Autotrophs require macro-elements such as nitrogen (N), phosphorus (P), potassium (K), and trace elements such as iron (Fe), cobalt (Co), and molybdenum (Mo), among others (Lampert and Sommer, 2007; Kaspari and Powers, 2016). These essential elements provide structural, electrochemical, and catalytic functions, but on Earth they are not particularly abundant in soluble forms that can be taken up by most primary producers (Sterner and Elser, 2002). N and P, necessary for protein synthesis, building nucleic acids, and energy transfer, are the primary limiting nutrients in most aquatic (Conley et al., 2009) and terrestrial (Du et al., 2020) ecosystems. In a global meta-analysis, Elser et al. (2007) compiled a dataset of growth experiments in which N, P, and N&P combined were factorially supplied to primary producers. The outcomes of the so called ‘factorial limitation scenarios’ showed that both N and P limitation of primary producer growth is strong and widespread across ecosystems and that combined N&P enrichment leads to a positive synergistic growth response compared to single N and P addition (Elser et al., 2007). Expanding on this previous study, Harpole et al. (2011) found not only a synergistic effect of N and P addition on autotroph growth, but also identified different types of factorial limitation scenarios that varied in magnitude depending on whether primary producer communities were limited by a single nutrient or co-limited. More recent studies highlighted the prevalence of co-limitation by N&P and other elements in different systems (Ma et al., 2019; Klupar et al., 2021; Browning and Moore, 2023).

The observation that autotrophs are frequently co-limited by more than one nutrient supports the Multiple Limitation Hypothesis (Gleeson and Tilman, 1992), which was developed in contrast to Liebig’s Law of the Minimum (Von Liebig, 1840). Von Liebig formulated the law to describe the limitation of crop yields by essential nutrients, but it has been generally used to explain how autotroph growth is limited in different systems (Paris, 1992). The law postulates that only a single nutrient is limiting at a time, depending on the availability of that nutrient and its demand by the organism (de Baar, 1994). On the other hand, the Multiple Limitation Hypothesis states that more than one nutrient can simultaneously limit growth (Gleeson and Tilman, 1992). These two concepts differ in their assumption of the type of the essential resource or nutrient (Tilman, 1982; Saito et al., 2008; Sperfeld et al., 2012). Liebig’s law assumes nutrients to be strictly essential, so that the limiting nutrient can change continually and even abruptly, but only one is limiting at a given time. Contrarily, the Multiple Limitation Hypothesis assumes interactively essential nutrients; thus, all nutrients are co-limiting simultaneously, but at different magnitudes depending on supply and demand. Sperfeld et al. (2016) integrated Liebig’s Law of the Minimum and the Multiple Limitation Hypothesis into a general (co-)limitation framework, which also comprises two approaches frequently used to study nutrient (co-)limitation, i) factorial limitation scenarios (Craine, 2009; Allgeier et al., 2011; Harpole et al., 2011) and ii) ‘response surfaces’ across a multi-dimensional, continuous nutrient space based on specific nutrient categories (Tilman, 1982; Saito et al., 2008; Sperfeld et al., 2012).

This theoretical (co-)limitation framework links the two approaches in that several factorial (co-)limitation scenarios emerge within a two-dimensional continuous supply space of two nutrients, depending on the initial concentration and ratio, and on the response surface type of the co-limiting nutrients (Sperfeld et al., 2016; Halvorson et al., 2017; Redoglio et al., 2022). The response surfaces in this framework were developed combining Tilman (1980, 1982) graphical approach of using growth isoclines for describing different resource categories and Saito et al. (2008) definitions of (co-)limitation types using simple mathematical equations. Depending on the location of a hypothetical basal nutrient supply point (i.e. a control, Ctrl) within the continuous nutrient space, factorial nutrient additions (i.e. N, P and N&P supply) lead to growth responses that result in different types of factorial limitation scenarios (see Figure 1 ; Supplementary Material ). Among scenarios classified as true co-limitation, simultaneous co-limitation arises when only the combined addition of two nutrients causes a positive response, while independent co-limitation results from a positive response to the addition of either nutrient alone or in combination (Harpole et al., 2011; Sperfeld et al., 2016; Figure 1 ). In these cases, the addition of both nutrients produces an additive or synergistic response compared to the single nutrient addition. On the other hand, single limitation occurs when only the supplement of one nutrient and the addition of both nutrients produces the same positive response, while serial limitation arises if the response to the addition of both nutrients is higher than when a single nutrient is supplied ( Figure 1 ). In these scenarios, not only the addition of both nutrients leads to a positive response, but also the supply of one of the two nutrients alone can cause a response comparable to that from the addition of both nutrients. Some of these factorial limitation scenarios can emerge across different response surfaces types and others are specific to a response surface type, also depending on where the basal nutrient supply point before factorial additions is situated within the nutrient space (Sperfeld et al., 2016).

The use of factorial growth assays and the resulting limitation scenarios is advantageous to address nutrient (co-)limitation not only because it can be reconciled with the theoretical background through response surfaces (Sperfeld et al., 2016), but also because it allows to directly test with nutrient addition experiments which nutrient alone or combination elicits a response, i.e. is limiting growth. The direct response to nutrient addition often measured in autotrophs at limiting conditions is biomass increase (Elser et al., 2007; Harpole et al., 2011; Browning and Moore, 2023). Nutrient limitation can also be indirectly estimated by measuring the nutrient concentration in the environment to infer what is limiting depending on scarcity, even though this approach cannot ‘prove’ what is limiting growth at a given time and nutrient concentration in the environment is usually also far from constant (Beardall et al., 2001). Autotroph elemental composition (e.g. cellular nutrient quotas or tissue concentrations) on the other hand can also point to the nutrients limiting growth and relates to the physiology of organisms (Marleau et al., 2015; Wirtz and Kerimoglu, 2016), but often the stoichiometric composition itself is influenced by other co-limiting nutrients (Bracken et al., 2015), making the direct identification of the (co-)limiting nutrients challenging.

Several mechanisms explaining the observed co-limitation patterns have been proposed at different levels of biological organization, but are still not well understood. At the biochemical level, for example, nutrients can substitute for the same function in a molecule (e.g. two elements equally substitutable within the same enzyme, such as zinc and cobalt in carbonic anhydrase; Sunda and Huntsman, 1995). The limitation of one nutrient can also lead to the inability to acquire another nutrient, and both nutrients must be simultaneously available to perform the biochemical function (Saito et al., 2008; Moore et al., 2013). At the cellular and organism level, nutrients can be in such a low concentration for uptake that both need to be added to observe a growth response (Arrigo, 2005; Sperfeld et al., 2016). In addition, the mechanisms described at the biochemical level can be extended to cells and organisms, considering that several elements might limit distinct biochemical pathways and that the mechanisms involved might differ for each process (Harpole et al., 2011).

At the community level, the presence of species and functional groups with contrasting physiologies and traits adds complexity (Reynolds et al., 2002; Litchman et al., 2007; Falkenberg et al., 2013), making our understanding of the mechanisms of nutrient (co-)limitation more difficult. Several studies have proposed mechanisms to explain the observed patterns of nutrient (co-)limitation of primary producer communities. Harpole et al. (2011) suggested that in the case of a community consisting of functionally equivalent species co-limited by the same nutrients, the mechanisms defined at the biochemical and organismal level should also apply. In this niche overlap setting, (co-)limitation at the community level reflects the patterns of single species or organisms and thus mechanisms at the biochemical level can be extended to communities. On the other hand, species may occupy separate niches within the community or vary in their nutrient requirements and thus can be each limited by different nutrients simultaneously (Arrigo, 2005). Marleau et al. (2015) also showed in a model study that differences in movement rates of nutrients can explain nutrient (co-)limitation patterns of autotrophs at the ecosystem level. Studies investigating the mechanisms of (co-)limitation at the community level are very scarce. Danger et al. (2008), using bacterial communities as a model, suggested that Liebig’s Law of the Minimum is not applicable at the community level and that communities as a whole are co-limited by nutrients due to competition and coexistence processes that optimize resource use. A model on terrestrial plants predicted responses to N and P limitation that resemble the theoretical response surfaces of both Liebig’s Law of the Minimum and the Multiple Limitation Hypothesis (Ågren et al., 2012). Kaspari and Powers (2016) also proposed to retire Liebig’s Law of the Minimum and to consider the interaction between elements that leads to co-limitation. Bannon et al. (2022) recently introduced the concept of Community Interaction Co-limitation, that is a limitation of interacting groups within the community by different nutrients that results in an inadequate production of a biologically produced nutrient. They also highlighted how this is likely to be impacted by community composition, because of the close interactions among different microbial groups. Despite extensive data on the frequency of (co-)limitation by N and P in primary producers across terrestrial and aquatic ecosystems (Elser et al., 2007; Allgeier et al., 2011; Harpole et al., 2011; Bratt et al., 2020), we still lack knowledge on the underlying mechanisms that result in the observed (co-)limitation scenarios in autotroph community biomass.

Using experimental microcosms with phytoplankton communities as a model system, we explored what leads to the frequently observed patterns of N and P (co-)limitation in autotroph communities. We assess the type of (co-)limitation at the community level as the response of overall community biomass to factorial nutrient additions after an initial growing phase at certain nutrient conditions. The communities were pre-grown at different N:P supply ratios and subsequently supplied factorially with N and P to assess the factorial limitation scenarios. Depending on the changes in species composition and growth patterns before and after nutrient additions in the different communities, we explored possible mechanisms based on biomass responses of individual species leading to nutrient (co-)limitation patterns at the community level (cf. Supplementary Figure S1 ). Similar to Ågren et al. (2012), we hypothesize the emergence of single or serial limitation scenarios after factorial N and P additions at imbalanced N:P ratios ( Figure 1 ). At more balanced N:P ratios, true co-limitation patterns (i.e. simultaneous and/or independent co-limitation) will likely arise after nutrient additions ( Figure 1 ). We further hypothesize that the community limitation pattern is driven by the dominant species, favored by the addition of the nutrient in shortest supply, if a community is limited by a single nutrient ( Supplementary Figure S1B ). For N&P co-limitation scenarios, we hypothesize that either the dominant species drives the pattern in overall community biomass ( Supplementary Figure S1C ) or sub-dominant species contribute to the observed community biomass patterns if the dominant species does not respond to nutrient additions ( Supplementary Figure S1A ). Finally, we propose that Liebig’s Law of the Minimum can be still applied to the community level, even though each species in the community is limited or co-limited by different nutrients depending on resource use efficiency.

We used phytoplankton communities in highly controlled microcosms to investigate potential mechanisms based on biomass responses of individual species that lead to the development of nutrient (co-)limitation patterns at the community level. We designed an experiment simulating several (co-)limitation scenarios and used biovolume data of three communities varying in species composition to explore what has led to the different community growth patterns based on their biomass responses.

At the end of the second growing phase, we identified single N limitation and co-limitation by N and P at the community level. These limitation scenarios were mostly based on responses of the dominant species (see below). Strictly defined co-limitation scenarios were frequently observed in our experimental communities, but single N limitation was the most common scenario, which is partially in agreement with the findings of the meta-analysis of Harpole et al. (2011). In a study with a similar experimental design and phytoplankton communities, but with an initial ‘balanced’ N:P = 16:1 (Redfield Ratio: Redfield, 1958) in the growth medium, we did not detect co-limitation and N was the single or primary limiting nutrient (Redoglio et al., 2022). Frank et al. (2020) also reported a stronger response to N supply in experimental phytoplankton communities, suggesting optimal N:P > 16. By shifting towards N:P = 35 in the initial growth medium as the more balanced nutrient ratio, we aimed to create conditions where co-limitation scenarios can emerge.

We expected co-limitation to arise at N:P = 35 ( Figure 1 ), but this was observed only for community 1 where S. elongatus was dominating. In community 2 and 3, where green algae A. obliquus and M. minutum were prevalent ( Figure 4 ), we observed single N limitation ( Figure 3 ). On the other hand, we detected co-limitation, or at least a tendency, at N:P = 70:1 in all three communities. Despite this medium being already N-rich from phase 1, P supply significantly favored growth only when combined with N. At N:P = 17.5, our expectation of N to become the limiting nutrient was met. Thus, we did not find a confirmation for the hypothesis of co-limitation emerging under the more balanced N:P = 35 across all communities, while the hypothesis of single or serial limitation arising at imbalanced N:P was partially supported, as we detected single N limitation at N:P = 17:5, but not single or serial P limitation at N:P = 70 in the communities.

These results suggest that community 2 and 3 had a high optimal N:P and were particularly N demanding, while the more balanced N:P of community 1 was closer to our expectations. This can be explained considering the phytoplankton species composition ( Figure 4 ; Supplementary Figure S4 ) and their traits regarding nutrient requirements (Litchman et al., 2007). Already at the end of phase 1 and later at the end of phase 2, our communities were dominated either by green algae or cyanobacteria, which have been shown to have an optimal N:P higher than the Redfield ratio (Arrigo, 2005; Hillebrand et al., 2013). The dominating green algae (A. obliquus and M. minutum) and the cyanobacterium S. elongatus are also fast growing species (Sperfeld et al., 2010). Concurrently, cyanobacteria such as Anabaena spp. are known to fix atmospheric nitrogen (Huisman et al., 2018). Thus, their higher occurrence relative to total biovolume in community 1, might have lowered the optimal N:P for that community compared to the others. In addition, we used species with different nutrient uptake capabilities ( Table 1 ), that may have assimilated the nutrients after nutrient spikes differently and thus grew at different rates during the non-stationary growth phase. Physiological traits of the phytoplankton species might thus have influenced the community structure and dominance of species (Litchman et al., 2007) at stationary growth in our experiment. This still represents a realistic scenario, as it is often observed in nature that certain species are dominating under particular conditions (Barcelos Ramos et al., 2017; Burson et al., 2018).

Overall, our hypotheses that mainly the dominant species in case of single nutrient limitation determines the community pattern ( Supplementary Figures S1B, C ), while also sub-dominant species in case of co-limitation can drive the community response ( Supplementary Figure S1A ), are only partially supported. Although (co-)limitation patterns of whole community biomass could theoretically be partly or entirely the result of biomass responses of rare or sub-dominant species ( Supplementary Figures S1A, D ), we found that both co-limitation and single N limitation patterns identified for the whole community were strongly driven by the dominant species. In community 1 and 2, S. elongatus and A. obliquus were respectively ten times more abundant than the other species, which indicates these species as better competitors for the nutrients in the given ratios and supplies (Litchman et al., 2007; Edwards et al., 2011). It appears that M. minutum, despite being relatively low in biovolume in community 2, showed similar single N limitation patterns to A. obliquus, which is expected as both green algae are assumed to have similar nutrient demands and traits (Sperfeld et al., 2010). In these cases (community 1 and 2), where the community has low diversity, and one species outcompetes the others, it seems that co-limitation arises only when the dominant species is co-limited and sub-dominant species do not contribute much to the community response. In community 3, species were more equally abundant, even though the green algae M. minutum and A. obliquus together prevailed ( Figure 4 ). Interestingly, although not significant, we can see a tendency of community co-limitation at N:P = 70, despite the single species showing contrasting limitation patterns. Nevertheless, S. elongatus, which mostly drove the overall pattern in community 3, seemed to be more responsive to both P and N&P additions ( Figure 3 ), while A. obliquus which also contributed to the overall pattern, seemed to increase only after N&P addition. In such a case of a community in which species are more evenly distributed, sub-dominant species can be (co-)limited by different nutrients and each species may contribute to the overall biomass pattern observed (cf. Supplementary Figure S1D ).

The biovolume of species at the end of the phase 2 is the result of a 43-days long experiment with nutrient manipulations where communities starting with different sets of species were allowed to build up biomass. Due to species interactions along the duration of the experiment, community processes such as competition for resources, species sorting and competitive exclusion, led some species to become dominant. Also natural communities can often show biomass distributions with one or few dominant species (Tokeshi, 1993). In particular at high nutrient supplies, phytoplankton communities can increase their productivity but decrease in diversity, with few abundant and several rare species in the ocean (Vallina et al., 2014) and in freshwater systems (Reinl et al., 2022). Nevertheless, even if rare species do not contribute considerably to overall community biomass, they can be important in competitive processes, as we observed in community 2 and 3. The rare species C. ovata, N. limnetica and S. hantszchii influenced the competitive outcome of the dominant green algae species M. minutum and A. obliquus ( Figure 4 ); in community 3, where C. ovata, N. limnetica and S. hantszchii were present, the two green algae showed a comparable biovolume, while in community 2, where the rare species were not part of the initial community assembly, A. obliquus strongly dominated the community. Likewise, the cyanobacterium S. elongatus showed much higher relative biovolumes in community 3 than in community 2 ( Figure 4 ), which depended solely on the presence/absence of the three rare species in the initial set of species, and further highlights the effect of low biomass species on the competition outcomes in our communities.

The community biomass responses ( Table 2 ) and the responses of individual species within the communities ( Figure 3 ) differed depending on the initial medium N:P, the nutrient supply treatment, and the community itself. As a reaction to the manipulation of resources, such as nutrient supply and ratio, species can adjust their morphology or physiology for uptake-related traits (Bonachela et al., 2011). In addition, phytoplankton species have been shown to exhibit different degrees of flexibility in their demands for optimal nutrient ratios, particularly depending on their growth rate (Hillebrand et al., 2013). This phenotypic or physiological plasticity can also influence species interactions and sorting, and thus overall community composition (Miner et al., 2005). For example, the cyanobacterium S. elongatus showed a similar growth response at different media (N:P = 35 and 70) in community 1 where it was dominating, but the growth response changed in community 2 and 3 and also differed in the aforementioned media in these communities ( Figure 3 ), indicating that S. elongatus exhibited plasticity in resource acquisition traits and that this plasticity was modulated by the community context and competition processes with other species.

The trait-based ecology approach proposes that community ecology and characteristics can be assessed based on traits that vary among species (Litchman and Klausmeier, 2008). Traits such as resource utilization or cell size can, among others, shape the species distribution and thus contribute to structuring phytoplankton communities (Litchman et al., 2010; Sommer et al., 2017). If we apply this approach to our experiment and to the predictions of nutrient (co-)limitation scenarios, we see that where one or closely-related species are dominant, the community (co-)limitation pattern can be predicted based on the response of the dominant species or functional group. Thus, where the patterns coincide, the community response might also be predicted based on the traits of the species or groups regarding nutrient acquisition and storage. When the species in the community are more evenly distributed, our data show that (co-)limitation patterns can differ between individual species of contrasting nutrient requirements and the whole community, which makes predicting a community response from traits of single species more difficult.

Arrigo (2005) proposed that the addition of multiple nutrients may favor the growth of different members of the community, thus resulting in community co-limitation even if the species of the community are each limited by a single nutrient. In our experiment, we found a confirmation of this only when the species composition of the community was more even. However, as this was the case only in one out of nine instances, further investigations are needed to support this hypothesis. The community as a whole can be also co-limited by multiple nutrients when species occupy separate niches and when they have different physiologies and nutrient requirements (Falkenberg et al., 2013). The patterns identified for the different species can thus lead to a community limitation pattern that does not arise in single species, as we saw in community 2 and slightly in community 3, N:P = 70. When a species became dominant, the community (co-)limitation pattern was driven by that species, regardless of the limitation type (single or co-limitation). Thus, based on observations of similar biomass responses, the nutrient requirements of the community seem to coincide with those of the most abundant species. Our results are partially in contrast with what has been studied by Danger et al. (2008), who used bacteria communities to test their assumption that Liebig’s Law of the Minimum does not hold at the community level. We found that phytoplankton communities can be limited by a single nutrient, following Liebig’s Law of the Minimum, as well as co-limited by multiple nutrients. Outcomes from our experiments are more in line with model predictions by Ågren et al. (2012) of limitation responses that can follow both Liebig’s Law of the Minimum and the Multiple Limitation Hypothesis.

With our experiment, we wanted to investigate the underlying mechanisms in terms of biomass responses of individual species driving (co-)limitation of primary producer communities. We found that the dominant species, or functional groups with species showing similar nutrient requirements (inferred based on observations of similar biomass responses), mostly determine the (co-)limitation pattern observed at the community level. We suggest that the dominance of species or groups with specific nutrient requirements is one important mechanism that determines the pattern of nutrient (co-)limitation observed at the community level. However, this mechanism might only be valid if nutrient limitation (bottom-up) is the main driver of autotroph biomass responses, as other processes like the strength of herbivory (top-down) could also change the outcomes. In addition, when the abundances of species in a community are more evenly distributed, a community co-limitation scenario can arise even though species are each limited by different nutrients. We propose that not only the Multiple Limitation Hypothesis but also Liebig’s Law of the Minimum can be applied to communities as a whole. A single nutrient limitation pattern that follows Liebig’s Law of the Minimum can emerge at the community level, which means that overall community biomass can be seemingly limited by a single nutrient despite the individual species being limited by other or different nutrients. This makes the understanding of co-limitation at the community level more complex, and ecological mechanisms of species coexistence, competition and predation may add a further layer of complexity. Furthermore, we highlight the importance of community properties, such as species abundance and composition, and of species traits related to growth rate and nutrient acquisition that can shape the response of the community to nutrient (co-)limitation. Our study is a first step aiming a better understanding of community (co-)limitation in terms of biomass responses. Future studies should additionally consider and integrate physiological and/or biochemical responses to have a deeper understanding on the mechanisms involved at lower levels of biological organization. Finally, our results show that community (co-)limitation outcomes can be predicted based on knowledge of nutrient use requirements of one or few dominant species, which may be a useful tool for lake restoration and oligotrophication efforts.

The original contributions presented in the study are publicly available. This data can be found here: FigShare, https://figshare.com/s/fa812b103c58d8087c87.

AR: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing. ES: Conceptualization, Funding acquisition, Project administration, Resources, Supervision, Validation, Writing – review & editing, Formal analysis.