Creamer 2015

7/25/2019 Creamer 2015

1/13

Ecological network analysis reveals the inter-connection between soilbiodiversity

and

ecosystem

function

as

affected

by

land

use

acrossEurope

R.E. Creamera,*, S.E. Hannulab, J.P.Van

Leeuwenc, D. Stonea,d, M. Rutgerse, R.M. Schmelze,P.C.de Ruiterg, N.Bohse Hendriksenh, T. Bolgeri, M.L. Bouffaudj, M. Bueek, F. Carvalhol,D. Costal,

T. Dirilgeni, R. Franciscom, B.S. Grifthsn, R. Grif thso, F. Martink,P.Martins da Silval, S. Mendesl, P.V. Moraism, C. Pereiral, L. Philippotj, P. Plassartj,D. Redeckerp, J. Rmbkef,J.P. Sousal, M. Woutersee, P. Lemanceauj

aTeagasc, Johnstown Castle Research Centre, IrelandbNetherlands Institute of Ecology, The NetherlandscWageningen University and Research Centre, The Netherlandsd Leeds University, UKeNational Institute for Public Health and the Environment, The NetherlandsfECT Oekotoxikologie GmbH, GermanygUniversity of Amsterdam, The NetherlandshAarhus University, DenmarkiUniversity College Dublin, Irelandj INRA, UMR 1347 Agrocologie, Dijon, Francek INRA, Laboratory of Excellence Advanced Research on the Biology of Tree and Forest Ecosystems (ARBRE), UMR 1136, Champenoux, France University of

Lorraine, UMR 1136, Champenoux, FrancelCentre for Functional Ecology, University of Coimbra, PortugalmCEMUC and Department of Life Sciences, University of Coimbra, PortugalnCrop and Soils Systems Research Group, SRUC, UKoCentre for Ecology and Hydrology, UKpUniversit de Bourgogne, UMR1347 Agrocologie, Dijon, France

A

R

T

I

C

L

E

I

N

F

O

Article history:

Received 15 April 2015Received in revised form 5 August 2015Accepted 11 August 2015Available online xxx

Keywords:

Soil biodiversityEcosystem functionCarbon cycling and storageNitrogenPhosphorus

Nutrient

cyclingNetwork analysis

A

B

S

T

R

A

C

T

Soil organisms are considered drivers of soil ecosystem services (primary productivity, nutrient cycling,carbon cycling, water regulation) associated with sustainable agricultural production. Soil biodiversitywas highlighted in the soil thematic strategy as a key component of soil quality. The lack of quantitativestandardised data at a large scale has resulted in poor understanding of how soil biodiversity could beincorporated into legislation for the protection of soil quality. In 2011, the EcoFINDERS (FP7) projectsampled 76 sites across 11 European countries, covering ve biogeographical zones (Alpine, Atlantic,Boreal, Continental andMediterranean) and three land-uses (arable, grass, forestry). Samples collectedfromacross these sites ranged in soil properties; soil organic carbon (SOC), pH andtexture. To assess therange in biodiversity and ecosystem function across the sites, fourteen biological methods were appliedas proxy indicators for these functions.Thesemethodsmeasured the following:microbial diversity: DNAyields (molecular biomass), archaea, bacteria, total fungi and arbuscularmycorrhizal fungi;micro fauna

diversity: nematode trophic groups; meso fauna diversity: enchytraeids and Collembola species;microbial function: nitrication, extracellular enzymes, multiple substrate induced respiration,community level physiological proling and ammonia oxidiser/nitrication functional genes. Networkanalysis was used to identify the key connections between organisms under the different land usescenarios. Highest network density was found in forest soils and lowest density occurred in arable soils.Key taxomonic units (TUs) were identied in each land-use type and in relation to SOC and pHcategorisations. Top-connected taxonomic units (i.e. displaying the most co-occurrence to other TUs)were identied for each land use type. In arable sites thiswas dominatedby bacteria andfungi, while in

* Corresponding author.E-mail address: [email protected] (R.E. Creamer).

http://dx.doi.org/10.1016/j.apsoil.2015.08.0060929-1393/ 2015 Elsevier B.V. All rights reserved.

Applied Soil Ecology xxx (2015) xxxxxx

G Model

APSOIL 2259 No. of Pages 13

Please

cite

this

article

in

press

as:

R.E.

Creamer,

et

al.,

Ecological

network

analysis

reveals

the

inter-connection

between

soil

biodiversity

andecosystem function as affected by land use across Europe, Appl. Soil Ecol. (2015), http://dx.doi.org/10.1016/j.apsoil.2015.08.006

Contents

lists

available

at

ScienceDirect

Applied Soil Ecology

journal homepage: www.elsevier.com/locate/apsoi l

7/25/2019 Creamer 2015

2/13

grassland sites bacteria and fungi were most connected. In forest soils archaeal, enchytraeid and fungalTUsdisplayed the largest numberof neighbours, reectingthe greatest connectivity. Multiple regressionmodelswere applied to assess the potential contribution of soil organisms to carbon cycling and storageandnutrientcyclingof specicallynitrogenandphosphorus. Key drivers of carboncyclingweremicrobialbiomass, basal respiration and fungal richness; these three measures have often been associated withcarbon cycling in soils. Regression models of nutrient cycling were dependent on the model applied,showing variation in biological indicators.

2015 Elsevier B.V. All rights reserved.

1. Introduction

Soil

organisms

are

considered

as

drivers

of

ecosystem

services,in

particular

those

soil

ecosystem

services

associated

withsustainable

agricultural

production.

These

include

primary

pro-duction

of

food,

bre

and

fuel,

nutrient

cycling,

carbon

cycling

andstorage,

and

water

inltration

and

purication

(Hooper

et

al.,2005).

As

such,

soil

biodiversity

is

therefore

highlighted

in

the

SoilThematic

Strategy

(EU

(European

Union),

2002) as

a

keycomponent

of

soil

quality.

Soil

quality

is

dened as

the

capacityof

soil

to

function,

within

natural

or

managed

ecosystemboundaries,

to

sustain

plant

and

animal

production,

maintain

orenhance

water

and

air

quality,

and

support

human

health

andhabitation

(Karlen

et

al.,

1997).

Many

of

these

functions

depend

onthe

diversity

and

activities

of

soil

organism

communities.Increasingly

we

require

a

multi-faceted

approach

to

landmanagement,

with

an

increasing

need

for

greater

food

production,while

simultaneously

delivering

other

ecosystem

services

or

soilfunctions,

such

as

carbon

(Tardy

et

al.,

2015)

and

nutrient

cycling(Fierer

et

al.,

2012).

Land

management

can

lead

to

the

degradationof

carbon

stocks

in

soils,

and

therefore

understanding

the

role

ofsoil

biota

in

carbon

cycling

and

storage

is

vital.

The

soil

carbon

poolis

3.3

and

4.5

times

the

size

of

the

atmospheric

(760

Gt)

and

thebiotic

pool

(560

Gt),

respectively

(Lal,

2004). It

is

essential

from

aclimate

change

perspective

that

we

protect

carbon

storagepotential

in

our

soils,

furthermore,

active

cycling

of

carbon,combined

with

large

amounts

of

organic

carbon

temporarily

stored in soils, increases primary productivity, stabilises soilstructure,

increases

nutrient

retention

and

water ltration (Turbet

al.,

2010

De

Vries

et

al.,

2013). Land

management

also

has

asignicant

impact

on

the

capacity

of

the

system

to

cycle

nutrients,providing

a

constant

supply

to

crops

as

needed

to

ensure

optimumproductivity.

This

has

traditionally

been

a

high

input

system,

withthe

addition

of

synthetic

fertilisers

to

promote

availability

ofessential

nutrient

for

plant

growth

(especially

nitrogen

(N)

andphosphorus

(P)),

however

it

is

becoming

increasingly

apparentthat

soil

organisms

have

a

strong

role

to

play

in

the

cycling

ofnutrients

due

to

their

involvement

in

the

geochemical

cycles(Lemanceau

et

al.,

2015).In

2012,

the

European

Commission

acknowledged

the

impor-tance

of

soil

biodiversity

in

the

role

of

ecosystem

functioning,

stating

that these functionsareworthyofprotection becauseof theirsocio-economic

as

well as

environmental

importance

(Jones

et

al.,

2012).However,

the

lack

of

quantitative

standardised

data

on

soilbiodiversity at the European scale has resulted in poor understand-ingof

both

the

role

that

soil

organisms

playin

soil

ecosystem

servicesand

the

need

to protect

soil

biodiversity

to ensure

the

futureprovision

of

such

functions.

This

was

also

highlighted

in

the

EUs

6thFramework programme nanced project: environmental assess-ment

of

soil

for

monitoring

(ENVASSO)

established

in

2005,

thatrecommended

pan-European

indicators

to assess

the

potential

lossof

soil

biodiversity

(Bispo

et

al.,

2009).

This

work

has

been

followedup by the Ecological Function and Biodiversity Indicators inEuropean

Soils

(EcoFINDERS)

project,

nanced

under

the

EUs

7thFramework

programme

and

established

in

2009,

to support

the

European

Union

soil

policy

making

byproviding

the

necessary

toolsto

design

and

implement

strategiesfor

sustainable

useof

soils,

with

aspecic

focus

on

soil

biodiversity

and

associated

ecosystemfunctioning.

There

have

been

many

studies

which

have quantied the impactof

land

management

and

land

use

on

the

diversity

and

functioning

ofsoil

biota

(afewexamples

include;

Trasar-Cepedaetal.,

2008;

Lohauset

al.,

2013;

Mills

and

Adl,

2011;

Bartz

et

al.,

2014).

Tsiafouli

et

al.(2015)

highlights

the

lack

of

integrative

approach,

with

many

ofthese

studies

focussing

on

one

aspect

of

soil

biodiversity

(e.g.

speciesrichness,

abundance,

food

webs,

community

structure),

promotingthe

need

for

more

multi-factorial

approaches.

Tsiafouli

et

al.

(2015)analysed

the

effect

of

agricultural

intensication across Europe onthe

structure,

diversity,

food

web

assembly

and

communitydynamics

of

soil

biota,

summarising

that

agriculture

intensicationreduces

soil

biodiversity,

resulting

in

fewer

functional

groups

andreduce

diversity.Traditional

methods

such

as

diversity

estimates

and

multivari-ate

statistical

techniques

describe

beta-diversity

and

can

reveal

therole

of

biotic

and

abiotic

factors

in

shaping

the

communities.However,

they

do

not

take

into

account

the

interactions

amongorganisms,

a

very

important

factor

shaping

any

natural

community(Bohan

et

al.,

2013;

Mulder

et

al.,

2011).Much

of

the

focus

in

natureconservation

has

been

on

protection

of

individual

species

whilebiotic

interactions

are

increasingly

at

risk

from

local

and

globalextinction

as

a

consequence

of

(anthropogenic)

environmentaldisturbances

(Pocock

et

al.,

2012).

Using

a

network

based

approach, the relationship between organisms within and acrosstaxonomic

units/trophic

levels

can

be

analysed

even

from

verylarge

datasets.

In

ecology,

networks

have

been

long

used

formacro-organisms

(Bascompte

et

al.,

2003) but

recently

theapproach

of

analysing

large

datasets

using

summarizing

networkanalysis

based

on

ecological

theories

has

become

popular

in

theeld

of

soil

microbial

ecology

(see

for

example

Barbern

et

al.,2012).

The

aim

of

this

study

was

to

investigate

the

biological

diversity(soil

microbial

and

faunal

communities)

associated

with

majorland

use

management

types

found

across

Europe

and

to

examinehow

these

various

ecological

networks

relate

to

twokey

ecosystemservices

in

soil;

(1)

carbon

cycling

and

storage

potential

and

(2)nutrient

cycling,

specically

nitrogen

(N)

and

phosphorus

(P).

To

achieve this, a pan-European transect was sampled in 2011 at81

sites,

across

11

European

countries,

covering ve biogeographi-

cal

zones

(Alpine,

Atlantic,

Boreal,

Continental

and

Mediterranean)and three land use types (arable, grass, forestry) (Stone et al., 2015,this

issue). These

sites

represent

a

wide

range

of

soil

properties,specically chosen to provide a wide spectrum of measurementsfor

SOC,

pH

and

texture

(sand,

silt

and

clay

content).

Fourteen

soilbiological properties were measured: (i) microbial diversity; DNAyields

(molecular

biomass),

archaea,

bacteria,

fungi,

arbuscularmycorrhizal

fungi

(AMF),

(ii)

micro

fauna

diversity;

nematodestrophic

groups,

(iii)

meso

fauna

diversity;

enchytraeid,

andCollembola species, (iv) functional indicators; nitrication, extra-cellular

enzyme

assays

(EEA),

multiple

substrate

induced

respira-tion

(MSIR)

and

community

level

physiological

proling (CLPP),

2 R.E. Creamer et al./Applied Soil Ecology xxx (2015) xxxxxx

G Model

APSOIL 2259 No. of Pages 13

Please

cite

this

article

in

press

as:

R.E.

Creamer,

et

al.,

Ecological

network

analysis

reveals

the

inter-connection

between

soil

biodiversity

andecosystem function as affected by land use across Europe, Appl. Soil Ecol. (2015), http://dx.doi.org/10.1016/j.apsoil.2015.08.006

7/25/2019 Creamer 2015

3/13

and

the

abundance

of

key

functional

genes

involved

in

ammoniaoxidation

and

denitrication.

Network

analysis

was

used

toidentify

the

key

connections

between

organisms/trophic

groupsunder

the

different

land

use

management

types

and

multipleregression

analyses

were

then

employed

to

examine

the

relation-ship

between

the

various

organisms/trophic

groups

and

soilfunctions.

2. Methods

2.1. Site sampling and sample processing

81

sites

were

sampled

across

Europe,

as

part

of

the

EcoFINDERSproject;

this

study

is

known

as

the

European

transect (Fig.

1).Sites

were

selected

from

within

European

Union

countries

using

aspatial

random

sampling

model,

weighed

to

derive

a

spectrum

ofsites

representative

of

the

range

of

soil

properties;

(SOC,

texturalclass

(representing

by%

clay)

and

pH),

land-use

and

biogreo-graphical

zones

across

Europe

(EEA,

2012).

Data

used

to

spatiallyderive

potential

locations

for

sampling

were

based

on

theEuropean

Food

Safety

Authority

(EFSA)

database,

the

Corinelandcover

map

and

soils

database

(Gardi

et

al.,

2011)

and

theEuropean

Environment

Agency

map

of

biogeographical

zones

(EEA, 2012). Full details of the development of the site selection

model

and

sampling

can

be

found

in

Stone

et

al.

(2015). In

brief,

soilwas

sampled

from

each

site

following

a

pre-agreed

standardoperating

procedures

(SOPs)

within

EcoFINDERS,

guaranteeingthat

all

sites

were

sampled

in

a

consistent

manner.

Soil

was

takenfrom

the

top

5

cm

of

the

prole using plastic cores. All cores werepacked

in

pre-labelled

bags

and

posted

(24

h

delivery)

in

cooledboxes

to

Teagasc

Research

Institute,

in

Ireland.

On

receipt,

soilswere

sieved

to

7/25/2019 Creamer 2015

4/13

following

(ISO

(International

Organization

for

Standardization),1995)

and

organic

carbon

(OC)

was

determined

by

LECO

elementalanalysis,

this

was

conducted

on

0.25

mm

milled

dry

soil

sub-samples

(Massey

et

al.,

2014). Cation

exchange

capacity

(CEC)

wasmeasured

using

BaCl2 extraction method (ISO (InternationalOrganization

for

Standardization),

1994). pH

was

measured

in

a1:2.5

soil

in

water

suspension

using

a

glass

electrode

(vanReeuwijk,

2002).

N

mineralisation

was

analysed

using

the

Illinoissoil

nitrogen

test

for

amino

sugar-N

(McDonald

et

al.,

2014).

ThisIllinois

soil

nitrogen

test

(ISNT)

method

was

developed

by

Khanet

al.

(2001)

and

modied

by

Klapwyk

and

Ketterings

(2005)

toestimate

the

amount

of

amino-sugars

plus

NH4-N in the soil. Theconcentration

of

ISNT-N

liberated

by

NaOH

and

captured

as

NH4-Nby

the

boric

acid

was

quantied

by

colorimetric

analysis

with

anAquakem

600A

(Aquakem

600A,

1621, Vantaa,

Finland).

Phospho-rus

was

measured

using

the

Mehlich

3

methodology

(Mehlich,1984)

and

analysed

on

a

Varian

Vista

MPX

ICP-OES.

2.3.

Measurements

of

soil

biodiversity

2.3.1. Microbial diversity

The

methodology

used

for

phospholipid

fatty

acids

(PLFA)extraction,

separation,

transmethylation

and

GC

analysis

was

the

MIDI PLFA hybrid method described by Francisco et al. (2015, thisissue).

Briey, soils were lyophilized and lipids extracted using theBligh

and

Dyer

(1959)

extraction

procedure.

Lipid

extracts

wereseparated

by

solid-phase

extraction

(SPE)

using

an

SI-column

andorganic

solvents

as

eluents.

Phospholipids

were

eluted

withmethanol.

Phospholipids

were

derivatised

and

transmethylatedusing

the

MIDI

FAME

protocol

(MIDI,

Inc.,

Newark,

DE,

UnitedStates).

Fatty

acid

methyl

esters

(FAME)

were

measured

by

GasChromatography

(GC)

(Agilent

Technologies,

Wilmington,

DE,USA),

identied and quantied using standards (internal FAME19:0

and

calibration

mixtures)

and

Sherlock

MIS

data

base,

basedon

the

calculated

equivalent

chain

lengths

(ECL).

The

biomarkerswere

dened according to Francisco et al. (2015, this issue).DNA

for

all

molecular

work

was

extracted

using

the

method

described in Plassart et al. (2012). Crude DNA extracts wereresolved

by

electrophoresis

in

gel,

stained

with

ethidium

bromideand

a

standard

curve

of

DNA

was

used

to

estimate

the nal DNA

concentration

in

the

extracts

allowing

the

assessment

of

so-calledmolecular

microbial

biomass

(Dequiedt

et

al.,

2011).

In

this

paperthis

will

be

referred

to

as

molecular

microbial

biomass.

Aftermicrobial

DNA

extraction,

Terminal

restriction

fragment

lengthpolymorphism

(T-RFLP)

was

applied

to

measure

the

threemicrobial

domains

(bacteria,

archaea,

fungi),

based

on

the

lengthand

abundance

of

unique

restriction

fragments

found

in

eachsample.

Bacterial

and

archaeal

T-RFLP

community

proles weregenerated

by

amplifying

specic 16S

rRNA

gene

sequences,

whilethe

fungal

T-RFLP

community

proles

were

generated

byamplifying

the

ITS1-ITS4

region

as

described

by

Grifths et al.

(2011)

and

Plassart

et

al.

(2012).To

determine

fungal

richness

and

relative

frequency,

fungalITS2

region

was

amplied

from

these

metagenomic

DNA

samples(Ihrmark

et

al.,

2012).

Amplicon

libraries

were

pyrosequenced

andfungal

community

diversity

was

generated

from

the

analysedsequences

as

described

by

Coince

et

al.

(2014).Fungal

copy

numbers

were

determined

using

the

same

primersas

for

454-pyrosequencing

using

real-time

PCR

mix

from

Rotor-Gene

SYBR

Green

PCR

Kit

(Qiagen).

T4

Gene

32

protein

(Roche)

wasused

to

enhance

the

reaction

and

ensure

similar

amplication

fromall

soils.

The

samples

were

analysed

on

a

Rotor-Gene

3000

machine(Gorbett

Research,

Sydney,

Australia).

The

reaction

mixtures

wereprocessed

using

a

pipetting

robot

(Gorbett

Research,

Sydney,Australia)

in

20

ml

volume

and

contained

0.3

mM

each

primer,

0.25

ml

T4

and

1.0

10.0

ng

template

DNA.

The

cycling

conditions

were: 40

sec

at

95 C,

1

min

at

58 C

and

1

min

at

72 C.

Plasmidsextracted

from

pure

fungal

cultures

were

serial

diluted

and

used

asa

reference

for

the

copy

numbers.

As

ITS2

region

can

vary

in

length,three

different

plasmids

extracted

from

three

different

specieswere

used

as

standards.

All

samples

were

analysed

in

at

least

twodifferent

runs

and

in

two

different

concentrations

to

conrm

thereproducibility

of

the

quantication

and

lack

of

inhibition

due

toi.e.

humic

acids.To

analyse

AMF

diversity,

nested

PCRs

were

performed

on

threereplicates

from

all

samples.

The

rst

PCR

was

performed

using0.4

U

of

Phusion

High

Fidelity

DNA

polymerase

(Thermo

FisherScientic, Courtaboeuf, France), 1x Phusion HF buffer, 0.5mM ofthe

primers

SSUmCf

and

LSUmBr

(Krger

et

al.,

2009),

0.2

mM

ofeach

dNTPs

and

1

ml

of

genomic

DNA,

in

a

nal

volume

of

20

ml.The

PCR

conditions

used

were

5

min

at

99 C,

35

cycles

of

10

s

at99 C,

30

s

at

63 C

and

1

min

at

72 C,

followed

by

10

min

at

72 C,using

an

Eppendorf

Mastercycler

epgradient

S

(Vaudaux-Eppen-dorf,

Schnenbuch,

Switzerland).

The

nested

PCR

was

done

using1

U

of

Phusion

High

Fidelity

polymerase,

1 HF

buffer,

0.5

mM

ofthe

primers

ITS3m

(Zhong

et

al.,

2010) and

ITS4

(White

et

al.,

1990)with

barcodes,

0.2mM of each dNTPs and 2ml of PCR product

diluted

at

1:50,

in

a

total

volume

of

50

ml.

PCR

conditions

were

30

sat

98 C,

30

cycles

of

10

s

at

98 C,

30

s

at

64 C

and

20

s

at

72 C,

followed by 10 min at 72 C, in an Eppendorf Mastercyclerepgradient

S.

The

three

PCR

replicates

of

each

sample

were

pooledand

puried

using

the

High

Pure

PCR

Product

Purication

Kit(Roche

Applied

Science,

Meylan,

France)

following

the

manufac-turers instructions. After quantication with Picogreen, thepuried PCR products were mixed equimolarly to preparesequencing

libraries.

The

libraries

were

sent

to

Beckman

CoulterGenomics

(Grenoble,

France)

for

sequencing

using

454

GS

FLXtechnology.

2.3.2.

Micro-

and

meso-fauna

Enchytraeids

were

extracted

from

three

replicate

soil

cores(5.0

cm

depth 5.0 cm width) per site with OConnors hot/wet

funnel

method

(OConnor,

1962)

following

ISO

standards

(ISO

(International Organization for Standardization), 2006). Speci-mens

were

identied to species using light-microscopically in vivo,applying

the

keys

and

techniques

in

Schmelz

and

Collado

(2010,2012),

together

with

primary

literature.Collembola

were

extracted

from

three

replicate

soil

cores(5.0

cm

depth 5.0 cm width) per site. The samples were

transferred

by

courier

mail

to

IMAR,

University

of

Coimbra,Portugal,

where

they

were

extracted

using

a

modied

MacfadyenHigh

Gradient

Extractor

(Macfadyen,

1961)

for

seven

days.Collembola

were

mounted

on

slides

and

identied to specieslevel,

in

most

cases,

using

primary

literature

on

EuropeanCollembolan

identication.Nematode

trophic

groups

were

determined

by

morphologicalanalysis,

using

a

Doncaster

counting

plate

(Doncaster,

1962),

and

identied

to

trophic

level

(plant-parasites,

bacterial-feeders,fungal-feeders,

omnivores

and

predators)

by

observing

thehead/mouth

structures

under

an

inverted

microscope

(100

and200 magnication).

2.4.

Measurements

of

soil

biological

functioning

To

determine

multiple

substrate

induced

respiration

(MSIR)proles, the MicroResp methodology adapted from Campbell et al.(2003)

and

reported

by

Creamer

et

al.

(2009)

was

applied

in

thisstudy.

A

spectrum

of

seven

substrates

was

selected: D-(+)-galac-

tose, L-malic acid, gamma amino butyric acid, n-acetyl glucos-

amine, D-(+)-glucose, alpha ketogluterate, citric acid and water for

basal

respiration

measurements.

Details

of

the

methodology

are

described

in

Creamer

et

al.

(2015,

this

issue).

4 R.E. Creamer et al./Applied Soil Ecology xxx (2015) xxxxxx

G Model

APSOIL 2259 No. of Pages 13

Please

cite

this

article

in

press

as:

R.E.

Creamer,

et

al.,

Ecological

network

analysis

reveals

the

inter-connection

between

soil

biodiversity

andecosystem function as affected by land use across Europe, Appl. Soil Ecol. (2015), http://dx.doi.org/10.1016/j.apsoil.2015.08.006

7/25/2019 Creamer 2015

5/13

Community

level

physiological

proles (CLPP) using BiologECO-plates

were

analysed

in

the

European

transect

and

theNetherlands

Soil

Monitoring

Network

(Rutgers

et

al.,

this

issue).This

method

is

considered

to

determine

multiple

functionalendpoints

represented

by

a

sample

of

the

heterotrophic

soilbacterial

community

(Winding

et

al.,

2005).Extracellular

enzyme

activities

(EEA)

in

the

soils

were

deter-mined

on

8

different uorogenic model substrates related to the

hydrolysis

of O-glycosyl linkages of ve di- and poly-saccharides

including

starch,

cellulose,

hemicellulose

and

chitin,

ester

linkagesof

organic

phosphates

and

sulfates

and

peptide

linkages

ofproteins.

The

assay

was

performed

in

microtiter

plates

as

describedby

Johansen

et

al.

(2005).Potential

nitrication

was

measured

using

the

method

de-scribed

by

Kandeler

(1996)

but

adapted

to

the

microplate

(Ng

et

al.,2014).

Soil

samples

(2

g

moist

soil)

were

incubated

for

5

h

onrotatory

shaker

at

room

temperature

in

20

ml

(NH4)2SO4(10 mM)and

0.1

ml

NaClO

(1.5

M).

Control

was

kept

at

20 C

duringincubation

and

thawed

at

room

temperature

after

incubationperiod.

After

incubation,

6

ml

KCl

(2

M)

solution

was

added

tosamples

and

controls

and

shaken

(30

min)

followed

by

centrifuga-tion

(4

min,

3000

rpm).

5

ml

of

ltrate

was

mixed

with

3

ml

NH4Cland

2

ml

colour

reagent

(2

g

sulphanilamide

and

0.1

g

N-(1-

naphthyl)-ethylenediamine hydrochloride in 150 ml distilledwater

and

20

ml

concentrated

phosphoric

acid)

and

allowed

to

stand

(15

min,

room

temperature).

NO2 was

measured

spectro-photometrically

(OD

540

nm)

on

a

microplate

reader.

The

NO2-N

concentration

was

calculated

using

a

calibration

curve

made

with

astandard

solution

of

NaNO2(10mg NO2-N

ml1).Quantication of the bacterial and archaeal ammonia-oxidizers

(AOA

and

AOB)

and

of

the

nitrous

oxide

reducers

(nosZ1

andnosZ2)

was

performed

according

to

Tourna

et

al.

(2008), Leiningeret

al.

(2006)

and

Jones

et

al.

(2013), respectively.

The

real-time

PCRassays

were

carried

out

in

a

ViiA7

(Life

Technologies,

USA)

with

a15

ml

reaction

volume

containing

the

SYBR

green

PCR

Master

Mix(Absolute

Blue

QPCR

SYBR

Green

Low

Rox

Mix,

Thermo,

France),1

mM of each primer, 250 ng of T4 gene 32 (QBiogene, France) and0.5

ng

of

DNA.

Standard

curves

were

obtained

with

serial

plasmiddilutions

of

a

known

amount

of

a

plasmid

DNA

containingfragment

of

the amoA, nosZ1 and nosZ2 genes.

2.5. Statistical procedures

2.5.1.

Network

analysis

for

soil

biodiversity

linkages

To

construct

networks,

TRFLP

data

were

used

for

fungi,

archaeaand

bacteria,

species

numbers

for

Enchytraeids

and

Collembola,trophic

groups

for

nematodes,

and

amplicon

sequence-datagrouped

into

family

level

for

AMF

(Table

1biodiversity).

All

of

the levels are further considered as taxonomic unit. Taxonomicunits (TU) that were present in only one sample per category (land

Table 1

Biological Indicators applied at 81 sites across Europe, to assess soil biodiversity or functions; C-cycling and nutrient (N&P) cycling.

Main Indicator Measures Paper recommending indicator

C storage and cyclingExtracellular enzyme activity(EEA)

Beta-glucosidase; sum of enzyme activity Kivlin and Treseder, (2014);Sinsabaugh et al. (2008)

Multiple substrate inducedrespiration (MicroResp)

Basal respiration, L-malic acid, D-(+)-glucose, alpha ketogluterate, PCA1, PCA2 Campbell et al. (2003),Creamer et al. (2015)

Biolog 1/GG50 Rutgers et al. (2015), Rutgersand Breure, (1999)

Phospholipid fatty acids (PLFA) Fungal: bacterial, ergosterol (18:2w6,9), AMF (16:1w5c and 18:1w9c) Francisco et al. (2015);

Herman

et

al.

(2012);Fernandes et al. (2013)DNA yields (ng microbial DNA g soil1) Dequiedt et al. (2011)Enchytraeids

Relative

abundance

of

Enchytraeid

acidity

indicators

Graefe

and

Beylich,

(2003);Cole et al. (2000)

Nematodes Feeding guild richness (plant-feeders, fungal-feeders, omnivores, bacterial-feeders, predators) Grifths et al. (2007)AMF families Acaulosporaceae, Ambisporaceae, Archaeosporaceae, Claroideoglomeraceae, Diversisporaceae,

Gigasporaceae, Glomeraceae,Pascisporaceae, ParaglomeraceaeVan Der Heijden et al. (2008)

Fungal abundance and richness Fungal copy numbers, fungal richness Coince et al. (2014)

Nutrient cycling of N and PEnchytraieds Species richness and Abundance per m2 Cole et al. (2000) (N

mineralisation only)Nematode

Plant-feeders,

Fungal-feeders,

Omnivores,

Bacterial-feeders,

Predators

and

total

abundance

Xiao

et

al.

(2010)Grifths and Bardgett (1997)

Extracellular Enzyme Activity(EEA)

Arylsulfatase, phosphomonoesterase, Leucin aminopeptidase Sinsabaugh et al. (2014)

Biolog L_Arginine, L-asparagine, L-phenylalanine, L-serine, N-acetyle-D-glucosamine, L-threonine, D-glucosaminic acid, glycyl-L-glutamic acid, phenyl-ethylamine, putrescine

Van Eekeren et al. (2008)

Nitrication potential Amount of NO2-N released (ng/g soil dm/h) Schloter et al. (2003)Functional Gene A nosZ1 (denitrier) gene, nosZ2 (denitrier) gene Jones et al. (2013); Philippot

et al. (2013)Functional Gene B AOB (Bacteria: ammonia oxidizers) gene; AOA (Archea: ammonia oxidizers) gene Leininger et al. (2006)Molecular microbial biomass (ng microbial DNA g soil-1) Dequiedt et al. (2011)Fungal abundance and richness Fungal copy numbers, fungal richness Coince et al. (2014)

BiodiversityEnchytraeid Species diversity Rmbke et al. (2013)Nematode Feeding guild richness Donn et al. (2012)

Yeates et al. (1993)Collembola Species richness Bispo, et al. (2009)AMF families Acaulosporaceae, Ambisporaceae, Archaeosporaceae, Claroideoglomeraceae, Diversisporaceae,

Gigasporaceae, Glomeraceae, Pascisporaceae, ParaglomeraceaeHart and Reader (2002)

archaea, bacteria and fungi T-RFLP copies per ng DNA (abundance); Dikarya Richness; Plassart et al. (2012)

R.E. Creamer et al./Applied Soil Ecology xxx (2015) xxxxxx 5

G Model

APSOIL 2259 No. of Pages 13

Please

cite

this

article

in

press

as:

R.E.

Creamer,

et

al.,

Ecological

network

analysis

reveals

the

inter-connection

between

soil

biodiversity

andecosystem function as affected by land use across Europe, Appl. Soil Ecol. (2015), http://dx.doi.org/10.1016/j.apsoil.2015.08.006

7/25/2019 Creamer 2015

6/13

use,

pH

or

SOC)

were

removed

prior

to

analysis.

The

remainingnumber

of

TUs

after

singleton

removal

per

category

is

presented

inSupplementary

Table

A

and

the

numbers

were

used

to

scale

thesizes

of

the

nodes

in

the

network.

Spearman-rank

correlationmatrixes

were

calculated

in

R

(R

Core

Development

Team,

RFoundation

for

Statistical

Computing,

Vienna,

Austria)

usingabundance

data.

Only

signicant

positive

correlations

were

usedin

further

analysis.

The

percentage

of

signicant correlations fromtotal

possible

correlations

was

used

as

a

measure

of

interactionstrength

between

TUs,

and

edge

size

and

darkness

were

scaled

tothis.

Within

taxa

correlations

(groups/species

or

trophic

group)were

calculated

but

not

displayed.

The

data

were

transferred

toCytoscape

(Shannon

et

al.,

2003) for

further

analysis

andvisualization.

Network

density,

clustering

coefcient

and

averagenumber

of

neighbours

was

calculated

using

network

analysis

toolswithin

Cytoscape.

2.5.2.

Potential

C

storage

and

cycling

Using

total

C

content

and

land-management

type

(arable,grassland

and

forest),

sites

have

been

ranked

according

to

C

cyclingand

storage

potential.

For

this

ranking,

coefcients for land usetype

were

calculated

using

linear

regression

with

total

C

asdependent,

and

land

use

type

as

categorical

independent

variables.

The coefcients thus found were: 0.414 for arable elds, 0.917 forgrasslands

and

2.040

for

forests.

These

coefcients were multipliedby

the

amount

of

normalised

SOC

content

in

the

topsoil

(sampledivided

by

the

average)

to

geta

quantication

of

C

storage

potentialcorrected

for

land

use.Carbon

storage

potential

was

subsequently

used

as

a

dependentvariable

in

a

forward

stepwise

multiple

linear

regression. All

suitablebiological

measurements

(Table

1carbon

cycling)

were

included

inthe

analysis,

and

all

non-standardized

parameters were log-trans-formed

before

inclusion

in

the

model.

Based

on

amount

of

explainedvariance,

and

the

Akaike

information

criterion

(AIC),weselected

themost

parsimonious

model

for

predicting

C

storage

potential.

2.5.3.

Nutrient

cycling

of

N

and

P

Three models have been applied to assess the cycling ofnutrients

at

these

sites:

(i)

normalised

N

mineralisation

alone

hasbeen

modelled

to

address

the

contribution

of

soil

organisms

to

thenitrogen

mineralisation

in

soils.

(ii)

a

P

availability

model

usinginversely

normalised

P

availability

(average

divided

by

sample)only

to

assess

the

role

of

soil

biology

to

P

cycling

in

soils.

(iii)

acombined

model

to

assess

overall

nutrient

cycling.

This

uses

theproduct

of

the

normalised

N

mineralisation

(sample

divided

byaverage)

multiplied

by

the

inversely

normalised

P

availability(average

divided

by

sample)

in

the

topsoil

to

geta

quantication ofcombined

N

and

P

cycling.

For

all

three

models

forward

stepwisemultiple

linear

regression

was

applied.

All

suitable

biological

measurements

(Table

1nutrient cy-cling)

were

included

in

the

analysis,

and

all

non-standardizedparameters

were

log-transformed

before

inclusion

in

the

model.Based

on

amount

of

explained

variance,

and

the

Akaike

informa-tion

criterium

(AIC),

we

selected

the

most

parsimonious

model

forpredicting

nutrient

cycling.

This

procedure

was

also

executed

withnormalised

N

mineralisation

and

inversely

normalised

P

availabil-ity

as

dependent

variables

separately,

to

disentangle

the

respectivenutrient

cycles.

Statistical

analyses

were

carried

out

using

SPSS(20.0.0)

and

R

(R

Core

Development

Team,

2012).

3. Results

3.1.

Review

of

soil

properties

and

measurements

In

total,

76

sites

were

analysed

from

the

81

sites

sampled

(Stoneet

al.,

this

issue),

as

these

sites

had

a

complete

set

of

parameters.Table

2

shows

the

range

of

soil

properties

for

the

76

sites;

SOCranged

from

0.45%

to

51.1%

(Fig.

2a),

lowest

mean

SOC

was

found

inarable

sites,

while

the

highest

SOC

concentrations

were

found

inforest

sites.

pH

varied

considerably

across

sites,

ranging

from

3.7

to8.2,

with

the

lowest

mean

pH

found

in

forest

sites

and

highestmean

pH

in

arable

sites

(4.99

and

7.07,

respectively).

Soil

texture

varied across all sites and is represented in this paper by claycontent.

Clay

content

varied

from

7/25/2019 Creamer 2015

7/13

group

of

organisms

and

between

categories,

and

reveals

thestrongly

connected

TUs.ThekeyconnectivityofTUs

occurring

across

thedifferent

land

usecategories

wereillustrated

in

Fig.

3a

which

shows

the

strength

of

theconnection

between

key

TUs.

The

highest

density

networks

werefound

in

the

forest

soils (density

of

0.041)

followed

by grasslands(density

0.027)

and

arable

lands

(0.025)

(Supplementary

Table

A).The

higher

the

density

the

larger

the

number

of

signicantconnections

found.

In

the

arable

sites

the

AMF

families

andnematode

trophic

groups

showed

the

strongest

association.

Thisis

due

to strong

positive

correlation

between

plant-feedingnematodes

and

AMF

(Fig.

3a).

This

connection

is

completely

absentin

grassland

soils

and

only

weak

in

forest

soils.

A

similar

co-

occurrence

can

be

seen

for

AMF

families

and

archaea,

where

aconnection

is

visible

in

arable

systems,

very

weak

in

grasslandsystems

and

missing

in

forest

soils.

There

was

a

strong

connectionbetween

enchytraeid

species

and

Collembola

species

in

arable

andgrassland

sites

but

is

much

weaker

in

forest

soils.

In

the

grasslandsites

the

dominant

connections

were

found

between

the

bacteriaand

archaea

TUs,

with

signicant correlations also found betweenenchytraeids

and

nematodes

with

archaea.

In

forest

systems,

there

isa

strong

connection

between

enchytraeids

and

AMF

families.

Theconnection

between

bacteria

and

fungi,

the

two

largest

TUs

of

soilorganisms

(Francisco

et

al.,

2015,

this

issue),

gets

stronger

when

landuse

intensity

diminishes

(arable< grasslands 30

neigh-bours))

compared

to

the

other

land

use

classes

(SupplementaryTable

A).

The

top

twenty

connected

TUs

present

in

each

land-useclass

are

displayed

in

Fig.

3b,

with

the

number

of

neighbours

(otherTUs

which

show

a

signicant

correlation

with

this

TU)

representedon

they-axis, the higher the number of neighbours the more stable

a

network.

Archaea

(peak

IDs;

trf_246,

and

trf_359

and

bacteria(peak

IDs;

trf_470))

(Fig.

3b),

were

found

to

be

the

top

threeconnected

TUs

in

arable

sites

and

overall

archaea

and

bacteria

TUswere

the

most

connected

in

these

soils.

Only

one

species

ofCollembola

(Isotomaviridis)was found in the top twenty connectedTUs and two fungi TUs (peak IDs; trf_205, trf_266). AMF families,nematode

functional

groups

and

enchytraeid

species

were

notfound

among

the

top

twenty-connected

TUs

in

arable

soils.In

the

grassland

sites,

the

top-connected

TUs

were

bacteria(Peak

IDs;

trf_412,

trf_88,

trf_413,

and

trf_226

(Acidobacteria))

andfungi

(Peak

IDs;

trf_205

and

trf_230)

which

were

found

to

havemore

than

50

neighbours.

Some

enchytraeid

species

(Fridericiacylindrical (CYL), Enchytroniaparva (PAR),) were present in the top-twenty

connected

species

in

grassland

sites. Fridericia cylindrica

was

the

most

connected

enchytraeid

species,

this

species

wasobserved

only

in

grassland

sites.

This

species

was

found

to

have

strong connectivity in sites that were categorized by SOC 215%and

pH

57 (Supplementary graphs A and B), suggesting it is notcommonly

connected

in

more

extreme

environments,

but

ratherprefers

slightly

acidic

to

neutral

pH

and

none

peaty

conditions

(i.e.SOC

7/25/2019 Creamer 2015

8/13

molecular

microbial

biomass

and

fungal

richness,

with

a

Std.

Errorof

the

estimate

of

0.591.Model:

Ln(Rank-C)

=

1.818

+

0.849

Ln(MicroResp_Water)

+0.983

Ln(Molecular Microbial Biomass) 1.920 Ln(FungalRichness).

In

the

network

analysis

used

to

determine

the

key

connectionsof

soil

TUs

in

relation

to

carbon

cycling

and

storage,

sites

were

analysed

according

to

the

SOC

categories

dened in Stone et al.(2015,

this

issue)

which

represents

sites

with

low

carbon

content15%).

The

highest

network

densities

were

detected

in

sites

of

mediumSOC

content

(215%)

(density

of

0.040).

In

sites,

where

soil

SOCexceeded

15%

and

in

sites

with

less

than

2%

SOC,

density

was

very

Fig. 3. (a) Network of biotic interaction based on signicant positive Spearman correlations in each land use type. The nodes are sized to the number of species included in theanalyses. The size and darkness of the connecting edges is sized to the proportion of signicant positive correlations from all possible correlations between taxonomic units in

the land-use type. For arable soils the threshold for positive interaction was Spearman correlation >0.40, for grasslands >0.35 and for forests >0.45, respectively. Land usecategories were classied on the basis of Stone et al., 2015 (this issue). (b) Top 20 connected species in the three land-use categories (i) arable, (ii) grassland and (iii) forests.Colour legend indicates the following; fungi (pink), archeae (purple), bacteria (blue), enchytraeids (green), collembolan (orange), nematodes (light blue-turqouise) and AMF(dark red).

Fig. 4. Network of biotic interaction based on signicant positive Spearman correlations in each organic matter (SOC%) category. The nodes are sized to the number of speciesincluded in the analyses and their darkness is relational to the connectedness of the node to other nodes. The size and darkness of the connecting edges is sized to theproportion of signicant positive correlations from all possible correlations between taxonomic units in each organic matter% category. For soils with organic matter%0.44, for soil with org matter 215% >0.27 and for soils with organic matter >15% >0.77, respectively. Carboncategories were classied on the basis of Stone et al., 2015 (this issue).

8 R.E. Creamer et al./Applied Soil Ecology xxx (2015) xxxxxx

G Model

APSOIL 2259 No. of Pages 13

Please

cite

this

article

in

press

as:

R.E.

Creamer,

et

al.,

Ecological

network

analysis

reveals

the

inter-connection

between

soil

biodiversity

andecosystem function as affected by land use across Europe, Appl. Soil Ecol. (2015), http://dx.doi.org/10.1016/j.apsoil.2015.08.006

7/25/2019 Creamer 2015

9/13

low

(0.023

and

0.024,

respectively)

(Supplementary

Table

B).Interactions

were

more

evenly

distributed

at

sites

with

215%

SOCcontent,

compared

to

low

and

high

SOC

categories

where

someconnections

were

missing.

For

example,

the

connection

betweencollembolan

species

and

enchytraeid

species

was

much

stronger

in15% SOC(12.5%

connections

signicant)

categories

compared

to

215%category

where

only

5.9%

of

the

signicant connections wereevident

between

these

two

groups.As

the

regression

model

highlighted,

the

fungal

richness

waskey

in

explaining

the

variance

accounted

for

in

the

carbon

cyclingand

storage

model.

Using

the

network

analysis

we

can

assess

thecomposition

and

connectivity

of

the

fungal

community

in

thedifferent

SOC

categories.

The

highest

density

of

fungal

rst

degreenetworks

was

found

in

the

category

215% (0.027)

while

in

bothlow

SOC

(15%) the fungi formed much loosernetworks

(densities

of

0.019)

(Fig.

5).

3.4.

Nutrient

cycling

of

N

and

P

Nitrogen

mineralisation

availability

varied

across

the

76

sitesfrom

39

to

1092

(mg

N

kg1dry

soil).

N

mineralisation

was

loweston

average

in

arable

sites

in

the

Mediterranean

region

and

greatest

in the Alpine forest sites (Fig. 2b). Phosphorus availability wasequally

diverse,

ranging

from

3.64

to

451

(mg/l1) between

sites.The

lowest

available

P

was

quantied

in

arable

Mediterraneansites,

while

the

greatest

availability

of

P

was

measured

in

arableContinental

sites

(Fig.

2c).Three

models

were

applied

to

quantify

nutrient

cycling

acrossthe

sites.

The

rst

model

used

normalised

N

mineralisation

data

asthe

dependent

variable

and

was

regressed

against

nine

biological

indicators

chosen

to

represent

nutrient

cycling

(Table

1).

Thismodel

described

a

large

proportion

of

the

variation

across

siteswith

an

adjusted R2 of 0.734 and a Std. Error of the estimate of

0.332.

The

model

included

the

following

signicant parameters:Model

1:

(normalised

N

min)

=7.71 + 0.75 ln(Mol.Microbial.

Biomass)

0.11

Biolog

L_Threonine.P

availability

(Model

2)

only

accounted

for

a

small

amount

ofthe

variability

using

the

same

biological

indicators

as

model

1, withan

adjusted R2 of 0.196 and a Std. Error of the estimate of 2.904. The

model

included

the

following

signicant

parameters:Model

2:

(normalized P) = 12.97 1.93 ln(EEA_phosphomo-

noesterase) 1.03 ln(potential nitrication) + 1.35 ln(EEA_-

Leucin.aminopeptidase) 1.01 ln(Enchy_SpRichness).

A

combined

model

of

N

P

was

applied

to

assess

thecontribution

of

soil

organisms

to

nutrient

cycling

in

general.

Thismodel

resulted

in

an

adjusted R2 of 0.482 and a Std. Error of the

estimate

of

1.978,

including

the

following

signicant parameters:Model

3

(N

P)

=

12.72

+

3.28

Ln(Molecular

microbial

bio-mass)

1.16

Ln(Potential

nitrication)

+

0.55

Ln(AOA)

3.72

Ln(Fungal

Richness).Accounting

for

48%

of

the

variation

found

across

sites,

thismodel

suggests

a

microbial

driven

system

for

cycling

of

nutrients.

4. Discussion

This

study

sought

to

determine

the

covariation

in

TUs

in

soilsacross

three

broad

European

land

uses.

This

has

been

achievedidentifying

the

major

connections

between

TUs

for

the

differentland

uses

and

the

stability

(number

of

connections)

associatedwith

the

land

use

types.

In

addition,

this

paper

has

summarised

key

Fig. 5. Fungal rst degree networks in different organic matter categories. All signicant connections between fungi (red) and the other organisms are depicted here. Thedarkness of the edges is scaled to the interaction strength and the size of the nodes to the average abundance of the TRFs in each organic matter category. Interactions betweenfungal TRFs are not drawn. Colour legend indicates the following; fungi (pink), archeae (purple), bacteria (blue), enchytraeids (green), collembolan (orange), nematodes (light

blue-turqouise)

and

AMF (dark

red).

R.E. Creamer et al./Applied Soil Ecology xxx (2015) xxxxxx 9

G Model

APSOIL 2259 No. of Pages 13

Please

cite

this

article

in

press

as:

R.E.

Creamer,

et

al.,

Ecological

network

analysis

reveals

the

inter-connection

between

soil

biodiversity

andecosystem function as affected by land use across Europe, Appl. Soil Ecol. (2015), http://dx.doi.org/10.1016/j.apsoil.2015.08.006

7/25/2019 Creamer 2015

10/13

indicators

for

large

scale

monitoring

or

measurement

of

twoecosystem

services

(cycling

of

carbon,

nitrogen

and

phosphorus).

4.1. Co-occurrence of soil biota in different land use types

The

density

of

network

connections

provide

a

useful

insight

into the potential food web dynamics taking place in soils acrossEurope and how these change with land use. Assessment of thethree

land-use

types

(arable,

grass

and

forest)

clearly

showed

thatland

use

intensication resulted in lower density networks, areduction

in

the

strength

of

the

connections

between

bacteria

andmost other TUs (with the exception of collembolan and AMF) andan

overall

reduction

in

the

average

number

of

neighbours.

Forestsoils

displayed

the

greatest

density

of

network

connections

of

thethree

land

use

categories,

suggesting

a

more

stable

system

with

astrongly developed food web in place (Digel et al., 2014). Incomparison

the

arable

sites

revealed

relatively

poor

density,

with

adominance

of

a

few

taxonomic

groups,

suggesting

a

partial

foodweb

driven

by

AMF

and

plant

feeding

nematodes.

These

two

TUsare

well

known

to

co-exist

on

plant

roots

(Hol

and

Cook,

2005),

competing for root space and potential feeding sites (Francl, 1993).These

ndings corresponds with the work of Tsiafouli et al.(2015)

who

found

that

increasing

land-use

intensity

resulted

in

adecrease

in

soil

faunal

taxonomic

groups,

diversity

amongfunctional

groups

and

a

reduction

in

the

average

trophic

level

inthe

soil

food

web.

De

Vries

et

al.

(2013)

also

found

that

land

useintensication reduced the abundance of most functional groups ofsoil

organisms

in

four

climatically

different

regions

in

Europe.The

difference

in

community

connectivity

with

land

use

typealso

reected the trend to have more connections in forest soilsthan

in

arable

with

more

groups

connected

(i.e.

co-occurred

atmore

than

1

site,

with

other

TUs).

Archaea

and

bacterial

TUs

werethe

most

connected

in

arable

sites,

while

the

interconnectiondominance

shifted

to

bacteria

and

fungi

in

grassland

sites.

In

forest

soils

archaeal,

enchytraeid

and

fungal

TUs

displayed

the

largestnumber

of

neighbours,

reecting the greatest connectivity.Changes

in

community

composition

may

reect

the

substrateavailability

and

disturbance

associated

with

the

different

land

usesystems.

For

example,

the

absence

of

enchytraeids

from

arablesystems

in

the

top

5

cm

can

be

related

to

physical

or

chemicaldisturbance

(ploughing,

soil

compaction

or

contamination),

ormoisture

conditions

near

the

surface

(Didden,

1993;

Rhrig

et

al.,1998;

Didden

and

Rmbke,

2001).

In

forest

systems

the

acidicnature

of

the

soils,

reduces

the

presence

of

competing

earthworms,resulting

in

the

occurrence

of

certain

enchytraeid

species,

notablyCognettia sphagnetorum (CON) (Huhta et al., 1986; Graefe andBeylich,

2003;

Rty,

2004). Archaeal,

bacterial

and

fungal

TUs

werefound

across

all

land

uses

and

are

considered

the

rst

order

primary

consumers

of

all

food

web

systems

(Powell,

2007).

Due

to

the

large

number

of

TUs

extractable

from

these

kingdoms,

it

wouldbe

expected

to

have

a

large

number

of

connections

found

in

allsites.

4.2. Carbon cycling and storage

This paper assessed whether it was possible at a large scale torelate the ecosystem service; carbon cycling and storage to keycomponents

of

the

soil

biota.

Sites

were

ranked

by

a

combinationof

SOC

(%)

and

disturbance

intensity

of

the

soil,

for

example

arablesites

are

ploughed

on

a

regular

basis,

creating

high

disturbanceintensity over a short time period, oxidising the more labile soilorganic

carbon

fractions

and

releasing

carbon

to

the

atmosphere

asCO2 (Chan et al., 2002). At the other end of the spectrum ofdisturbance

intensity,

forest

sites

are

left

relatively

undisturbed

forthe growth period of the trees, this can range from 15 years to100

years.In

this

study

basal

respiration,

molecular

microbial

biomass

andfungal

richness

were

strong

indicators

associated

with

thefunctional

capacity

of

a

system

to

cycle

and

store

soil

organic

carbon over time. Unsurprisingly, the rst two indicators describethe

capacity

of

the

system

to

turnover

carbon

(Vance

et

al.,

1987;Meidute

et

al.,

2008).

Microbes

are

the

primary

decomposers

ofplant

material

due

to

the

diversity

of

the

enzymes

produced

andtheir

unique

ability

to

produce

enzymes

to

break

down

both

simplemolecules

such

as

cellulose

and

more

complex

plant

derivedcompounds

such

as

lignin

(Romani

et

al.,

2006). While

themicrobial

(bacterial

and

fungal)

community

are

commonlyassociated

with

transformations

of

SOC

in

soils

(Tardy

et

al.,2015) it

has

also

been

shown

that

the

interaction

betweenmicrobes

and

soil

fauna

(including

mites,

earthworms,

collembo-lans,

enchytraeids

and

nematodes)

aid

this

process

and

typicallysimulate

decomposition

thus

affecting

carbon

cycling

(Nielsenet

al.,

2011).

In

this

study

the

highest

network

density

was

found

in

sites

with

SOC

between

2

and

15%,

showing

a

hump-back

model,describing

the

response

of

the

biotic

community

to

extremeconditions

of

SOC.

The

connectivity

of

the

fungal

TUs

was

greatestin

the

forest

sites,

suggesting

their

importance

in

the

food

webs

offorest

systems

and

in

terms

of

cycling

SOC.

Francisco

et

al.

(2015,this

issue)

supported

this

nding

showing

that

fungal

abundancewas

greatest

in

the

forest

sites

and

lowest

in

arable

sites.

Howeverthe

fungal

richness

was

low

in

forest

sites

compared

to

the

arableand

grassland

sites

(Stone

et

al.,

2014), this

could

be

due

to

apredominance

of

key

soil

fungi

in

forest

systems

(Bouffaud,

inpress).

Authors

have

dened

soil

microbial

biomass

and

fungi

asprincipally

responsible

for

carbon

sequestration

in

soil

(Giller

et

al.,1997;

Clemmensen

et

al.,

2013). These

last

authors

showed

that5070%

of

stored

carbon

in

a

chronosequence

of

boreal

forested

islands

derives

from

roots

and

root-associated

microorganisms.

Fig. 6. Network of biotic interaction based on signicant positive Spearman correlations in each pH category. The nodes are sized to the number of species included in theanalyses and their darkness is relational to the connectedness of the node to other nodes. The size and darkness of the connecting edges is sized to the proportion of signicantpositive correlations from all possible correlations between taxonomic units in the land-use type. For pH0.57, for pH 57 correlation >0.31 and for pH >7 correlation >0.38, respectively. pH categories were classied on the basis of Stone et al., 2015 (this issue).

10 R.E. Creamer et al./Applied Soil Ecology xxx (2015) xxxxxx

G Model

APSOIL 2259 No. of Pages 13

Please

cite

this

article

in

press

as:

R.E.

Creamer,

et

al.,

Ecological

network

analysis

reveals

the

inter-connection

between

soil

biodiversity

andecosystem function as affected by land use across Europe, Appl. Soil Ecol. (2015), http://dx.doi.org/10.1016/j.apsoil.2015.08.006

7/25/2019 Creamer 2015

11/13

Therefore

the

inclusion

of

soil

microbial

biomass,

respiration

andfungal

richness

have

been

found

to

be

key

indicators

of

carboncycling

and

potential

storage

and

should

be

considered

in

furthersoil

monitoring

frameworks

assessing

this

ecosystem

service.

4.3. Nutrient cycling of nitrogen and phosphorus

The

assessment

of

nutrient

cycling

at

76

sites

across

Europe

wasachieved

by

comparing

N

mineralisation

and

P

uptake

by

plants

atthese

sites.

Three

initial

models

were

derived,

the

models

werestatistical,

and

often

do

not

necessarily

reect

true

causalrelationships

(Mac

Nally,

2002).

However,

they

are

very

usefulto

quantitatively

describe

complex

soil

systems.

A

model

using

Ncycling

only,

accounted

for

73%

of

the

variation

in

N

mineralisationbetween

sites

and

was

described

by

the

measures

of

molecularbiomass

and L-threonine substrate utilisation in the Biolog assay. L-

threonine

contains

nitrogen

(C:N

ratio

of

4:1),

suggesting

itsrelevance

as

an

indicator

for

N-cycling.

This

suggests

a

reliance

onthe

SOC

availability

to

support

N

mineralisation.

Fierer

et

al.

(2012)suggest

that

phylogenetic

or

physiological

reponses

of

themicrobial

community

to

N

concentrations

may

be

the

result

ofthe

amount

or

type

of

organic

carbon

substrate

present

in

soils.However,

in

both

the

European

and

Dutch

datasets

(Rutgers

et

al.,

2015 this issue), L-threonine demonstrates higher variation thanaverage

within

samples.While

the

N

model

only

adequately

described

the

biologicalcontribution

to

N

cycling

across

these

sites

it

did

not

account

forthe

contribution

of

soil

biota

to

P

cycling

in

soils.

The

second

modeladdressed

P

cycling

only,

but

this

resulted

in

a

very

poor

estimateof

variance.

Soil

enzymes

were

most

signicant

in

describing

thevariation

of

P

availability

across

the

sites,

specically

phospho-monoesterase

and

Leucin

aminopeptidase.

Phosphomonoesteraseis

responsible

for

the

mineralisation

of

organic

P

to

the

inorganic

Pform,

which

is

utilised

by

plants

and

microbes

(Nannipieri

et

al.,2011).

Hendriksen

et

al.

(2015,

this

issue)

found

that

climatic

andland-use

had

no

signicant impact on the behaviour of soil enzymeactivity

but

that

soil

organic

carbon

content

was

a

strong

regulator

of enzyme activity and pH had a signicant effect on specicenzymes

such

as

phosphomonoesterase.Enchytraeid

species

richness

also

contributed

signicantly tothe

model;

however,

this

may

be

a

statistical

rather

than

anecological

phenomenon

(Mac

Nally,

2002) as

no

previous

researchhas

identied the role of enchytraeid species richness incontributing

to

P

availability

and

variation

between

sites

was

low.Finally,

a

combined

model

including

molecular

microbialbiomass,

potential

nitrication,

abundance

of

archaeal

ammo-nia-oxidizers

and

the

structure

of

the

fungal

community

provideda

better

description

of

variance

for

P

availability,

but

less

robust

forN

mineralisation

potential

across

sites.

This

model,

while

itaccounts

for

less

variation

across

sites,

compared

to

the

modelusing

N

only,

provides

a

reasonable

assessment

of

biological

indicators

for

both

N

and

P

cycling

in

soils

across

Europe.

Thepositive

relationship

between

nutrient

cycling

(N

P)

andmicrobial

biomass

can

be

attributed

to

the

important

role

thatmicro-organisms

play

in

nutrient

mobilisation.

The

microbialbiomass

is

essentially

a

labile

pool

of

P, which

is

resistant

toxation

by

abiotic

conditions

(clay

content,

Fe,

Al,

Ca)

and

loss

byleaching

(Brookes,

2001).

Enhanced

P

availability

for

plants

is

oftenattributed

to

arbuscular

mycorrhizae,

for

example

Van

der

Heijdenet

al.

(1998)

showed

that

increasing

arbuscular

mycorrhizaldiversity

and

hyphal

length

were

signicant

for

increasing

plantP

concentrations.

Interestingly

N

nitrication

was

identied

in

thecombined

model,

but

not

considered

signicant in the modelconsidering

N

cycling

only.

Van

der

Heijden

et

al.

(2008)

reportedthat

mycorrhizal

fungi

and

nitrogen-xing

bacteria

were

responsible

for

up

to

80%

of

all

nitrogen,

and

up

to

75%

ofphosphorus,

that

is

acquired

by

plants

annually.

5. Conclusion

This

is

the

rst

pan-European

study

to

measure

such

a

range

ofsoil

biological

parameters

across

Europe.

The

network

analysis

hasdemonstrated

the

variation

in

co-occurrence

of

TUs

across

thethree

different

land

use

classes.

It

has

shown

the

impact

of

land

useintensity

on

the

density

of

network

connections,

highlighting

thatarable

systems

display

much

lower

network

density

compared

tograss

and

forest

systems.

There

were

also

changes

associated

withpH

and

SOC.

Key

biological

indicators

were

identied in relation tothe

cycling

of

carbon

and

nutrients

(N

and

P).

Most

of

the

indicatorsidentied

were

comparable

to

those

identied

in

more

mechanis-tic

studies,

showing

the

applicability

of

these

indicators

for

largerscale

studies

or

monitoring

networks.

In

some

cases,

statisticalrelationships

were

acknowledged

where

no

prior

research

wasavailable

explain

the

underlying

mechanisms

and

therefore

usingsuch

large

scale

sampling

campaigns

must

be

interpreted

withsome

caution.

This

collation

and

analysis

of

soil

biodiversity

datashows

the

importance

of

large

datasets

to

understand

thecommunity

dynamics

and

the

quantitative

patterns

for

prediction

of soil ecosystem services at a continental scale.

Acknowledgements

This

work

was

supported

by

the

European

Union

within

theprojects

EcoFINDERS

(FP7-264465)

and

by

the

Rural

and

Environ-ment

Science

and

Analytical

Services

Division

of

the

ScottishGovernment.

Many

thanks

to

the

laboratory

staff

at

Teagasc,Johnstown

Castle

Research

Centre

for

their

support

in

abioticlaboratory

analyses,

in

particular

Pat

Sills,

Paul

Massey,

CarmelOConnor

and

Olivia

Fagan.

Thanks

to

Rogier

Schulte,

Teagasc,Johnstown

Castle

Research

Centre

for

advice

in

modelling

designfor

nutrient

cycling

and

carbon

cycling

statistical

models.

Appendix A. Supplementary data

Supplementary

data

associated

with

this

article

can

befound,

in

the

online

version,

at

http://dx.doi.org/10.1016/j.apsoil.2015.08.006.

References

Barbern, A., Bates, S.T., Casamayor, E.O., Fierer, N., 2012. Using network analysis toexplore co-occurrence patterns in soil microbial communities. ISME J. 6,343351.

Bartz, M.L.C., Brown, G.G., da Rosa, M.G., Klauberg Filho, O.,James, S.W., Decans, T.,Baretta, D., 2014. Earthworm richness in land-use systems in Santa Catarina,Brazil. Appl. Soil Ecol. 83, 5970.

Bascompte, J., Jordano, P., Melian, C.J., Olesen, J.M., 2003. The nested assembly ofplantanimal mutualistic networks. Proc. Natl. Acad. Sci. U. S. A. 100,

9383

9387.Bispo, A., Cluzeau, D., Creamer, R., Dombos, M., Graefe, U., Krogh, P.H., Sousa, J.P.,Peres, G., Rutgers, M.,Winding, A., Rmbke,J., 2009. Indicators for monitoringsoil biodiversity. Integr. Environ. Assess. Manag. 5, 717719.

Bligh, E.G., Dyer, W.J., 1959. A rapid method for total lipid extraction andpurication. Can. J. Biochem. Physiol. 37, 911917.

Bohan, D.A., Raybould, A., Mulder, C., Woodward, G., Tamaddoni-Nezhad, A.,Bluthgen, N., Pocock, M.J.O., Muggleton, S., Evans, D.M., Astegiano,J., Massol, F.,Loeu