Three-dimensional confocal images are suitable for the analysis of nuclear morphology
Our nuclear staining method was based on a protocol developed for analogous analyses of nuclear morphology in nematode-infected roots of Arabidopsis thaliana [22, 26] and modified according to Carotenuto et al. (2019) - to analyse M. truncatula root tissues engaged in AM. In particular, the use of DAPI staining on 100-μm thick Vibratome sections had previously proven successful in preserving root nuclei [22] and allowing detailed measurements of their areas, volumes and spatial distribution. In our samples (Fig. 1), plant nuclei were marked very brightly, with a high signal/noise ratio and weak non-specific labelling of cell walls. Importantly, DAPI also stained Gigaspora margarita nuclei (easy to discriminate from the plant nuclei based on their small size) in intraradical and extraradical hyphae.
In order to validate the results obtained through automated image analysis, the size of equatorial nuclear sections was first measured manually, as described in Carotenuto et al. (2019), by outlining each DAPI-stained nucleus in single optical sections from six z-stacks recorded in uninoculated and colonized roots. We measured an amount of 1020 and 1150 nuclei in uninoculated and colonized root sections respectively (Additional files 1, 9). Quantitative analysis showed that the average area of equatorial nuclear sections was significantly smaller (independent sample’s median test, P < 0.05) in uninoculated (32.7 ± 0.33 μm2) compared to mycorrhizal root segments (35.9 ± 0.48 μm2). In addition, the curve distribution of nuclear areas (Fig. 2a; Additional file 1) showed significantly higher values of skewness (P < 0.001) and kurtosis (P < 0.01) in mycorrhizal (ske = 1.14 ± 0.16; kur = 1.52 ± 0.49) compared to uninoculated root segments (ske = 0.74 ± 0.20; kur = 0.75 ± 0.56), due to the wider range of values observed in the former treatment (15–142 μm2) compared to the uninoculated one (15–62 μm2).
In conclusion, the manual measurement of nuclear cross sections in z-stacks from chemically fixed and DAPI-stained root segments highlighted a marked increase in nuclear size upon AM colonization, in line with our previous studies [19]. The image data set was therefore considered suitable for determining nuclear morphology by automated image analysis.
TrackMate analysis of nuclear diameters proved limitations in estimation of cross section areas and volumes
TrackMate is a plugin originally designed to track cytoskeletal elements within time-lapse imaging of living cells [27]. In order to apply this software to our dataset, we chose to swap the time-axis with the z-axis, therefore forcing TrackMate to trace the selected objects (optical sections of DAPI-stained nuclei) across a confocal z-stack. The TrackMate detector automatically identified nuclear sections with coincident centroids, tagging them with a univocal ID and detecting their diameters, which were then used to calculate the cross-section area of individual nuclei. The resulting frequency distribution of nuclear cross-section areas showed an overall range increase between uninoculated (15–114 μm2) and mycorrhizal root sections (15–149 μm2), although no statistically significant difference between skewness and kurtosis values of the two curves was observed (Fig. 2c; Additional files 1, 9).
As a second step, assuming that nuclear sections could be approximated to circular objects, we used the TrackMate output data to estimate nuclear volumes based on confocal z-step and the diameters of all optical sections with the same ID tag (see Methods and Additional file 2).
Also in this case, the distribution curve of nuclear volumes (Fig. 3a; Additional file 1) showed no significant difference in skewness and kurtosis values between uninoculated (23–624 μm3) and mycorrhizal samples (23–891 μm3), although the occurring overall range increased.
Compared to manual measurements, our application of TrackMate to automatically detect and measure nuclear section diameters across z-stacks revealed critical limitations. In particular, our use of the TrackMate proved unreliable for nuclear size estimation from confocal z-stacks. Furthermore, the TrackMate analysis showed a low detection efficiency compared to the manual method: a marked number of cell nuclei remained undetected in both colonized (> 44%) and uninoculated (> 38%) roots.
A customized plugin (round surface detector) efficiently identifies and measures nuclear cross-section
As an alternative to the previous approach, we wrote an image analysis algorithm using the Fiji embedded macro language (see Methods; Additional file 3; Additional file 4). The resulting plugin allowed automated nuclear tagging and identification based on roundness and the size range of the detected objects.
The detection efficiency of this Round Surface Detector plugin (Additional files 1, 9) produced results comparable to that achieved by manual measurement, identifying 1001 and 1147 nuclei in uninoculated and colonized root segments, respectively. In contrast to the manual method, no significant increase in average nuclear area between uninoculated and mycorrhizal root sections was detected by the Round Surface Detector, although the distribution range of nuclear cross section areas was considerably larger in mycorrhizal (15–142 μm2) than uninoculated roots (15–68 μm2). However, the curve distribution of nuclear areas (Fig. 2b) showed significantly higher values of skewness (P < 0.001) and kurtosis (P < 0.01) in mycorrhizal (ske = 1.37 ± 0.20; kur = 2.83 ± 0.91) compared to uninoculated root segments (ske = 1.04 ± 0.10; kur = 1.62 ± 0.52).
Using 3D object counter for nuclear volume measurements
As the last method, we analysed our confocal datasets using 3D Object Counter, a Fiji plugin dedicated to 3D image analysis (see Methods; Additional file 5). This approach detected 594 and 893 nuclei in uninoculated and mycorrhizal roots, respectively, with a larger frequency distribution range (Additional files 1, 9) in mycorrhizal (20–220 μm3) than uninoculated roots (15–118 μm3). Compared to the manual approach, nuclear detection was less efficient (from 33% in colonized to 40% in uninoculated roots). This was probably caused by two problems within the algorithm. First, the selection of a single intensity threshold separating all nuclei from the background was difficult due to variations in staining intensity across the z-stack. Second, clustered nuclei - that were identified as a single lobed object - had to be manually deleted, thus reducing the total number of measurements.
Nevertheless, 3D Object Counter analysis highlighted a significant 1.25-fold increase (independent sample’s median test, P < 0.05) in average nuclear volume (Additional file 1) between uninoculated (40.3 ± 0.68 μm3) and mycorrhizal roots (50.7 ± 1.13 μm3), in line with manual measurements and previous studies [19]. The analysis of the distribution curve (Fig. 3b) confirmed this difference, showing significant higher values of skewness (P < 0.01) in mycorrhizal (1.47 ± 0.15) compared to uninoculated root segments (1.14 ± 0.22), whereas kurtosis value did not show any significant difference mainly because of the high variance.
In short 3D Object Counter resulted to be the most reliable method to automatically detect and measure nuclear size in mycorrhizal and uninoculated M. truncatula roots, representing a valid alternative to manual measurements. On this basis, we decided to apply this method to two specific and potentially challenging case studies.
Case study 1: analysis of nuclear morphology in two mutant phenotypes
Based on our previous tests, we decided to use 3D Object Counter to compare nuclear volumes in the cortical tissue of uninoculated and inoculated roots from wild type and two mutants of M. truncatula (dmi2–2 and dmi3–1). Such mutants are impaired for key CSSP genes: the receptor-like kinase DMI2 (for Does not Make Infection) and the nuclear calcium/calmodulin-dependent protein kinase (DMI3). Both strongly affect the development of functional AM symbiosis [28, 29], with an arrest of fungal colonization at the root epidermis, and a lack of cell cycle and endoreduplication reactivation in cortical cells [18, 19]. On this basis - and to prevent the abovementioned limitations of the 3D Object Counter algorithm - we restricted our measurements to the cortical tissue by cropping the original z-stacks (Additional files 6, 9).
A total of 440 and 690 nuclei were detected in uninoculated and mycorrhizal wild-type samples; 497 and 477 nuclei in uninoculated and inoculated dmi2–2; 623 and 511 nuclei, respectively, in dmi3–1 (Fig. 4). The analysis indicated an average nuclear volume of 58.6 μm3 for cortical cells in colonized wild-type roots. By contrast, uninoculated wild-type and both uninoculated and inoculated dmi2–2 and dmi3–1 mutants displayed nuclei of approximatively 44 μm3 (Fig. 4).
While demonstrating the efficiency of the 3D Object Counter approach, this significant difference has relevant biological implications. The enlargement of cortical cell nuclei, recorded through manual measurement, had previously been related to AM colonization [19]. The present results - based on the use of an automated image analysis method and intrinsically excluding any possible bias related to manual measuring - confirm that this is a hallmark of AM fungal accommodation and is dependent on the functionality of the CSSP where DMI2 and DMI3 play key roles.
Relationship between nuclear volume and endoreduplication
Increase in nuclear volume is the most obvious consequence of increments in ploidy [25, 30,31,32]. We therefore decided to check whether the population of nuclear volumes could be internally subdivided into a number of coherent classes, possibly reflecting progressive rounds of ploidy increase. To this aim, we applied the Sturges rule [33] to our dataset. The analysis, as reported in Additional file 7, indicated 25 μm3 as the optimal class width and identified four classes in the population of nuclear volumes from uninoculated wild-type roots, with average volumes of 30.3 μm3 for class I, 54.6 μm3 (class II), 77.6 μm3 (class III), 107 μm3 (class IV). By contrast, eight such classes were identified in mycorrhizal wild-type with average volumes of 29.6 μm3 (class I), 55.2 μm3 (class II), 81 μm3 (class III), 105.9 μm3 (class IV), 131.8 μm3 (class V), 155.3 μm3 (class VI), 181 μm3 (class VII) and 210.2 μm3 (class VIII).
Notably, the two populations of nuclear volumes – identified by applying descriptive statistics [33] to our dataset of measurements – matched with the corresponding ploidy levels and population of nuclear areas described by Carotenuto et al. (2019). On this basis, we used this clustering as an indication of putative ploidy levels, as shown in Additional file 8.
Case study 2: cortical cell division vs ploidy increase in AM
In addition to the onset of diffuse and recursive endoreduplication events [19], the triggering of ectopic cell divisions, leading to the appearance of split cells (Fig. 1), has also been described in the AM colonized area of the root [18]. Such split cells are easily recognized for their roughly square shape [18], which is maintained over time due to the limited capability of wall extension of the fully differentiated surrounding tissue.
Using the clustering of nuclear volumes into Sturges classes as an indication of putative ploidy levels, we focussed our analysis on all split cells found in our dataset of confocal z-stacks from uninoculated and mycorrhizal samples at progressive stages of fungal colonization (Fig. 5). Our aim here was to clarify how endoreduplication and cell division events combine during AM colonization. In fact, the progressive colonization of the root cortex from a single penetration point - characteristic of AM interactions [10] - allowed the observation of cell responses at different distances from intraradical hyphae, and provided a range of images depicting the whole progression of cell responses related to fungal accommodation.
Thus, nuclear volume measurements were extended to 10 uninoculated and 10 inoculated independent root sections of wild-type root organ culture (ROC) line (Fig. 5, Additional file 9). A total of 51 undivided and 8 split cortical cells were analyzed in uninoculated roots. The average nuclear volume in such undivided cells from uninoculated roots was 61 μm3 (Fig. 5a), corresponding to a putative 4C ploidy (Fig. 5b), as expected for this tissue, based on literature data [34, 35]. The few split cells observed in uninoculated samples displayed an average nuclear volume of about 36 μm3 (Fig. 5a, b), corresponding to putative 2C ploidy and suggesting that such cells originated from the division of a regular cortical cell.
In mycorrhizal M. truncatula roots we analyzed 129 undivided and 89 split cells (Fig. 5a). Among undivided cortical cells, 30 contained arbuscules (arbusculated), 35 were located next to an arbuscule (neighbouring) and 64 were at a distance of at least two cells from any arbuscule (far). Among split cells, we identified 27 arbusculated, 36 neighbouring and 26 far cells.
Far from the arbuscules (Fig. 5c), the situation was very similar to that of uninoculated roots: undivided far cells contained putative 4C nuclei (average nuclear volume of 61.1 μm3; Fig. 5a) while the nuclei of far split cells had an average volume of 38.9 μm3 (putative 2C ploidy).
By contrast, neighbouring cells displayed enlarged nuclei, with putative 16C nuclei (average of 110.1 μm3) in undivided cells and 8C nuclei (70.7 μm3 on average) in split cells (Fig. 5d). This suggested the occurrence of two endoreduplication rounds in each daughter cell.
Lastly, both undivided and split arbusculated cells (Fig. 5e-g) had nuclear volumes ranging between 8C and 128C (with an average nuclear volume of 157 μm3 for undivided and 143.7 μm3 for split cells; Fig. 5a). Remarkably, 13 cases (in a total of 81 z-stacks) were observed where only one of the split cells was hosting an arbuscule. Such images (exemplified in Fig. 5e, f) clearly show that the colonized cell hosts the largest nucleus. This provides convincing support to our current model of arbuscule accommodation (Carotenuto et al., 2019), where root colonization by AM fungi reactivates the cell cycle in the root cortex, initially leading to ectopic cell divisions and later to the onset of repeated endoreduplication rounds as cortical cells prepare to host an arbuscule.