Cities have become focal points for addressing sustainability issues and have emerged as important political protagonists in global sustainable development (SD)1,2. Numerous SD policy agendas at the urban level aim to balance and negotiate between environmental, social and economic goals3,4. A growing number of cities are taking steps to reduce their carbon footprint, welcome migrants, develop alternative economic development agendas and protect biodiversity within and beyond the city. Some city initiatives contain radical, progressive elements that prioritize the needs of the most disadvantaged5. These urban sustainable development (USD) policies frequently exceed national policies in terms of their depth, innovativeness and/or progressiveness. In doing so, they thereby challenge the primacy of national governments in SD.

This article examines an array of urban policy measures that fall within the realm of SD. It aims to better understand specific urban takes on SD and to examine how they are welcomed and/or contested in multi-level governance settings. We draw on the concept on multi-level governance (MLG) stemming from political science, to examine the manifold interactions between state and non-state actors across multiple policy-relevant scales without a structuring authority6. The concept of MLG challenges the state-centric and hierarchical distribution of power and responsibility; it encompasses state and non-state actors and assumes that these policy-making interactions contain degrees of negotiation but also policy disputes7,8. The MLG framework explicitly focuses on diverse strategies and forms of cooperation that actors at different policy-making levels adopt to advance their policy agendas.

While we acknowledge the work on multi-level processes in the sustainability transitions literature9,10, here we focus specifically on national–urban policy-making relationships for which the MLG framework is well suited. The framework allows us to consider in what cases USD policies align and advance national policy agendas and when the two are in conflict. Where policies are contested, cities may (explicitly or implicitly) deploy positioning strategies to contradict, stretch, or circumvent such higher-level agendas11,12,13.

Shifting the focus to urban policies and positioning strategies may thus advance the understanding of the role of cities in global SD agendas and highlight that urban policy-making for SD is not as de-politicized and uncontested between different levels of government as the literature sometimes suggests5,14. At the same time, we broaden the focus by not only looking at one relatively narrow policy area (such as migration governance) but discussing an array of social, environmental and economic policy measures that contribute to the wider goal of SD.

In this article, we introduce 12 illustrative case study vignettes from 11 European cities that have formulated and implemented SD policies (see Table 1). In each case, we focus on specific policy measures, not on whole policy programs or policy agendas. Our intent is not to map USD agendas or to present specific policies and positioning strategies that have been particularly successful. Rather, we seek to shed light on how specific urban policies are supported by corresponding positioning and cooperation strategies at different governance levels. By categorizing urban policy types and positioning strategies, we propose a way of reflecting on how cities seek to place specifically urban takes of SD on global agendas.

Table 1 Types of urban sustainable development policies.

The 12 vignettes all present USD policy measures with varying thematic emphases, which are either discussed in the academic literature or showcased through international city networks. We have purposefully selected policy measures of high visibility. After an initial screening of the literature, we chose the cases and conducted semi-structured interviews with 16 experts involved in the implementation of these policy measures (see Supplementary Table 1 for a list of interview partners and Table A2 for the interview questionnaire). The interviews allowed us to better understand the nuances and dynamics of each case.


Urban policy types for sustainable development and corresponding positioning strategies

Based on the selected vignettes, we developed a framework that conceptualizes urban perspectives on MLG interactions in USD policy-making. The framework consists of two dimensions: (1) congruence between the national and urban policy levels, and (2) the scale of the policy issue. Accordingly, the first dimension identifies whether urban policies are aligned with the national agenda, or whether local authorities adopt policies that diverge from the national policy thrust. Building on reflections on MLG interactions, this factor is important as different levels of government may have different interpretations of policy problems, leading to the integration of contradictory goals in policy designs. Such differing interpretations across government tiers may thus produce ‘decoupled’ policies, in which government players follow divergent, or sometimes even conflicting, policy paths15. The second dimension indicates the scale of the policy, meaning whether the particular policy issue is to be addressed at a more global or national level or mainly on an urban scale. This factor is important as an indicator of whether cities approach these issues by ‘looking up’—i.e. with an ambition to scale up their activities—or by ‘looking around’, i.e. seeking guidance and exchange while retaining a primary focus on the implementation of the policy within their city or region16. Categorizing policies along these two dimensions yields a 2 × 2 table with four cells representing different Weberian ideal types of approaches to USD (Table 1).

For each policy type, we furthermore identified the primary cooperation and positioning strategies at different governance levels. Cities use various strategies to defend, legitimize or advance their policies in MLG settings11,17,18. These strategies can be either explicit or implicit16.

In analysing the vignettes, we found that cities build on different cooperation strategies linked to particular policy measures to position themselves within or against the national agenda. Depending on the policy type, these positioning strategies are deployed to (1) legitimate their actions, thus taking ownership of certain policy areas; (2) gain visibility; or (3) move beyond the local scale, thus seeking to contribute to the wider goal of global SD. These cooperation strategies may target multiple scales (a metropolitan area, cities in the same country, the national government, or international cities) that have been identified as relevant for MLG policy-making from an urban perspective19,20.

Table 2 summarizes our findings and links the urban policy types with corresponding ideal-typical positioning strategies: best practices, first practices, defiant practices and dodging practices. Our understanding of the latter two has been informed by Oomen et al. (2022), who examined strategies of divergence in migration governance (We borrowed the strategies presented by Oomen et al. (2022) for the cases that fall within our category of divergence (i.e. when national and urban policy are not congruent). We applied the strategies of defiance and dodging provided by Oomen et al. (2022), but not deviation and dilution as they do not align with explicit strategies in our cases.). Table 2 also displays our empirical findings with regard to possible cooperation strategies at the different governance levels, which cities employ to work towards policy implementation and to position themselves in the respective SD field. Although we acknowledge that in each type of positioning strategy, cities cooperate simultaneously with actors on multiple scales, the table displays the most important cooperation strategies in bold, to recognize the central role of this particular cooperation strategy for the advancement of the respective positioning strategy.

Table 2 Urban positioning and cooperation strategies.

We will now turn to each of the four policy types and discuss them by reference to the illustrative case study vignettes. We focus on the cities’ positioning strategies and how these are linked to the specificities of the four ideal types of USD policies.

Policy alignment with global focus: best practices

This USD policy type emerged from analysing climate-positive policies in Stockholm and Sønderborg, along with the circular economy policies in Berlin and in the Capital Region of Denmark. Stockholm is developing a large-scale fossil-free urban district that features low resource consumption and sustainable construction, while Sønderborg is aiming for citywide CO2-neutral development. Berlin and the Capital Region of Denmark are metropolitan regions that actively promote a circular economy in the building industry and in waste management. Accordingly, both contract with builders that selectively dismantle buildings and reuse intact materials.

These policy examples are mainly located within the domain of environmental SD. Environmental protection is an important policy area at the national level with corresponding strong and well-funded national agendas and international networks1,21,22. As both the national and the urban levels support the same policy thrust, there is no apparent goal conflict between the two scales2,21,23. Measures that cities take in this field may even help the national government meet the targets articulated as part of national or international agendas. This is also expected as urban areas drive a large share of the national carbon footprint24. This policy type is labelled in Table 1 as ‘policy alignment with global focus’.

To promote their environmental agendas, the examined cities and metropolitan region actively seek cooperation partners at different scales. As their policy aims align with national and international goals, cities are also active at the national and international scale (Interviews 1 to 5; hereafter interviews will be identified as I). In Stockholm for example, the city official states that they “always try to get financial support from the national level, of course, and it also generates good PR” (own translation of I1). There are well-developed city networks that address environmental sustainability issues, and cities can explicitly position themselves as central actors for progressive measures against climate change2,25. For Stockholm, Sønderborg, and the Capital Region of Denmark, being part of networks like C40 or Cityloops/ICLEI not only grants them visibility but also helps them obtain funding for other climate change–related initiatives (I1, I2, I5). As for Cityloops, it was crucial to be able to rely on the network of ICLEI: “You need somebody who can help you identify partners if you do not already have a network on that” (I5). International and national networks also provide a platform for the exchange of knowledge and experiences (I2, I4, I5). Sønderborg, together with other Danish municipalities, developed a nationwide initiative to support locally adapted climate action plans. The Capital Region of Denmark, along with other cities in the Cityloops project, formulated a European Circular Cities Declaration, which seeks to support other cities in their transition towards a more circular economy. The local and metropolitan scale is also of importance for city actors, as they need local partners to implement projects.

In the examined cases, the cities and capital region present themselves as champions, or “municipal lighthouse” (I2), in the respective policy area. They seek to gain visibility by showcasing their progressive policy approaches as best practices. Doing so not only serves as a tactic to attract funding but also functions as an advocacy strategy to scale their approach up to the national level and to other cities (I1, I2, I3, I4, I5). The city official of Berlin, for example, explains that with the success of their measures, “people come from everywhere, take a look at it and bring it to the outside world. This leads to actors from outside also inquiring about what is being initiated in Berlin” (own translation of I4). We thus refer to the positioning strategy of the aligned policy type with global focus as ‘best practices’ (Table 2).

Policy alignment with urban focus: first practices

The second USD policy type has emerged from examining social innovation projects in Vienna and Newcastle upon Tyne, as well as ‘24 h’ policies in London. Vienna, as part of its ‘Smart City Framework Strategy,’ is not only centered on digitalization, but emphasizes ensuring a high quality of life for everyone and, as a city official explains, “addresses all aspects of city life. Not only mobility, buildings and energy, but also awareness raising, participatory processes, [or] creating new work opportunities” (I10). Digitalization is thus seen as a tool to support not only technical innovation, but more importantly also social innovation. Similarly, in Newcastle upon Tyne the University originally established a platform to experiment with projects and build on different forms of cooperation and enhanced digital developments. Originally a platform of engagement, the platform was then used to “feed new ideas into policy” (I11), for research and for innovation through projects. London, on the other hand, has created the policy focus, ‘24 h London,’ which recognizes and addresses contemporary changes in the city associated with the gradual growth of nocturnal activities.

Cities, as ‘burning glasses’ of our society8, experience the immediate consequences of spatial, socio-economic, environmental and political transformations and thus often have to find interdisciplinary approaches to address emerging challenges. Against this backdrop, cities around the world have increasingly introduced experimental projects and open-outcome processes to find alternative solutions and produce transformations towards SD26. One specific cultural and socio-economic change occurring in cities is that the classic dichotomy between day and night is becoming eroded27. As a result, 24 h policies have emerged as a means to develop the evening and nighttime economy of cities and to create safer city centres, since people who do not work from 9:00 a.m. to 5:00 p.m. often face more dangerous and stressful working conditions27,28. Whereas such initiatives do not directly conflict with the national policy thrust, they address issues that arise mainly at the urban level and are of less importance at the national scale. This policy type, shown at the top right of Table 1, is referred to as ‘policy alignment with urban focus’.

Promoting innovative, locally grounded policies for SD requires cities to establish partnerships with actors within the same city. In both Vienna and Newcastle, local networks and forms of cooperation at the urban level are central (I10, I11). These cooperative undertakings include partnerships across the public and private sectors, academia, and civil society for the immediate implementation of local projects (I9, I10, I11). Cooperation is thus project-specific and geared towards incorporating locally anchored stakeholders who are affected by specific policy issues. London, for example, has built its 24 h policy approach on inner-city partnerships and focuses on establishing good cooperation with all its boroughs (I14). For Newcastle, a member of the project teams explains that “a series of methods” have been developed as part of the platform, “to encourage that sort of cross-sector cooperation and partnerships” (I11). Beyond the city scale, all three examples are embedded in national and international city networks. This collaboration with other cities is geared largely towards the exchange of knowledge and experiences (I10, I11, I14).

Local, national, and international platforms allow cities to position themselves as pioneers (I10, I11, I14). This helps them gain visibility and recognition for their forward-thinking role and results in inquiries from other cities seeking to learn about their experiences in the respective policy fields. We thus conceptualize the aligned policy type with urban focus as ‘first practices’ (Table 2).

Policy divergence with global focus: defiant practices

The third USD policy type emerged from examining urban migration policies in Barcelona and Amsterdam. Barcelona formulated an inclusive and radical urban asylum approach (sometimes referred to as ‘Barcelona City of Refuge’) amidst the so-called European refugee crisis of 2015–2016. It called for the direct reception of refugees crossing the Mediterranean Sea and for shared responsibility and city-to-city refugee relocations between Southern and Northern European cities. This proactive approach openly challenged the national and European asylum policies11,29. In a similar vein, Amsterdam developed its ‘24 h Reception for Undocumented Migrants’ plan, which provides a very comprehensive support structure for irregular migrants in terms of access to housing, health care, legal counselling and police protection30.

Both cities’ actions provoked open conflict with the national government. Whereas the cities adopted immigration and asylum policies that supported people regardless of their residency status, their national governments wanted them to pursue integration only of people who possessed a formal residency status. A further decoupling from the national level can be seen in the interpretation of policy authority, as both cities challenged the national authorities. Amsterdam explicitly defied national migration policy by emphasizing the city’s responsibility to guarantee basic needs (including housing) for everyone who lives in the city31. An expert explains that the city “denounced the Spanish and EU migration policies to claim a bigger role for cities in the governance of migration” (I7). Such initiatives directly conflict with the national policy thrust while addressing an issue of global nature. Hence, this policy type, shown in the lower left quadrant of Table 1, qualifies as ‘policy divergence with global focus’.

To navigate the contested migration policy context, Barcelona and Amsterdam have organized themselves in city networks to build up pressure from below. Barcelona is active in creating alliances and engages in strong networks with other Spanish cities, as well as with cities throughout Europe (I6, I7). Parallel to other efforts within Spain, Barcelona’s mayor, Ada Colau, has openly called on European countries to accept more asylum seekers29. Barcelona has also entered into direct discussions with EU representatives, which have enabled the city to leapfrog the Spanish national government (I6, I7). A city official explains the intentions: “We are very active in city-networks because we often feel that the Spanish state is not helping much (…). If we do not have the support of the national government, let’s go abroad, let’s go to other cities, let’s go the EU” (I6). Furthermore, the interview partner stresses the importance of accompanying and supporting such an urban migration policy with a political positioning strategy: “It is was not only a practical operation, but also political operation to gain more autonomy” (I6).

Meanwhile, Amsterdam has invested in joining forces with Dutch cities to create a broad urban alliance to directly negotiate migration policy issues with the Dutch state (I8). Dutch cities’ unified position on migration policy allowed them to enter into negotiations with the national government in 2017 (I8). The resulting ‘24 h Reception for Undocumented Migrants’ policy represents a compromise between Amsterdam and the Dutch government: “The compromise was more about the process of dealing with undocumented migrants than the actual policy, in which we actually disagree [with the national government]” (I8). However, Amsterdam is determined to make use of its implementation discretion to secure the basic needs of people without a regular residency status (I8).

Both Barcelona and Amsterdam seek to stretch, counter or circumvent their policy-making limitations in migration policy. Given this conflictual policy-making setting, cities have networked with other cities throughout Europe to create a critical mass and to articulate a specifically urban perspective on migration policy. Oomen et al. (2022) describe these strategies as defiant, in that local governments are directly, sometimes vocally, opposing policies at the national level. Accordingly, we refer to this positioning strategy as ‘defiant practices’ (Table 2).

Policy divergence with urban focus: dodging practices

The fourth USD policy type emerged from our examination of Preston, London, and Milan. In Preston, we focus on their measures linked to Community Wealth Building, which are directed towards embedding and democratizing economic processes at the local level. Through these measures, the city aims to strengthen locally anchored institutions (e.g. universities and hospitals), as well as to leverage staff and public procurement processes to support the local population and small businesses32. Moreover, Community Wealth Building promotes so-called public-common partnerships in which political authorities work with civil society organizations to provide public services. In London and Milan, we highlight measures that help to provide more public spaces. In London, such approaches are linked to the regulation of privately owned public spaces so as to prevent the privatization thereof, while in Milan, measures are geared towards reclaiming streets for slow traffic and public recreation.

These examples illustrate locally embedded measures on specifically urban issues. They are typically developed in cities because the impact of austerity policies and the commercialization of public spaces manifest themselves there12,33. An expert explains that “Cities certainly create the conditions for people to come together to contest some of those conditions of capital accumulation, the concentration of those issues (…). This creates the conditions for this kind of politics.” (I13). These interventionist approaches conflict with liberal, free-market policies at the national level. However, as these challenges are prevalent only at the urban scale, the cities’ approaches contest national policy only indirectly. Especially the examples of Community Wealth Building “are beneath the radar of national politics”, as an expert puts it (I13). Accordingly, the cities try to circumvent national-level policy paradigms rather than directly challenging them; the focus is primarily on improving local conditions. Hence, in contrast to the former policy type (policy divergence with global focus), there is no conflict in policy authority because of the specific urban relevance of the policy area. We therefore label this policy type, located in the bottom right quadrant of Table 1, as ‘policy divergence with urban focus’.

The local focus of these policies makes cooperation with other stakeholders at the metropolitan scale essential (I12, I13, I15, I16). A city official explains that the expansion of bike lanes beyond city boundaries “gives an alternative to many people not just within the city but also outside the city to better connect with the inner core” (I16). The main goal of these cities is to build strong local partnerships so that the cities have the resources and flexibility to independently solve emerging problems. Cooperation with actors outside metropolitan borders is of secondary importance. Nevertheless, the cities in these examples do engage in national and international networks with other cities. Here, the goals are exchange of knowledge and engagement in international debates on these issues (I12, I13, I15, I16). Whereas cooperation with higher-tier governments seems generally less important, the example of Milan shows that cooperation with the national level can still prove useful. Despite an apparent policy incongruence, the Italian government supported Milan’s initiative by financing public provision of bicycles and making regulatory changes to enable bike boxes and protected bike lanes (I16).

As noted above, the cities in these examples are reacting to inherently urban problems. Accordingly, they invest mainly in creating local partnerships and in horizontal networks with other cities to exchange knowledge and experiences. Cooperation with the national level may help to remove possible legal obstacles that could derive from decoupled policy agendas. Here again we find inspiration from Oomen et al. (2022), who describe these strategies as dodging, since the local governments do not necessarily proclaim their opposition to the national government but rather circumvent the national policy thrust and, in doing so, expand their local authority. Consequently, we label this positioning strategy as ‘dodging practices’ (Table 2).


Our analysis highlights that the different policy types require different positioning strategies. Policy types represent the contextual setting, which influences the positioning strategy cities choose, the actors they cooperate with, and the scales on which they cooperate. Through this analysis, we detected interesting patterns. In the context of national and urban policy divergence, cities tend to jump scales instead of building (linear) connections across them. Cities with SD policies that diverge from national agendas seek cooperation with other (international) cities to deliberately leapfrog the national level. However, when cities promote policies that are aligned with the national policy framework, a bridging of scales along the formal vertical lines of authority seems to be the most pragmatic and opportunistic approach, and it is also in the interest of the national government. These findings corroborate existing MLG literature showing that cities make use of versatile strategies to target different actors in the context of specific policies11,20.

Although a specific policy type certainly informs a city’s positioning and cooperation strategy at different governance scales, we recognize that this typology glosses over real-life complexities and that the categories are not as clear-cut in reality as we have presented them. Indeed, the differences between them can be marginal at times. Furthermore, the positioning strategies described in this paper are linked to specific policy measures and do not necessarily represent the overall positioning of the city. Cities formulate diverse measures and agendas depending on the policy issue at stake. We therefore do not intend to portray this connection between the policy type and the corresponding strategies as a purely deterministic relationship. Nevertheless, examining these ideal-typical positioning strategies is important for understanding how cities aim to put specifically urban takes of SD on global agendas.

In conclusion, this article has examined and compared different SD policy measures in 11 European cities. While appreciating the value of the multi-level framework in the sustainability transition literature10, we argue that a MLG perspective can enrich the theorization of relations, contestations and strategies between levels of government. We therefore contribute to the recent literature on (contested) relations between local governments and higher governmental authorities11,13,16 by specifically focusing on SD policy-making and introducing a framework to understand national–urban relations and multi-level policy-oriented strategies.

The framework suggests four ideal types of USD policy (Table 1) and maps out their corresponding cooperation and positioning strategies (Table 2). The different strategies reveal how cities can take ownership of certain policy issues and how they can strengthen collaborations across scales or, if necessary, defend their position in potential MLG contestations. We have shown that cities frequently formulate their own SD policies, which often reflect a specifically urban take on SD. To support their policy takes, cities build collaborative partnerships at the local, regional, national and international levels. Cities are thus not mere implementers of national or global SD agendas but active policy-makers and policy influencers.

Not all the policy types and strategies in our analysis receive the same level of attention in academic literature and policy debates. Existing discussions of SD policies tend to focus on certain policy domains such as climate change or social innovation. In these domains, we usually observe alignment with the national level’s perception of how SD should be implemented in cities. Divergent urban policy examples, such as urban migration policies or measures for Community Wealth Building, receive less attention in SD debates. However, these types of urban policy are important for challenging the common perception that USD is geared primarily towards environmental measures14,34 and in showing that cities can bring forth more radical and inclusive visions of SD5. These types of social or economic urban policies are defended in MLG contestations through broad coalitions of urban actors11,29. We therefore believe it is important for further research to highlight these more diverse—and sometimes contested—approaches that cities are taking to address global SD issues.

We suggest that future studies should test and consolidate our findings. For example, this analysis focuses mainly on the relationship between the local and national governments. This delimitation does not do justice to the complexity of the actors’ networks and ignores the important role of local communities, social movements, and epistemic communities involved in addressing SD agendas. Similarly, we have not delved into the comparative differences which exist among path-dependent legal systems and governance structures. Moreover, all the selected vignettes are European case studies of high visibility. Contrasting these findings with observations of smaller, lesser-known cities, especially ones outside Europe, could be an interesting future contribution.

Although the frameworks presented in this article provide an incomplete glimpse of the diversity of USD policies and strategies, we nevertheless think that our analysis yields interesting results. We have demonstrated that there exist specifically urban take to SD, and that these approaches sometimes oppose or contradict national (or even international) SD policy agendas, especially when they challenge national-level policy-making competencies or focus on serving those who are most disadvantaged.


Case selection

We selected our 12 case studies based on the rationale that they should present USD policy measures with different thematic focuses—that is, social, environmental and economic policy measures. Of the 12 case studies, four focus on the environmental dimension of sustainability (Stockholm. Sonderborg, Capital Region of Denmark and Berlin), four on the social dimension (Barcelona, Amsterdam, London and Milan), and four on the economic dimension (Newcastle, Vienna, London and Preston). We selected the cases because they are covered in the academic literature (in urban studies and related fields, such as planning, public policy or sustainability studies) and/or are highlighted as examples on the websites or in the reports of international city networks. We thus purposefully selected USD policy measures of high visibility. With regard to finding the international city networks relevant for our research, we consulted overview articles such as Acuto and Leffel35 and conducted additional web searches.

Interviews (data collection, interview questionnaire, coding)

The data for the analysis of the 12 cases came from a literature review and 16 semi-structured interviews. As suggested by Berry36 and Rathbun37, we first reviewed all publicly available academic and grey literature on these cases before conducting the interviews. This ‘frontloading’ of desk research helped to pose more targeted and informed interview questions. We conducted at least one interview per case because we wanted to ensure the triangulation of non-reactive data (i.e., written documents) with reactive data (i.e., interviews)38,39. We conducted semi-structured interviews with 16 experts involved in the implementation of these policy measures (see Supplementary Table 1 for the list of interview partners). Semi-structured interviews were considered the most suitable method because they allowed us to better understand nuances within cases and obtain information about policy processes37. In the interviews, we could ask specifically about the processes behind USD policy measures and positioning strategies (how?) and the reasons for adopting them (why?). Without conducting these interviews, we would have been forced to rely on written statements about the USD policy measures, and the positioning strategies would have remained largely a black box.

All three authors conducted the interviews, using the same interview questionnaire that featured the categories of policy design, implementation, cooperation and positioning strategy (see Supplementary Table 2 for the interview questionnaire). The interviews lasted between 30 and 90 min. We conducted them by videoconference (and once by phone) and recorded them if we had the interviewee’s approval. Immediately after the interviews, we wrote down interview protocols.

We then thematically coded the interview protocols as well as the gathered literatures (see Supplementary Table 3 for a summary of the interviews)40,41. We then categorized and conceptualized the cases, and we discussed the results of the different case among all three authors. We explicitly looked for patterns in our coding and interpretation of the data40,41. In the discussion, we recognized the pattern that the USD policy measures differ regarding national-urban goal congruence and the global/urban scale of the measures. We grouped the USD policy measures accordingly, resulting in the four groups displayed in Table 1. In a next step, we examined whether we can detect specific forms of cooperation and positioning strategies inside the four groups of USD policy measures40. While we found specific forms of positioning strategies for each group, the distinction of cooperation strategies were less clear-cut and these positioning strategies are not mutually exclusive. We, however, highlight the cooperation strategies that were necessary for the respective positioning strategy. We display these findings in Table 2.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.