen spread in mother trees and daughter branches was studied in a commercial nursery. The mother tree planting was scouted for tumors prior to and after propagation. Branches selected for propagation were tagged to track disease development. The spatial distribution of crown gall in the mother tree planting was characterized with runs, join-count, and spatial autocorrelation analyses. The association of disease in mother trees and daughter branches was characterized with cross-correlation analysis. The incidence of crown gall in the mother tree planting increased from 7% prior to propagation to 32% eight months after propagation. Of the 4193 daughter branches monitored, 3.8% developed tumors. Runs analysis indicated significant clustering of diseased mother trees. Significant cross-correlations between mother trees and daughter branches with symptoms of crown gall were detected out to a distance of two plants from the source. Although pruning shears were routinely soaked in a disinfectant in this nursery, the degree of sterilization achieved apparently was not sufficient to prevent pathogen transmission. This study suggests that alternative sanitation measures should be sought and that infected mother trees and their neighbors should be avoided for propagation.


Crown gall on weeping fig (Ficus benjamina L.) has been a problem in commercial production of F. benjamina since 1991 (1). The disease is characterized by tumors on the main trunk of the plant as well as aerial tumors on branches (Figs. 1A, B). The bacterial pathogen causing crown gall of Ficus spp. is likely endemic to South Florida. Crown gall was first reported on the roots of Ficus aurea in the Florida Everglades in 1919 (8). Aerial tumors appear to be associated with pruning wounds that are made during propagation and general plant maintenance (1). The pathogen is thought to have been introduced into Florida’s commercial braided Ficus production after the freeze of 1989 when aerial tumors were found on 1-year-old trees produced from cuttings obtained from old hedges in South Florida (1). It was later discovered that bacteria isolated from aerial tumors were distinct from previously identified strains of Agrobacterium based on their unique opine metabolism and 16S rRNA sequence differences (1). In 1995, the tumorigenic Agrobacterium strain isolated from F. benjamina was identified as a new species and was named Agrobacterium larrymoorei sp. nov. (2).

The braided Ficus plant that is commonly encountered in retail stores is produced by first creating air layers on branches of similar size on mother trees of mature F. benjamina (Fig. 1C). Air layering is a propagation technique used to induce root formation on a branch while it is still attached to the parent plant, and is used on plant species that are difficult to root. To create an air layer, a 4-cm ring of bark and cambium layer is removed from a pencil-sized diameter branch approximately 40 cm from the distal end of a branch. Alternatively, a 3- to 5-cm cut can be made at an angle into about the center of the branch, and a small sliver of wood (e.g., a piece of toothpick) is inserted into the wound to hold it open and prevent it from healing. Next, a clump of dampened sphagnum moss is wrapped around the branch at the site of the wound and aluminum foil is wrapped around the clump of moss to hold it securely in place and prevent moisture from escaping. The foil is secured to the branch either by twisting it around the branch protruding from either side of the clump of moss or with twist-ties. Once the roots have formed − about 3 to 4 weeks after creating the air layer − the daughter branches are pruned from the parent plant just below the root ball with sterilized shears and 3 to 4 similarly sized daughter branches (not necessarily originating from the same mother tree) are braided, potted in sterile potting media, and then placed in nursery beds for 3 to 4 weeks to allow the plants to fuse and heal. Wounds to the mother tree and the newly formed daughters created during the process of air layering, as well as wounds produced when braiding the daughter branches, are typical sites of tumor formation (22).


Fig. 1. (A) Tumors caused by Agrobacterium larrymoorei sp. nov. on the main trunk of weeping fig (Ficus benjamina L. cv. Wintergreen) as well as (B) aerial and crown tumors on daughter branches. (C) Example of air-layers (wrapped in blue foil) on weeping fig. (D) Example of air layers that were numbered and tagged with weather-resistant aluminum tags as discussed in the text. (E) Braided plants that were placed on ground-cloth-covered plots to allow the branches to fuse and heal and form a single braided plant.


Having knowledge specific to the epidemiology of crown gall on Ficus is useful when developing disease management strategies. This knowledge can provide insight into which production practices might be effectively altered to disrupt disease development. For Ficus, the extent to which crown gall is transmitted in the mother planting during the process of propagation and the extent to which crown gall is passed to the daughter plants are not well understood. The objectives of this project were to: (i) determine the effect of propagation on transmission of crown gall within the mother planting; and (ii) determine the likelihood that air-layered daughter branches originating from symptomatic mother trees result in infected daughter plants.

Disease Monitoring and Data Collection

The study was conducted in a commercial nursery located in St. Lucie Co., FL. The mother tree planting of F. benjamina cv. Wintergreen consisted of 16 evenly spaced rows of 78 plants and was initially surveyed for tumors by nursery personnel just prior to the beginning of propagation in 2007. This was followed by a confirmatory scouting by laboratory personnel of the author during the process of tagging branches selected for air layering (discussed below). Three periods of propagation occurred within this field which coincided with orders placed at the nursery (Table 1). Beginning with the first row of the mother tree planting and winding through the rows in a serpentine pattern, employees of the nursery inspected each tree of the planting and selected branches for air layering based on a combination of vigor, length, and circumference of the branch until the necessary number of branches was chosen to meet the order. The branches were selected from mother trees with no visible symptoms of crown gall. Subsequent periods of propagation began in the row where the last propagation event ended, except that row 7 was utilized in both the 1st and 2nd times. Within 3 days of the start of propagation, the air layers were numbered and tagged with weather-resistant aluminum tags (Al Tag, Forestry Suppliers Inc., Jackson, MS) for future tracking (Fig. 1D). After three to four weeks, air-layered branches that had formed roots, herein referred to as daughter branches, were cut from the mother tree, braided with 2 to 4 other daughter branches, planted in pots containing sterile potting media, and then placed on ground-cloth-covered plots to allow the branches to fuse and heal and form a single braided plant (Fig. 1E). This process was completed by employees of the nursery.

Tumor formation was tracked on all tagged daughter branches (3 to 4 per pot) at approximate monthly intervals. With the braided plants, it was sometimes difficult to associate a tumor to a single daughter branch because some tumors formed evenly at the center of the interface where branches intertwined. In these cases, the tumor was assigned arbitrarily to a particular daughter branch, but often this was inconsequential because for any single braided plant the daughter branches were routinely pruned from the same mother plant. In 2008, a final survey of the mother tree planting was conducted by lab personnel approximately 6 to 8 months after the last propagation period concluded (Table 1). The lag time between the last propagation period and the final disease rating was to allow sufficient time for tumor development so they could be more easily identified within the dense foliage. The typical incubation period for crown gall diseases (i.e., the time between inoculation and symptom development) is 2 to 4 weeks.

Table 1. Dates of the propagation activities conducted in a commercial planting of weeping fig (Ficus benjamina L. cv. Wintergreen).

Air layers tagged /
1st mother tree rating
9-10 Oct 2007 4 Dec 2007 10 Jan 2008
Air layers cut & nursery plot establishedy  19 Oct 2007 5 Jan 2008 18 Jan 2008
Final nursery
plot rating
15 Feb 2008 8 May 2008 8 May 2008
2nd mother tree
6-31 Jun 2008 8-30 Aug 2008 19-26 Sep 2008

 x Air layering and subsequent production of daughter plants was done in three distinct time periods: Period I utilized rows 1-7 of the mother tree planting; Period II utilized rows 7 to 12; and Period III utilized rows 13 to 16. Mother tree rows selected in each period of propagation typically began where propagation ended in the previous period except for row 7 which was utilized in both Periods I and II.

 y The nursery plot was a ground-cloth covered area where potted, braided Ficus plants were placed to be managed while they fused and healed.

Distribution and Incidence of Crown Gall

The distribution of mother trees with symptoms of crown gall and the number of daughter branches from each mother tree that developed tumors in their nursery plots are shown in Figure 2. A total of 1129, 1805, and 1259 daughter branches were collected and monitored for symptoms of crown gall from the first, second, and third period of propagation, respectively. The minimum, mean, and maximum number of air layers produced per row in each of the three periods of propagation was, respectively: 78, 161.3, and 397 in Period I; 62, 300.8, and 383 in Period II; and 241, 314.8, and 372 in Period III. Air-layers were not prepared on all mother trees (shown as an "×" in Figure 2).


Fig. 2. Distribution of mother trees (MT) with symptoms of crown gall in a commercial planting of weeping fig (Ficus benjamina L. cv. Wintergreen) and the number of daughter branches (DB) sourced from the mother tree that developed one or more tumors (triangles) in their nursery plots. Mother trees were inspected for tumors prior to propagation (open circles) and again after propagation (closed circles); dates of activities are listed in Table 1. Daughter branches were inspected for tumors in their nursery plots about 4 months after propagation from the dates listed in Table 1.


The percentage of mother trees with crown gall in the mother tree planting increased from the initial estimate of 7% prior to the start of the three periods of propagation to nearly 32% eight months after the final period of propagation (Table 2). The final percentage of trees with one or more tumors ranged from 11.5 to 50% among rows. Of the 4193 daughter branches that were monitored, 3.8% developed tumors (Table 2). The correlation between the percentage of mother trees within a row with symptoms of crown gall and the percentage of daughter branches originating from the same row with symptoms crown gall was 0.496 (P = 0.051) according to Spearman’s rank correlation coefficient, indicating that rows with a high percentage of symptomatic mother trees produced a proportionately higher number of symptomatic daughter branches.

Table 2. Number and the corresponding proportion of mother trees per row with one or more tumors observed in the 2007 (pre-propagation), 2008 (survey of the number of new tumors), and final surveys (a combination of the 2007 and 2008 data) of a mother tree planting of weeping fig (Ficus benjamina L. cv. Wintergreen trees) located in St. Lucie Co., FL; and the number and corresponding incidence of daughter branches with one or more tumors arranged by the row from which they were sourced.

Row  Mother tree plantingx  Nursery plantingy
N 2007 2008 Total p n x p
1 78 5 4 9 0.115 145 6 0.041
2 78 8 18 26 0.333 100 6 0.060
3 78 3 24 27 0.346 78 12 0.154
4 78 7 14 21 0.269 132 2 0.015
5 78 4 21 25 0.321 183 10 0.055
6 78 3 19 22 0.282 94 4 0.043
7 78 4 17 21 0.269 459 16 0.035
8 78 3 20 23 0.295 368 17 0.046
9 78 0 24 24 0.308 359 7 0.020
10 78 6 33 39 0.500 377 26 0.069
11 78 10 17 27 0.346 256 10 0.039
12 78 2 30 32 0.410 383 25 0.065
13 78 6 10 16 0.205 241 2 0.008
14 78 4 21 25 0.321 323 9 0.028
15 78 16 21 37 0.474 323 4 0.012
16 78 5 17 22 0.282 372 3 0.008
Sum 1248 86 310 396 4193 159
Incidence 0.069 0.248 0.317 0.038

 x N = the number of mother trees in the indicated row. The number of mother trees with one or more tumors found in the 2007 and 2008 surveys is shown in the columns labeled as such, and the total number of tumors is the sum of the number of tumors found in both surveys; p = incidence of the total number of tumors.

 y n = the number of daughter branches removed from the indicated row; x = the number of daughter branches that developed tumors and the corresponding incidence, p.

Spatial Analyses of the Crown Gall Epidemic

Mother tree planting. To assess the spatial pattern of symptomatic mother trees (i.e., trees showing tumors on their trunks or branches) ordinary runs, join-count analyses, and spatial autocorrelation analyses were conducted. Ordinary runs analysis is used to characterize patterns in transect data by comparing the number of observed runs to the number of runs expected if disease were distributed randomly, where a run is a group of consecutive diseased or healthy trees (10,20). To perform the analysis, the rows were joined together in a serpentine pattern to mimic the pattern of movement of the workers (i.e., the abutting ends of rows were joined) to create a single transect of n = 1248 trees. The pattern of symptomatic mother trees was evaluated for the 2007 survey data, the number of new trees with tumors found in the 2008 survey, and the total number of symptomatic trees found in the planting at the end of the 2008 survey (i.e., final survey data). Runs analyses were done using Minitab (ver. 15, State College, PA).

Join-count analysis is used to characterize pattern in an evenly spaced lattice by tallying the number of neighboring trees that are jointly symptomatic, referred to as a black-black or "BB" join, and the number of neighboring trees where only one of the neighbors is symptomatic, referred to as a black-white or "BW" join (5,19). Joins were analyzed separately in the rook, bishop, or queen orientations in analogy to the directions that these pieces can move in a game of chess to identify anisotropy (the property of being directionally dependent) for the 2007 and 2008 datasets, and the final survey results. Similar to a runs analysis, the number of observed BB and BW joins is compared to the number of joins expected if symptomatic trees were distributed randomly in the planting. Join-count analysis was done using the ROOKCASE add-in for Excel (24).

Significant clustering of symptomatic mother trees was found in both the 2007 and 2008 (new tumors) survey data, and the 2008 final survey data according to ordinary runs analysis (Table 3). Results from the join-count analysis differed slightly and showed an absence of clustering of symptomatic mother trees for any orientation in the 2007 survey of mother trees (Table 4). This changed for the 2008 survey as well as the final survey data where symptomatic mother trees were found to be clustered for all orientations, agreeing with the ordinary runs analysis (Table 4).

Table 3. Results of an ordinary runs analysis to assess the pattern of mother trees with symptoms of crown gall in a commercial mother tree planting of weeping fig (Ficus benjamina L. cv. Wintergreen) located in St. Lucie Co., FL, using survey data from 2007 (pre-propagation), 2008 (i.e., number of new tumors) and their combination (2007 and 2008 data).

Year n x Observed
no. of runs
no. of runs
Variance z p-value
2007 1248 86 151 161 20.4 -2.245 0.025
2008 1248 316 401 473 178.3 -5.390 0.000
Total 1248 402 463 546 237.8 -5.384 0.000

 x n = total number of mother trees; x = number of mother trees with one or more tumors; z = standard normal Z; p-value = probability of observing the absolute value of a standard normal variable greater than |z|.

Table 4. Join count analysis of the number of tumors found in the 2007 (pre-propagation), 2008 (i.e., number of new tumors) and final surveys (2007 and 2008 data) in a commercial mother tree planting of weeping fig (Ficus benjamina L. cv. Wintergreen) located in St. Lucie Co., FL.

Year TN x Join count analysis
JN BB E(BB) zBB p-val BW E(BW) zBW p-val
  Rook’s case
2007 1248 86 2402 17 11 1.82 0.069 298 308 -1.53 0.126
2008 1248 316 2402 203 154 5.27 0.000 832 909 -4.12 0.000
Total 1248 402 2402 318 249 6.37 0.000 934 1050 -5.39 0.000
  Bishop’s case
2007 1248 86 2310 9 11 -0.60 0.548 302 297 0.65 0.516
2008 1248 316 2310 177 148 3.08 0.002 858 874 -0.86 0.390
Total 1248 402 2310 299 239 5.35 0.000 934 1010 -3.54 0.000
  Queen’s case
2007 1248 86 4712 26 22 0.87 0.384 600 605 -0.43 0.667
2008 1248 316 4712 380 301 5.78 0.000 1690 1783 -3.44 0.000
Total 1248 402 4712 617 488 8.05 0.000 1868 2059 -6.26 0.000

 x TN = total number of mother trees; x = number of mother trees with one or more tumors; JN = number of possible joins for the rook’s, bishop’s, and queen’s orientations; BB (BW) = observed number of BB (BW) joins; E(BB/BW) = expected number of BB (BW) joins; zBB (zBW) = standard normal statistic under randomization hypothesis; p-val = probability of observing the absolute value of a standard normal variable greater than |z|.

Spatial autocorrelation analysis (SPA) was then conducted on the total number of symptomatic trees (the final survey data) to characterize the degree of correlation among mother trees located 1, 2, 3,…, n trees apart in the planting. Two different spatial orientations of the data were examined. First, the data was treated as a single transect, as above, where the individual rows in the mother tree planting were joined in a serpentine pattern to give a single transect of n = 1248 trees in length. Analysis of the transect orientation was done to characterize the within-row transmission properties, similar to the rooks orientation of the join-count analysis. In addition to analyzing the full set of rows, the data was split into three different sets to include only the rows specific to that period of propagation. For each of the three periods, the rows were joined together in a serpentine pattern to create three single transects of ni trees, where i = 1st, 2nd, of 3rd period of propagation. Analysis by the period of propagation was done to isolate any effects that might have been associated with the particular time of propagation. Second, a radial analysis was considered for each of the data sets described above where correlations were calculated between trees located within the circumference of a circle with a radius of 1, 2, 3,…, n trees apart. This orientation is similar to the queen’s orientation of the join-count analysis.

For all the correlation analyses, approximate 95% confidence intervals around the correlation coefficients were calculated as: ±2/√n [(14) p. 83]. Correlations that exceeded these values were considered significant at α = 0.05 level. SPA was done using the ROOKCASE add-in for Excel (24). Although Moran’s I is considered a correlation coefficient, it is important to note that the expected value of this coefficient under no spatial autocorrelation is not zero but −1/(n −1) and that it is possible for Moran’s I to exceed one, particularly with empirical data (26).

Significant spatial autocorrelation of symptomatic mother trees was found in the 2008 survey and the final survey data, but not the 2007 survey. Treating the planting as a single transect of n = 1248 trees, significant positive autocorrelations were detected extending out to three trees for the 2008 survey (data not shown) and out to 6 trees for the combined survey data (Fig. 3A). Considering the individual periods of propagation, there was an increase in positive correlations with each succeeding time of propagation (Figs. 3B-D). Significant positive autocorrelation was most prevalent in the second period of propagation. For the radial analysis, spatial autocorrelation extended out to the neighboring tree for the 2008 survey (data not shown) and the final survey data (Fig. 3E). Significant correlations in disease status of neighboring plants were present in the first and second periods of propagation, but not the third (Figs. 3F-H).


Fig. 3. Autocorrelation coefficients for the transect analysis (A-D) and radial analysis (E-H) of mother trees with symptoms of crown gall in the mother tree planting for each of the three periods of propagation as well as the entire planting. Data are the sum of 2007 and 2008 surveys. The spatial lag represents mother trees that are 1, 2, 3,…, n trees apart. The broken lines represent the 95% confidence bands.


Daughter tree planting. Cross-correlation analyses were performed to determine the correlation between the disease status of daughter branches ~120 days after being cut from the mother tree and the disease status of the mother tree from which the branch was cut, as well as the disease status of mother trees located within 1, 2,…, n trees apart from their mother tree of origin. This was done only for the rook transect orientation. The disease status assigned to mother trees in the final survey data was used for the analysis. Confidence intervals were constructed for the cross-correlation coefficients using the approximation for the autocorrelation coefficient described above. Cross-correlation analyses were done using Minitab (ver. 15, State College, PA).

There were significant positive correlations between symptomatic mother trees and daughter branches propagated directly from symptomatic mother trees as well as from trees neighboring a symptomatic mother tree up to a distance of two trees in either direction from the symptomatic mother tree (Fig. 4). Relative to the timing of propagation, there was a directional effect during the first period of propagation that was not evident during the second period of propagation (Fig. 4). Significant correlations in the third period of propagation were sporadic and did not follow the same patterns as the first two periods (Fig. 4).


Fig. 4. Cross-correlation coefficients for the number of daughter branches with symptoms of crown gall originating from mother trees n lag distances apart for each plot. The spatial lag represents mother trees that are 1, 2, 3,…,n trees apart and is bi-directional. The broken lines represent the 95% confidence bands.



The distribution of tumors in the mother tree planting of weeping fig exhibited patterns of disease that was characteristic of close-proximity transmission. Specifically, the increase in the number of tumors found in the 2008 survey coupled with their distribution exhibiting low-order spatial autocorrelation is evidence of transmission due to cultural activities such as propagation and pruning. Dissemination of crown gall bacteria in cuttings or through the process of pruning has been shown in blackberry (9) and grape (3). It is possible, however, that tumor formation was not the result of direct transmission of the bacteria to the wounding site on contaminated pruning shears, but rather a response by the bacteria to induce tumors at the wounding sites in plants that have been systemically invaded by the bacteria previously (7). Agrobacterium has been shown to move systemically in several plant species resulting in the production of secondary or aerial tumors in rose, grapevine, chrysanthemum, cherry and other plants (4,6,7,21). Indeed this manner of disease development may account for some of the tumor formation seen in F. benjamina, and the observed correlations could arise under this scenario if the bacteria were transferred to neighboring plants through non-wounding activities such as root grafts which occur between plants in close proximity.

A significant correlation between symptomatic daughter branches and their mother tree source was also found through cross-correlation analysis. Specifically, daughter branches originating from mother trees with symptoms of crown gall or from mother trees that were neighboring symptomatic mother trees were more likely to develop tumors than daughter branches originating from mother trees located two or more trees away from a symptomatic mother tree. It is not surprising that a correlation exists between the incidence of crown gall on daughter branches and the disease status of the mother tree from which they came − diseased parents tend to produce diseased offspring. Correlations in disease status of neighboring plants were most likely the result of pathogen transmission on contaminated pruning shears or propagating on latently infected trees, where infection of the mother tree may have occurred prior to (e.g., root grafts) or at the time of propagation (e.g., when the air-layer was first established). In the latter case, it would be conceivable that both the mother tree and the eventual daughter branch(es) were infected simultaneously. Or, the correlations could have arisen because trees with established tumors simply escaped detection (sampling error; workers would not intentionally propagate from mother trees known to be infected). This is no different than propagating directly from an infected mother tree except that the mother tree was erroneously classified disease free. However, the statistical analysis cannot be used to distinguish between daughter branches originating from latently infected mother trees from trees that simply escaped detection.

In trees exhibiting symptoms, it is not clear how the bacteria are distributed within the plant. For example, in a tree with a single tumor it is not known how many (if any) additional branches are latently infected. The systemic movement of Agrobacterium spp. from a single point of inoculation to distal parts of the plant has been demonstrated for other plant species, but the extent of migration is variable and depends upon on a number of factors including plant species, bacterial strain, and the time elapsed since inoculation (6).

In considering standard management practices along with the results shown here, it is advisable to avoid propagating from any tree showing symptoms. Results shown here also suggest that it would be reasonable to avoid propagating from trees neighboring symptomatic trees even if the tree appears tumor free. In the absence of a rigorous sampling protocol, finding tumors can be very difficult, particularly young and/or small tumors − which may appear in as little as 3 to 4 weeks on daughter branches after being cut from the mother tree. Thus the cautious nurseryman may wish to avoid propagating from these trees since neighboring trees are more likely to develop tumors.

Workers are taught to adhere to good sanitation practices, including sterilization of tools and pruning shears used during propagation and pruning because it is known that several pathogens, particularly bacteria (3,11,15,17) and viruses (12,13,16,18,23,25), are spread on contaminated tools during propagation and pruning. In this commercial nursery, pruning shears were routinely soaked in a quaternary ammonium-based disinfectant. It could be argued that the degree of sterilization was insufficient to substantially reduce the spread of bacterium. To pinpoint the exact reason(s) why sterilization was ineffective is a challenge. Some possibilities include: workers did not adhere to the sanitation protocol; the disinfectant was ineffective against Agrobacterium; and the tools were insufficiently sterilized either because the bacteria were shielded from exposure [see ref. (15)] or were exposed to the disinfectant for an insufficient period of time. Workers were also told not to propagate from trees that contained tumors. Indeed, the 7% of mother trees that were initially identified with tumors in the 2007 survey were not utilized in propagation. Thus, the 25% increase of symptomatic mother trees and the 3.8% of daughter branches that developed crown gall probably resulted from a mixture of propagation on latently infected trees, the use of contaminated tools during the establishment of air layers (as a result of working on asymptomatic trees or one where the infection was simply missed), and/or the use of contaminated pruning shears when daughter branches were pruned from the mother tree.

Correlation analyses were used to analyze the data in this study. It should be noted that correlation does not infer causation and speculation on the possibilities of why symptomatic mother trees tended to be neighbors or why daughter branches were more likely to develop crown gall if their mother was symptomatic were addressed in the preceding paragraphs. Although significant correlations were discovered in all analyses, their values were generally small lying within the range of 0.1 and 0.27. It might be easy to dismiss the low correlations as trivial except one must consider that the study was done in a commercial nursery. In a commercial setting, environmental conditions fluctuate and are not continuously optimal or even favorable for pathogen establishment / infection and/or disease development through the course of the season. Additionally, the pathogen is unlikely to be evenly distributed through the planting and a plant’s susceptibility to infection may change through the course of the season. Moreover, a quaternary ammonium-based disinfectant was used during propagation which presumably had to reduce bacterial populations. These factors, along with factors mentioned in the previous paragraphs, could perceivably reduce the correlations that might have been observed if a designed experiment were conducted under controlled conditions.

The use of runs, join-count, and spatial autocorrelation analyses (SPA) could be considered redundant with SPA being the choice method because it yields similar information as runs and join-count analyses. In addition, SPA provides detailed information on the strength of correlations as a function of distance allowing inferences to be made about the scale of clustering. However, the statistical assumptions are more rigorous for SPA than either runs or join count analyses and, consequently, there is a greater chance for reporting spurious correlations if assumptions are violated. For example, the data are inherently binary and it is only through the use of larger sample sizes that violation of the normality assumption is disregarded. Results are also dependent on the spatial weights matrix. A binary weighting matrix was chosen for this analysis, thus correlations were calculated identically for all pairs and were not weighted, for example, based on their distance apart.

In conclusion, a small but significant correlation existed between a mother trees chance of developing crown gall and its proximity to an infected mother tree. A small but significant correlation also existed between a daughter branches chance of developing crown gall and the infection status of the mother tree from which it was cut as well as the infection status of the mother tree’s neighbors. Yet despite the low correlation, the impact is significant. A single mother tree − which can produce numerous daughter branches (Fig. 1C) − is removed from propagation if a single tumor is discovered on it. Moreover, if growers were to adhere to the suggestion of avoiding trees that neighbor symptomatic mother trees, the eligible pool of usable trees is quickly reduced as two additional trees would be considered unusable with every single tree classified as infected. Unfortunately, because the time between periods of propagation is short and the latent period is relatively long (or certainly the time before symptoms become readily obvious is long) and spotting tumors within dense foliage is difficult, it is likely that propagation will involve infected mother trees and this hinders the possibility of slowing the rate of epidemic development. Unless serious efforts are made to identify infected trees (which may be prohibitively expensive) and/or a better sanitation practices are implemented, the likelihood of high levels of infection will continue.


I would like to thank Gregory Hess and Linda Rios for their technical assistance and thoughtful input. I would also like to thank John Martin for access to his nursery and for his informative discussions about the propagation, management, and business of Ficus production. Financial support was provided by USDA-ARS CRIS project 6618-22000-026-00D. The use of trade, firm or corporation names in this publication is for the information and convenience of the reader. Such use does not constitute an official endorsement or approval by the United States Department of Agriculture or the Agricultural Research Service of any product or service to the exclusion of others that may be suitable.

Literature Cited

1. Bouzar, H., Chilton, W. S., Nesme, X., Dessaux, Y., Vaudequin, V., Petit, A., Jones, J. B., and Hodge, N. C. 1995. A new Agrobacterium strain isolated from aerial tumors on Ficus benjamina L. Appl. Environ. Microbiol. 61:65-73.

2. Bouzar, H., and Jones, J. B. 2001. Agrobacterium larrymoorei sp. nov., a pathogen isolated from aerial tumours of Ficus benjamina. Int. J. Syst. Evol. Microbiol. 51:1023-1026.

3. Burr, T. J., and Katz, B. H. 1984 Grapevine cuttings as potential sites of survival and means of dissemination of Agrobacterium tumefaciens. Plant Dis. 68:976-978.

4. Burr, T. J., and Otten, L. 1999. Crown gall of grape: Biology and disease management. Annu. Rev. Phytopathol. 37:53-80.

5. Cliff, A. D., and Ord, J. K. 1981. Spatial Processes: Models and Applications. Pion Ltd., London, UK.

6. Cubero, J., Lastra, B., Salcedo, C. I., Piquer, J., and López, M. M. 2006. Systemic movement of Agrobacterium tumefaciens in several plant species. J. Appl. Microbiol. 101:412-421.

7. Escobar, M. A., and Dandekar, A. M. 2003. Agrobacterium tumefaciens as an agent of disease. Trends Plant Sci. 8:380-386.

8. Galloway, B. T. 1919. Giant crown galls from the Florida Everglades. Phytopathology 9:207-208.

9. Garrett, C. M. E. 1978 Crown gall of blackberry: Field spread and susceptibility to disease. Plant Pathol. 27:182-186.

10. Gibbons, J. D. 1985. Nonparametric Methods for Quantitative Analysis, 2nd Edn. American Sciences Press Inc., Columbus, OH.

11. Goodman, C. A., and Hattingh, M. J. 1988. Mechanical transmission of Xanthomonas campestris pv. pruni in plum nursery trees. Plant Dis. 72:6 3.

12. Hadidi, A., Giunchedi, L., Shamloul, A.M., Poggi-Pollini, C., and Amer M.A. 1997. Occurrence of peach latent mosaic viroid in stone fruits and its transmission with contaminated blades. Plant Dis. 81:154-158.

13. Kamenova, I., and Adkins, S. 2004. Transmission, in planta distribution, and management of Hibiscus latent Fort Pierce virus, a novel tobamovirus isolated from Florida hibiscus. Plant Dis. 88:674-679.

14. Kendall, M. G., and Ord, J. K. 1990. Time Series, 3rd Edn. Oxford Univ. Press, New York, NY.

15. Kleinhempel, H., Nachtigall, M., Ficke, W., and Ehrig, F. 1987. Disinfection of pruning shears for the prevention of the fire blight transmission. Acta Hortic. 217:211-218.

16. Kyriakou, A. 1992. Incidence in Cyprus of citrus exocortis viroid and its mechanical transmission. Plant Pathol. 41:20-24.

17. Lecomte, P. Risk of fire blight infection associated with pruning of pear trees. Acta Hortic. 273:83-90.

18. Lewandowski, D. J., Haynes, A. J., and Adkins, S. 2010. Surprising results from a search for effective disinfectants for Tobacco mosaic virus-contaminated tools. Plant Dis. 94:542-550.

19. Madden, L. V., Hughes, G., and van den Bosch, F. 2007. The Study of Plant Disease Epidemics. American Phytopathological Society, St. Paul, MN.

20. Madden, L. V., Louie, R., Abt, J. J., and Knoke, J. K. 1982. Evaluation of tests for randomness of infected plants. Phytopathology 72:195-198.

21. Martí, R., Cubero, J., Daza, A., Piquer, J., Salcedo, C. I., Morente, C., and López, M. M. 1999. Evidence of migration and endophytic presence of Agrobacterium tumefaciens in rose plants. Eur. J. Plant Pathol. 105:39-50.

22. Moore, L. W., Bouzar, H., and Burr, T. J. 2001. Agrobacterium. Pages 17-35 in: Laboratory Guide for Identification of Plant Pathogenic Bacteria, 3rd Edn. N. W. Schaad, J. B. Jones, and W. Chun, eds. American Phytopathological Society, St. Paul, MN.

23. Pategas, K. G., Schuerger, A. C., and Wetter, C. 1989. Management of tomato mosaic virus in hydroponically grown pepper (Capsicum annum). Plant Dis. 73:570-573.

24. Sawada, M. 1999. Rookcase: An Excel 97/2000 Visual Basic (VB) add-in for exploring global and local spatial autocorrelation. Bull. Ecol. Soc. Am. 80:231-234.

25. Wintermantel, W. M. 2011. A comparison of disinfectants to prevent spread of potyviruses in greenhouse tomato production. Online. Plant Health Progress doi:10.1094/PHP-2011-0221-01-RS.

26. Wong, D. W. S., and Lee, J. 2005. Statistical Analysis of Geographic Information with ArcView GIS and ArcGIS. John Wiley & Sons Inc., Hoboken, NJ.