Sources of inocula influence mycorrhizal colonization of plants in restoration projects: a meta-analysis

Inoculation may influence mycorrhizal colonization and provide benefits to plants in restoration projects. However, it is unclear whether inoculation has consistent effects across ecosystem types, if it has long-term effects on colonization, and whether sources of inocula differ in their effectiveness. To address these issues, we performed a meta-analysis of published restoration studies across a variety of ecosystems to examine the effects of mycorrhizal inoculation on mycorrhizal establishment and plant growth under field conditions. Although we included trials from a variety of geographic locations, disturbance types, and ecosystem types, the majority were based in temperate ecosystems in the Northern Hemisphere, and fewer trials were from tropical ecosystems. Across ecosystem types, we found that inoculation consistently increased the abundance of mycorrhizal fungi in degraded ecosystems, and thus improved the establishment of plants. These benefits did not significantly attenuate over time. Moreover, inocula from different sources varied in their effects on mycorrhizal colonization. Inocula sourced from reference ecosystems and inocula with specific fungal species yielded higher increases in mycorrhizal colonization than did inocula from commercial sources. These results suggest that inocula source matters, and that an initial investment into mycorrhizal inoculation could provide lasting benefits for facilitating the establishment of the below- and aboveground components of restored ecosystems.


Introduction
Individual studies have indicated that mycorrhizal fungi can improve the success of native plant establishment in restored ecosystems (e.g. Stahl et al. 1988;Sylvia et al. 1993;Allen et al. 2003a;Graham et al. 2013). Indeed, mycorrhizal fungi provide nutrients to plants, usually leading to increased photosynthetic rates and improved plant productivity (Mosse 1957;Allen et al. 1981;McGonigle & Fitter 1988;Lekberg & Koide 2005;Hoeksema et al. 2010). As other inhabitants rely on vegetation for food and cover, restoration of mycorrhizal fungi may have far-reaching benefits within the ecosystem. Nevertheless, little is known about the extent to which mycorrhizal fungi can be directly manipulated in restoration projects, and which sources of mycorrhizal inocula are most effective.
A relatively small portion of restoration projects have attempted to manipulate mycorrhizal fungi directly (Sylvia 1989;Greipsson 2010). Those who have addressed this issue typically apply inoculum (i.e. material containing spores or hyphal fragments) to soil in the field or to growth media in nursery pots prior to outplanting. The rationale for this practice is that the addition of mycorrhizal propagules should improve the likelihood that roots will encounter mycorrhizal spores or hyphae and become colonized by the fungus (Christensen & Allen 1980). Generally, plants that are more extensively colonized by arbuscular mycorrhizal (AM) fungi tend to grow better (Treseder 2013). In a recent meta-analysis, Piñeiro et al. (2013) reported that mycorrhizal inoculation was particularly effective in improving the survival and growth of planted seedlings in arid lands. Currently, it is unclear whether this technique is broadly successful across a range of ecosystem types.
Indeed, additions of mycorrhizal inoculum need not necessarily improve mycorrhizal colonization of plant roots in a given ecosystem. In some cases, mycorrhizal inoculation may have little effect on colonization because adequate inoculum is already naturally present in the ecosystem. The availability of mycorrhizal propagules is only one determinant of mycorrhizal abundance-a number of other factors may inhibit mycorrhizal fungi in ecosystems marked for restoration (Skujins & Allen 1986). For instance, harsh environmental conditions in degraded landscapes could reduce the viability of inoculum in the soil, or could inhibit the extension of hyphae from the roots of nursery-inoculated plants (Plett et al. 2015;van der Heijden et al. 2015). Thus, it is worth examining whether enhanced root colonization is sustained over the long term (i.e. months to years) in field-established plants.
In addition, the mycorrhizal species represented in a given inoculum may not always be beneficial for the revegetated plant community. In fact, many mycorrhizal species may exhibit host preferences, host selectivity, or some degree of host specificity (Helgason et al. 1998;Smith & Read 2008;Sánchez-Castro et al. 2012;van der Heijden et al. 2015), which may not allow them to form relationships with certain restored plants. Even species of arbuscular mycorrhizal fungi-which are thought to be relative generalists-can vary in the degree to which they colonize plant roots (Powell et al. 2009) and in their effects on plant growth (Hoeksema et al. 2010). Mycorrhizal fungi from neighboring reference ecosystems (i.e. ecosystems that exhibit characteristics intended of the restored ecosystem) may be better-suited for the local environment. Indeed, in a recent review of restoration studies from semiarid ecosystems, Barea et al. (2011) noted that inocula derived from exotic mycorrhizal fungi may not be as effective as inocula from indigenous fungi. Common sources of mycorrhizal inocula for restoration work include commercially available inocula, whole inocula derived from soil or roots collected from reference ecosystems, or species-specific inocula isolated in the laboratory. If we can identify sources of inocula that best improve mycorrhizal abundance, this information can be used by restoration ecologists to develop best practices.
Toward this end, we performed a meta-analysis of published restoration studies that examined effects of mycorrhizal inoculation on mycorrhizal establishment (as percent root length colonized, PRLC) and plant growth under field conditions. We tested the hypotheses that additions of mycorrhizal inocula would result in sustained increases in PRLC (hypothesis 1) and improved plant growth (hypothesis 2) in the field, and that whole inocula from neighboring reference ecosystems would elicit larger increases in PRLC than would commercial inocula (hypothesis 3).

Methods
We surveyed articles published in the literature and found 28 manipulative field-based restoration trials from 22 publications that addressed the influence of mycorrhizal fungal inoculation on percent colonization of plant roots. We used the Google Scholar (scholar.google.com) and Web of Science (webofknowledge.com) search engines, and also directly searched the archives of the Restoration Ecology journal (link.springer.com/journal/572). Our search terms were (restor* and mycorrhiz*), (restor* and inocul*), or (restor* and fung*). We used the following "decision rules" to select trials: the projects must have (1) used active ecological restoration of a degraded or constructed ecosystem, (2) incorporated an inoculated treatment and an uninoculated control, and (3) measured PRLC on (4) field-grown plants. We included trials from a variety of ecosystem types, geographic locations, and disturbance types. The landscapes in the selected trials were degraded primarily by human activities (Table 1).
In each trial, mycorrhizal colonization was directly manipulated via the addition of mycorrhizal inoculum as an active restoration technique. The fungal inoculum was sourced from reference ecosystems, commercial sources, or specific fungal isolates (Table 1). Commercial sources included mycorrhizal fungi mixed with unspecified/proprietary granular materials (e.g. Nutri-Link from Schenck & Smith, Native Plants Inc.) and pre-inoculated seedlings that were prepared by commercial nurseries using proprietary methods (Sylvia et al. 1993;de Aragón et al. 2013). Some of the species-specific inocula were prepared from soil originating from the experimental site, mycelial cultures obtained from curated collections, or sporocarps originating from either the experimental site or other ecosystems (Rincón et al. 2007;Alguacil et al. 2011). Some mycorrhizal species were selected based on their desired ecological traits, such as production of abundant fruit bodies or facilitation of plant establishment. Every selected trial included an inoculated treatment group compared with an uninoculated control group. Fungal inoculum was applied directly either to soils in the field study site or to plants in the greenhouse prior to outplanting in the field site.
Plants grew in the field for 4 months (Sylvia et al. 1993) to 54 months (de Aragón et al. 2013). Next, the plants were harvested and roots were analyzed for mycorrhizal colonization. In all cases, microscopy was used to evaluate  (Phillips & Hayman 1970;Ambler & Young 1977;Giovanetti & Mosse 1980;McGonigle et al. 1990;Brundrett et al. 1994). We excluded studies that lacked a direct field component or measured mycorrhizal abundance using spore density or other metrics besides PRLC. To maintain independence of trials, in the case of longitudinal studies that used several time points to measure PRLC, we included only the final PRLC measurement recorded (i.e. representing the longest duration of time). If a study reported multiple sets of results (e.g. geographic location or inoculum type) in which an independent untreated control group was compared with an inoculated treatment group, then each system was designated as a different trial (Sylvia et al. 1993;Al Agely & Sylvia 2008).
For each trial, we obtained the mean PRLCs and numbers of replicates (n) for the inoculated treatment and uninoculated controls. We used these data along with reports of standard error, standard deviation, or summary statistics to calculate a product-moment correlation (r) for each study as in Rosenthal (1991). We then calculated Fisher's z-transform as an effect size for each study, using the formula: In addition, the variance of z was calculated as: For all but one trial, the investigators also measured some aspect of plant performance, usually as shoot dry mass or seedling height (Table 1). We calculated Fisher's z plant and its variance from these data as described for Fisher's z PRLC , above.
To test hypothesis 1, we used a random-effects model to estimate a cumulative Fisher's z PRLC of all 28 trials (Rosenberg et al. 2007). Each trial was weighted by the reciprocal of the variance of z (v z ). In addition, 95% confidence intervals were calculated via bias-corrected bootstrapping with 999 iterations (Rosenberg et al. 2007). We compared the cumulative Fisher's z PRLC against a mean value of zero. Hypothesis 1 would be supported if the cumulative Fisher's z PRLC were significantly greater than zero. Likewise, hypothesis 2 would be supported if the cumulative Fisher's z plant of the 27 trials were significantly greater than zero.
To test hypothesis 3, and check for other aspects of mycorrhizal restoration methods that might influence outcomes, we performed a series of categorical model meta-analyses on grouped data (Rosenberg et al. 2007). Specifically, for hypothesis 3, we tested for differences in cumulative Fisher's z between trials that used various inoculum sources (e.g. whole inocula from reference ecosystem, single species inoculum, or commercial inoculum). Hypothesis 3 would be supported if the cumulative Fisher's z were significantly higher in trials that used whole inocula from references ecosystems than in those that used commercial inocula. We also tested for significant differences among trials that inoculated seedlings in the nursery (followed by outplanting) versus those that applied inoculum in the field. Likewise, we compared studies that measured AM versus ECM colonization of plant roots, and checked for differences among ecosystem types.
Finally, we performed two continuous model meta-analyses (Rosenberg et al. 2007) to examine whether inoculum effects decreased over time under field conditions. For each test, the continuous variable was the amount of time that inoculated plants grew in the field before being assayed for PRLC or plant growth (i.e. "duration in field"). Fisher's z PRLC was the effect size for one test; Fisher's z plant , the other.
We checked for "file-drawer" biases, by which failure to publish null effects of inoculation would influence our results. We used two tests: Kendall's tau test for rank correlations between effect size and sample size, and Orwin's fail-safe N test. We used MetaWin 2.0 for all analyses, and effects were considered significant when p < 0.05.

Characteristics of Selected Studies
The selected trials were conducted on landscapes degraded primarily by anthropogenic activities such as agriculture, logging, construction, desertification, grazing, and mining (Table 1). The majority were based in the Northern Hemisphere, with 15 studies from North America and nine from Western Europe. The remaining studies were located in Morocco, Indonesia, and South America. Coastal dunes, shrublands, and temperate forests were the dominant ecosystem types, followed by tropical forests, temperate grasslands, and tropical savanna. Inoculation of seedlings in the nursery, followed by outplanting in the field site, was more common than the application of inoculum directly to soil in the field.

General Responses to Inoculation
Across all 28 trials, inoculation with mycorrhizal fungi increased PRLC, as indicated by a cumulative Fisher's z PRLC of 0.65 (0.40-0.94, 95% confidence interval), which was significantly greater than the null value of zero (Fig. 1,  p < 0.001). The improvement in mycorrhizal abundance was accompanied by a significant increase in plant growth in the inoculated treatments, with a cumulative Fisher's z plant of 0.57 (0.36-0.91, 95% CI) across all 27 trials that measured plant responses (Fig. 2, p < 0.001). These results supported hypotheses 1 and 2, respectively. Moreover, the inoculation effects did not notably decline with longer exposure to field conditions (Fig. 3). Specifically, the effects of inoculation on PRLC (Q = 0.332, p = 0.498) and plant growth (Q = 0.695, p = 0.565) did not vary significantly with the duration of the field component.
The meta-analysis did not appear to be particularly sensitive to publication bias. Specifically, Kendall's tau tests were not

Variation Among Sources of Inocula
Sources of inocula elicited significantly different effects on PRLC (Fig. 1, p = 0.047). In particular, inocula from reference ecosystems and single species yielded higher increases in PRLC than did inoculum from commercial sources (Fig. 1). The pairwise differences between inocula from reference ecosystems and commercial sources supported hypothesis 3. We note that values of Fisher's z plant associated with the different inocula sources tended to display the same pattern as for Fisher's z PRLC , but in this case, differences among sources were not significant (Fig. 2, p = 0.317).

Discussion
Our findings suggest that restoration ecologists can intentionally increase the abundance of mycorrhizal fungi in degraded ecosystems, and thus improve the establishment of native plants. Moreover, these benefits can last up to several years. Specifically, mycorrhizal colonization of plant roots in field-based restoration projects significantly increased, on average, when mycorrhizal inocula was added. At the same time, plant performance in the field significantly improved. Neither effect declined significantly with time. These responses are consistent with our understanding of the ecology of mycorrhizal fungi-they are important plant mutualists (Allen et al. 2003b;Smith & Read 2008;Hoeksema et al. 2010) that are sensitive to anthropogenic disturbance (Cudlin et al. 2007;Compant et al. 2010;Pickles et al. 2012;Mohan et al. 2014) and may require restoration in their own right.
Moreover, certain sources of inocula were more effective than others. When fungal inoculum from a reference ecosystem or a single fungal taxon was used to inoculate field restoration projects, PRLC increased significantly more than when commercial inoculum was used. Likewise, Sylvia et al. (1993) examined the effects of different inoculum types in Florida dune restorations and found that native mycorrhizal inocula consistently yielded greater mycorrhizal colonization than commercial inocula. Higher root colonization from reference inocula could result from complementary interactions between the fungi and host plants that have developed over time under comparable conditions. In contrast, when a commercial source of inoculum is used, inoculation may shift the composition of the mycorrhizal community away from specialist native mycorrhizal fungi toward generalist "weedy" mycorrhizal fungi that might be less effective mutualists for native plants ( Barea et al. 2011). Indeed, Koch et al. (2011) found that incorporating exotic mycorrhizal fungi into a degraded site substantially changed mycorrhizal communities-even more than plant invasions did. Mycorrhizal community composition may influence restoration outcomes because mycorrhizal species differ in their ability to form relationships with a given plant species, and in their responses to environmental conditions (van der Heidjin 2002;Klironomos 2003;Morris et al. 2007;Hoeksema et al. 2010). The difference observed here in effectiveness of inoculum sources is consistent with this notion.
Researchers have long investigated the role of mycorrhizal fungi in ecological restoration (Ramos-Zapata et al. 2006;Siguenza et al. 2006;van der Wal et al. 2006;White et al. 2008;Vargas et al. 2010). For example, Skujins and Allen (1986) recommended facilitating mycorrhizal growth in degraded sites via inoculation and soil organic retention. Recently, Piñeiro et al. (2013) synthesized data from restoration projects in degraded drylands to compare effectiveness of various restoration techniques. They reported that inoculation with mycorrhizal fungi was generally better than tree shelters, organic amendments, and hydrogels in improving the growth and survival of seedlings. Likewise, Barea et al. (2011) reviewed the use of mycorrhizal inoculation in revegetation projects on degraded semiarid lands in Southeast Spain, finding that inoculation improved both plant development and soil quality. Barea et al. (2011) also compared native inoculum from reference ecosystems versus exotic commercial inoculum and found that in dry environments, native, drought-adapted mycorrhizal fungi improve plant performance more than non-native mycorrhizal fungi. Our meta-analysis extends these observations beyond drylands, and suggests that inoculation could improve colonization and plant growth in a variety of ecosystems.
Similar findings have been documented in studies that did not involve restoration. For example, Lekberg and Koide (2005) conducted a meta-analysis on agricultural systems, and reported that mycorrhizal colonization and plant biomass was augmented by inoculation. Moreover, Hoeksema et al. (2010) conducted a broad meta-analysis of the effects of mycorrhizal inoculation on plant growth in studies based in laboratories, agricultural fields, plantations, and natural ecosystems (including one restoration project). They observed increases in plant growth in response to inoculation in the field as well as the laboratory. None of these other meta-analyses compared inoculation methods in the nursery to those in the field.
In our meta-analysis, inoculation of mycorrhizal fungi increased plant performance, on average. Plant responses to inoculation treatments largely mirrored those of mycorrhizal colonization. If inoculation increases PRLC, then this in turn increases the surface area of mycorrhizal fungi interacting with plant roots and accessing nutrients in the soil environment. Plants with greater PRLC would then benefit from receiving more nutrients from mycorrhizal fungi than would uninoculated plants (Fitter 1985;Read 1999;Varma & Hock 1999;Allen et al. 2003b). In fact, Barea et al. (2011) demonstrated that inoculation increases outplanting performance, survival, and plant biomass in restoration projects in semiarid lands, ostensibly because mycorrhizal fungi improve plant resistance to the drought stress, nutrient deficiency, and soil degradation that are common in degraded ecosystems.
Our findings can be useful in developing best practices in restoration ecology. Intentional inoculation could improve the growth and establishment of a mycorrhizal community in restored ecosystems, which may benefit aboveground vegetation communities. In addition, using inoculum from a reference ecosystem or a species-specific inoculum may increase PRLC more than a commercial inoculum. The use of inocula from reference ecosystems could also be economically advantageous for practitioners who otherwise might purchase commercial inocula. In large-scale projects, the cost of commerical inocula would not be trivial. There was no significant difference between PRLC in species-specific inoculum and from reference ecosystems, yet it is more technically challenging and time consuming to isolate mycorrhizal species than it is to obtain inoculum from a reference ecosystem.
When obtaining inocula from undisturbed sites, practices that might harm reference ecosystems should be avoided. To reduce disturbance, sourcing small volumes of soil from reference sites to inoculate plants in the greenhouse prior to outplanting in the field may be preferable to transferring large volumes of native soil (Cairns 1995). Furthermore, sourcing inoculum from edges of reference ecosystems could decrease disturbance in untrammeled sites (Mitsch & Jørgensen 2004). Sourcing inocula from restored sites with similar historical disturbances could be advantageous, because fungal propagules from these restored sites may be adapted to the special conditions associated with a particular disturbance (Greipsson 2010;Orlowska et al. 2011). In cases where reference sites are slated for development, topsoil containing fungal propagules could be salvaged from the reference site prior to construction or development (Rowe et al. 2007). Nevertheless, it is worth considering that after several months, soil stockpiling may reduce mycorrhizal abundance and compromise the effectiveness of this inoculum source (Galatowitsch 2012). The most appropriate inoculum choice depends on several factors, but altogether, it may be more practical and economically feasible for restoration ecologists to integrate reference inoculum into their restoration protocols, instead of species-specific or commercial inoculum.
The number of trials represented in this meta-analysis is somewhat small, because we restricted our selection to restoration projects that assessed PRLC. Nevertheless, we were able to detect significant effects of inoculation across trials. Furthermore, inoculation effects were positive in the majority of trials-only four of 28 trials reported a decline in PRLC, and three of 27 in plant growth. Nevertheless, we note that even though a variety of ecosystem types were represented, the trials were mostly based in temperate ecosystems in the Northern Hemisphere. Additional research in tropical ecosystems and the Southern Hemisphere would be valuable.
Although mycorrhizal fungi are just one component of a broader belowground community, they are the only component of the microbial community we examined in this meta-analysis. This focus is deliberate; O'Neill et al. (1991) posited that mycorrhizal fungi are "keystone mutualists" in terrestrial ecosystems, and therefore may exert a disproportionate influence on other soil microbes on the site. If mycorrhizal fungi affect community structure and ecosystem processes, and thus restoration outcomes, then they may influence other members of the soil microbial community. Moreover, mycorrhizal inoculation may increase plant cover through added plant biomass, which could lead to increased protection of exposed soil surfaces from solar radiation and other harsh environmental conditions. This added plant biomass would also provide substrate for decomposer microbes that may provide additional benefits to restored ecosystems, such as erosion control or increased soil organic matter. Future studies exploring the role of mycorrhizal inoculation in influencing the total microbial community and ecosystem processes at restoration sites would clarify the direct and indirect roles of these fungi (Brunson et al. 2010;Kulmatiski 2011;Binet et al. 2013).
In conclusion, this meta-analysis indicates that mycorrhizal fungi can be directly manipulated in many restoration sites via inoculation. In turn, inoculation generally improved restoration success by increasing plant performance. Moreover, sources of inocula varied in their effects on mycorrhizal colonization. In particular, the use of inoculum from reference ecosystems may be particularly effective and practical, compared with species-specific and commercial inocula. Land managers may wish to consider incorporating mycorrhizal fungi in their restoration efforts to better facilitate the establishment of below-and aboveground components of ecosystems.