-
Notifications
You must be signed in to change notification settings - Fork 0
/
main.tex
1105 lines (923 loc) · 94.5 KB
/
main.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% This is just an example/guide for you to refer to when submitting manuscripts to Frontiers, it is not mandatory to use Frontiers .cls files nor frontiers.tex %
% This will only generate the Manuscript, the final article will be typeset by Frontiers after acceptance.
% %
% %
% When submitting your files, remember to upload this *tex file, the pdf generated with it, the *bib file (if bibliography is not within the *tex) and all the figures.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%% Version 3.4 Generated 2018/06/15 %%%
%%% You will need to have the following packages installed: datetime, fmtcount, etoolbox, fcprefix, which are normally inlcuded in WinEdt. %%%
%%% In http://www.ctan.org/ you can find the packages and how to install them, if necessary. %%%
%%% NB logo1.jpg is required in the path in order to correctly compile front page header %%%
\documentclass[utf8]{frontiersSCNS} % for Science, Engineering and Humanities and Social Sciences articles
\usepackage{url,hyperref,lineno,microtype}
\usepackage{graphicx}
\usepackage{subcaption}
\usepackage[onehalfspacing]{setspace}
\usepackage[english]{babel}
\usepackage[most]{tcolorbox}
\usepackage{tabularx}
% \usepackage[table]{xcolor} % loads also »colortbl«
\linenumbers
% Leave a blank line between paragraphs instead of using \\
\def\keyFont{\fontsize{8}{11}\helveticabold }
\def\firstAuthorLast{Lalejini {et~al.}} %use et al only if is more than 1 author
\def\Authors{Alexander Lalejini\,$^{1,2,3,*}$, Austin J. Ferguson\,$^{1,2,3}$, Nkrumah A. Grant\,$^{1,4}$, and Charles Ofria\,$^{1,2,3}$}
\def\Address{
$^{1}$BEACON Center for the Study of Evolution in Action, Michigan State University, East Lansing, MI, USA \\
$^{2}$Ecology, Evolution, and Behavior Program, Michigan State University, East Lansing, MI, USA \\
$^{3}$Department of Computer Science and Engineering, Michigan State University, East Lansing, MI, USA \\
$^{4}$Department of Biological Sciences, University of Idaho, Moscow, ID, USA
}
\def\corrAuthor{Alexander Lalejini}
\def\corrEmail{amlalejini@gmail.com}
\begin{document}
\onecolumn
\firstpage{1}
\title[Adaptive plasticity stabilizes evolution]{Adaptive phenotypic plasticity stabilizes evolution in fluctuating environments}
\author[\firstAuthorLast ]{\Authors} %This field will be automatically populated
\address{} %This field will be automatically populated
\correspondance{} %This field will be automatically populated
\extraAuth{}% If there are more than 1 corresponding author, comment this line and uncomment the next one.
%\extraAuth{corresponding Author2 \\ Laboratory X2, Institute X2, Department X2, Organization X2, Street X2, City X2 , State XX2 (only USA, Canada and Australia), Zip Code2, X2 Country X2, email2@uni2.edu}
\maketitle
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Utility commands
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newcommand{\code}{\texttt}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Evolutionary change rate experiment
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% data from experiment - 2021-02-08
\newcommand{\evolutionaryChangeRateReplicates}{100}
\newcommand{\evolutionaryChangeRatePlasticReps}{42}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Novel traits experiment
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newcommand{\novelTraitsReplicates}{100}
\newcommand{\novelTraitsReward}{10\%}
\newcommand{\novelTraitsPlasticReps}{42}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Deleterious hitchhiking experiment
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% - 2021-02-05 -
\newcommand{\deleteriousHitchhikingReplicates}{100}
\newcommand{\deleteriousHitchhikingPlasticReps}{43}
\newcommand{\instPoisonMagnitude}{10\%}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Metric definitions
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newcommand{\SweepsMetricName}{
Coalescence event count
}
\newcommand{\SweepsMetricDesc}{
Number of coalescence events that have occurred, which indicates the frequency of selective sweeps in the population.
}
\newcommand{\MutationCountMetricName}{
Mutation count
}
\newcommand{\MutationCountMetricDesc}{
Sum of all mutations that have occurred along a lineage.
}
\newcommand{\PhenotypicVolatilityMetricName}{
Phenotypic volatility
}
\newcommand{\PhenotypicVolatilityMetricDesc}{
Number of instances where parent and offspring phenotypic profiles do not match along a lineage.
}
\newcommand{\MutationalStabilityMetricName}{
Mutational robustness
}
\newcommand{\MutationalStabilityMetricDesc}{
Proportion of mutations (from the set of all possible one-step mutations) that do not change the phenotypic profile of a focal genotype. We also measured \textit{realized mutational robustness}, which is the proportion of mutated offspring along a lineage whose phenotypic profile matches that of their parent.
}
\newcommand{\TaskPerformanceMetricName}{
Final novel function count
}
\newcommand{\TaskPerformanceMetricDesc}{
Count of unique novel functions performed by the representative organism in a final population from experiment \hyperref[sec:methods:exp:novel-task-evolution]{phase 2B}.
This metric can range from 0 to 71 and measures how well the fitness landscape was exploited at a given point in time.
}
\newcommand{\TaskDiscoveryMetricName}{
Novel function discovery
}
\newcommand{\TaskDiscoveryMetricDesc}{
Number of unique novel functions ever performed along a given lineage in experimental \hyperref[sec:methods:exp:novel-task-evolution]{phase 2B}, even if a function is later lost.
This metric can range from 0 to 71 and measures a given lineage's level of exploration of the fitness landscape.
}
\newcommand{\TaskLossMetricName}{
Novel function loss
}
\newcommand{\TaskLossMetricDesc}{
Number of instances along a given lineage from experimental \hyperref[sec:methods:exp:novel-task-evolution]{phase 2B} where a novel function is performed by a parent but not its offspring.
This metric measures how often a given lineage fails to retain evolved traits over time.
}
\newcommand{\FinalPoisonMetricName}{
Final deleterious function count
}
\newcommand{\FinalPoisonMetricDesc}{
Number of times the deleterious function is performed by the representative organism from a final population from experiment \hyperref[sec:methods:exp:deleterious-instruction-accumulation]{phase 2C}.
}
\newcommand{\LineagePoisonMetricName}{
Deleterious function acquisition count
}
\newcommand{\LineagePoisonMetricDesc}{
Number of instances along a given lineage where a mutation causes an offspring to perform the deleterious function more times than its parent.
}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% ABSTRACT
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{abstract}
\section{}
Fluctuating environmental conditions are ubiquitous in natural systems, and populations have evolved various strategies to cope with such fluctuations.
The particular mechanisms that evolve profoundly influence subsequent evolutionary dynamics.
One such mechanism is phenotypic plasticity, which is the ability of a single genotype to produce alternate phenotypes in an environmentally dependent context.
Here, we use digital organisms (self-replicating computer programs) to investigate how adaptive phenotypic plasticity alters evolutionary dynamics and influences evolutionary outcomes in cyclically changing environments.
Specifically, we examined the evolutionary histories of both plastic populations and non-plastic populations to ask:
(1) Does adaptive plasticity promote or constrain evolutionary change?
(2) Are plastic populations better able to evolve and then maintain novel traits?
And (3), how does adaptive plasticity affect the potential for maladaptive alleles to accumulate in evolving genomes?
We find that populations with adaptive phenotypic plasticity undergo less evolutionary change than non-plastic populations, which must rely on genetic variation from \textit{de novo} mutations to continuously readapt to environmental fluctuations.
Indeed, the non-plastic populations undergo more frequent selective sweeps and accumulate many more genetic changes.
We find that the repeated selective sweeps in non-plastic populations drive the loss of beneficial traits and accumulation of maladaptive alleles, whereas phenotypic plasticity can stabilize populations against environmental fluctuations.
% via deleterious hitchhiking, whereas phenotypic plasticity can stabilize populations against environmental fluctuations.
This stabilization allows plastic populations to more easily retain novel adaptive traits than their non-plastic counterparts.
In general, the evolution of adaptive phenotypic plasticity shifted evolutionary dynamics to be more similar to that of populations evolving in a static environment than to non-plastic populations evolving in an identical fluctuating environment.
All natural environments subject populations to some form of change; our findings suggest that the stabilizing effect of phenotypic plasticity plays an important role in subsequent adaptive evolution.
\tiny
\keyFont{ \section{Keywords:} phenotypic plasticity, experimental evolution, digital evolution, changing environments, Avida}
\end{abstract}
\begin{raggedbottom}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% The introduction should be succinct, with no subheadings.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Introduction}
%%%%%%%%%%%%%%%%%%%%%%
% -- Different mechanisms evolve to cope with fluctuating environments, and those mechanisms
% affect evolution. --
%%%%%%%%%%%%%%%%%%%%%%
Natural organisms employ a wide range of evolved strategies for coping with environmental change, such as
periodic migration \citep{winger_long_2019},
bet-hedging \citep{beaumont_experimental_2009},
adaptive tracking \citep{barrett_adaptation_2008},
and phenotypic plasticity \citep{ghalambor_adaptive_2007}.
The particular mechanisms that evolve in response to fluctuating environments will also shift the course of subsequent evolution \citep{wennersten_population-level_2012,schaum_plasticity_2014}.
As such, if we are to understand or predict evolutionary outcomes, we must be able to identify which mechanisms are most likely to evolve and what constraints and opportunities they impart on subsequent evolution.
%%%%%%%%%%%%%%%%%%%%%%
% -- We focus on how plasticity affects subsequent evolution. --
%%%%%%%%%%%%%%%%%%%%%%
In this work, we focus on phenotypic plasticity, which can be defined as the capacity for a single genotype to alter phenotypic expression in response to a change in its environment \citep{west-eberhard_developmental_2003}.
Phenotypic plasticity is controlled by genes whose expression is coupled to one or more environmental signals, which may be either biotic or abiotic.
For example, the sex ratio of the crustacean \textit{Gammarus duebeni} is modulated by changes in photoperiod and temperature \citep{dunn_two_2005}, and the reproductive output of some invertebrate species is heightened when infected with parasites to compensate for offspring loss \citep{chadwick_parasite-mediated_2005}.
% -- @AML: move next two lines to later as suggested by editor --
% In this study, we conducted digital evolution experiments to investigate how the evolution of adaptive phenotypic plasticity shifts the course of evolution in a cyclically changing environment.
% Specifically, we examined the effects of adaptive plasticity on subsequent genomic and phenotypic change, the capacity to evolve and then maintain novel traits, and the accumulation of deleterious alleles.
%%%%%%%%%%%%%%
% Effects of phenotypic plasticity on subsequent evolution disputed.
% - baseline expections on rate of evolutionary response given adaptive vs. non-adaptive plasticity
%%%%%%%%%%%%%%
Evolutionary biologists have long been interested in how evolutionary change is influenced by phenotypic plasticity because of its role in generating phenotypic variance \citep{gibert_phenotypic_2019}.
The effects of phenotypic plasticity on adaptive evolution have been disputed, as few studies have been able to observe both the initial patterns of plasticity and the subsequent divergence of traits in natural populations \citep{ghalambor_adaptive_2007,wund_assessing_2012,forsman_rethinking_2015,ghalambor_non-adaptive_2015,hendry_key_2016}.
In changing environments, adaptive phenotypic plasticity provides a mechanism for organisms to regulate trait expression within their lifetime, which can stabilize populations through those changes \citep{gibert_phenotypic_2019}.
In this context, the stabilizing effect of adaptive plasticity has been hypothesized to constrain the rate of adaptive evolution \citep{gupta_study_1982,ancel_undermining_2000,huey_behavioral_2003,price_role_2003,paenke_influence_2007}.
That is, directional selection may be weak if environmentally-induced phenotypes are close to the optimum; as such, adaptively plastic populations may evolve slowly (relative to non-plastic populations) unless there is a substantial fitness cost to plasticity.
\begin{figure}[ht!]
\centering
\includegraphics[width=0.85\textwidth]{media-reaction-norms.pdf}
\caption{\small
\textbf{Hypothetical reaction norms for populations comprising genotypes placed in different environments.}
A reaction norm describes phenotypic change (or lack thereof) induced by environmental variation~\citep{west-eberhard_phenotypic_2008}.
In all panels, two environmental conditions (denoted E$_1$ and E$_2$) are shown on the x-axis.
The y-axis indicates the phenotype expressed in each environment with O$_{\text{E}1}$ and O$_{\text{E}2}$ designating the optimal phenotype for E$_1$ and E$_2$, respectively.
Each pair of points connected by a solid black line denotes a genotype, with the points themselves representing its hypothetical phenotypes in each environment.
We present three scenarios for how populations could respond to a change from E$_1$ to E$_2$.
(a) A non-plastic population where phenotypes do not change with environmental shifts.
% In such cases, we would expect strong directional selection toward O$_{\text{E}2}$ after the environment changes.
(b) An adaptively plastic population where phenotypes dynamically adjust to the new optimum.
% As such, we would expect this population to remain relatively stable after the environment changes.
% (c) A population exhibiting non-adaptive plasticity with substantial variation in how individuals respond to the environmental change. In this case, we expect the change in environment to result in a rapid evolutionary sweep by genotypes closest to the new optimal phenotype.
% (d) A population exhibiting maladaptive plasticity.
(c) A population exhibiting non-adaptive plasticity where environmental change induces phenotypes further away from the optimum.
% relative to the given environmental change.
% When the environment changes, there is little variation for selection to act on, and without beneficial mutations, this population could be at risk of extinction.
}
\label{fig:reaction-norms}
\end{figure}
% -- Plasticity as a source of cryptic variation --
Phenotypic plasticity allows for the accumulation of genetic variation in genomic regions that are unexpressed under current environmental conditions.
Such cryptic (``hidden'') genetic variation can serve as a source of diversity in the population, upon which selection can act when the environment changes \citep{schlichting_hidden_2008,levis_evaluating_2016}.
It remains unclear to what extent and under what circumstances this cryptic variation caches adaptive potential or merely accumulates deleterious alleles \citep{gibson_uncovering_2004,paaby_cryptic_2014,zheng_cryptic_2019}.
% -- Genes as followers --
The ``genes as followers'' hypothesis (also known as the ``plasticity first'' hypothesis) predicts that phenotypic plasticity may facilitate adaptive evolutionary change by producing variants with enhanced fitness under stressful or novel conditions \citep{west-eberhard_developmental_2003,schwander_genes_2011,levis_evaluating_2016}.
Environmentally-induced trait changes can be refined through selection over time (\textit{i.e.}, genetic accommodation).
Further, selection may drive plastic phenotypes to lose their environmental dependence over time in a process known as genetic assimilation \citep{west-eberhard_developmental_2005,pigliucci_phenotypic_2006,crispo_baldwin_2007,schlichting_phenotypic_2014,levis_evaluating_2016}.
In this way, environmentally-induced phenotypic changes can precede an evolutionary response.
% -- Plastic rescue + wrap up on plasticity background? --
Phenotypic plasticity may also ``rescue'' populations from extinction under changing environmental conditions by buffering populations against novel stressors.
This buffer promotes stability and persistence and grants populations time to further adapt to rapidly changing environmental conditions \citep{west-eberhard_developmental_2003,chevin_when_2010}.
Disparate predictions about how phenotypic plasticity may shift the course of subsequent evolution are not necessarily mutually exclusive.
Genetic and environmental contexts determine if, and to what extent, phenotypic plasticity promotes or constrains subsequent evolution.
Figure~\ref{fig:reaction-norms} overviews how we might expect different forms of phenotypic plasticity to result in different evolutionary responses after an environmental change.
In Figure~\ref{fig:reaction-norms}a, we would expect non-plastic populations to experience strong directional selection toward the new optimum (O$_{\text{E}2}$) after the environment changes.
We would expect an adaptively plastic population (Figure~\ref{fig:reaction-norms}b) to remain relatively stable after the environment changes, as plasticity shifts organisms' phenotypes to the new optimum.
In Figure~\ref{fig:reaction-norms}c, we would expect the non-adaptively plastic population to experience strong directional selection on their response to the new environmental conditions; indeed, such maladaptive plasticity may even put the population at risk of extinction in the absence of beneficial mutations.
%%%%%%%%%%%%%%%%%%%%%%
% -- Digital evolution --
%%%%%%%%%%%%%%%%%%%%%%
Experimental studies investigating the relationship between phenotypic plasticity and evolutionary outcomes can be challenging to conduct in natural systems.
Such experiments would require the ability to irreversibly toggle plasticity followed by long periods of evolution during which detailed phenotypic data would need to be collected.
Digital evolution experiments have emerged as a powerful research framework from which evolution can be studied.
In digital evolution, self-replicating computer programs (digital organisms) compete for resources, mutate, and evolve following Darwinian dynamics \citep{wilke_biology_2002}.
Digital evolution studies balance the speed and transparency of mathematical and computational simulations with the open-ended realism of laboratory experiments.
Modern computers allow us to observe many generations of digital evolution at tractable time scales; thousands of generations can take mere minutes as opposed to months, years, or millennia.
Digital evolution systems also allow for perfect, non-invasive data tracking.
Such transparency permits the tracking of complete evolutionary histories within an experiment, which circumvents the historical problem of drawing evolutionary inferences using incomplete records (from frozen samples or fossils) and extant genetic sequences.
Additionally, digital evolution systems allow for experimental manipulations and analyses that go beyond what is possible in wet-lab experiments.
Such analyses have included exhaustive knockouts of every site in a genome to identify the functionality of each \citep{lenski_evolutionary_2003},
comprehensive characterization of local mutational landscapes \citep{lenski_genome_1999,canino-koning_fluctuating_2019},
and the real-time reversion of all deleterious mutations as they occur to isolate their long-term effects on evolutionary outcomes \citep{covert_experiments_2013}.
Furthermore, digital evolution studies allow us to directly toggle the possibility for adaptive plastic responses to evolve, which enables us to empirically test hypotheses that were previously relegated to theoretical analyses.
%%%%%%%%%%%%%%%%%%%%%%
% -- Overview of what we're testing / doing --
%%%%%%%%%%%%%%%%%%%%%%
In this study, we conduct digital evolution experiments to investigate how the evolution of adaptive phenotypic plasticity shifts the course of evolution in a cyclically changing environment.
We use the Avida Digital Evolution Platform \citep{ofria_avida:_2009}.
Avida is an open-source system that has been used to conduct a wide range of well-regarded studies on evolutionary dynamics, including
the origins of complex features \citep{lenski_evolutionary_2003},
the survival of the flattest effect \citep{wilke_evolution_2001},
and the origins of reproductive division of labor \citep{goldsby_evolutionary_2014}.
Our experiments build directly on previous studies in Avida that characterized the \textit{de novo} evolution of adaptive phenotypic plasticity \citep{clune_investigating_2007,lalejini_evolutionary_2016} as well as previous work investigating the evolutionary consequences of fluctuating environments for populations of non-plastic digital organisms \citep{li_digital_2004,canino-koning_fluctuating_2019}.
Of particular relevance, \cite{clune_investigating_2007} and \cite{lalejini_evolutionary_2016} experimentally demonstrated that adaptive phenotypic plasticity can evolve given the following four conditions (as described by \citealt{ghalambor_behavior_2010}):
(1) populations experience temporal environmental variation,
(2) these environments are differentiable by reliable cues,
(3) each environment favors different phenotypic traits,
and (4) no single phenotype exhibits high fitness across all environments.
We build on this previous work, but we shift our focus from the evolutionary causes of adaptive phenotypic plasticity to investigate its evolutionary consequences in a fluctuating environment.
Specifically, we examine the effects of adaptive plasticity on subsequent genomic and phenotypic change, the capacity to evolve and then maintain novel traits, and the accumulation of deleterious alleles.
% - what we did -
Each of our experiments are divided into two phases: in phase one, we precondition sets of founder organisms with differing plastic or non-plastic adaptations;
in phase two, we examine the subsequent evolution of populations founded with organisms from phase one under specific environmental conditions (Figure \ref{fig:experimental-design}).
First, we examine the evolutionary histories of phase two populations to test whether adaptive plasticity constrained subsequent genomic and phenotypic changes.
Next, we evaluate how adaptive plasticity influences how well populations produced by each type of founder can evolve and retain novel adaptive traits.
Finally, we examine lineages to determine whether adaptive plasticity facilitated the accumulation of cryptic genetic variation that would prove deleterious when the environment changed.
%%%%%%%%%%%%%%%%%%%%%%
% - Overview of findings -
%%%%%%%%%%%%%%%%%%%%%%
We found that the evolution of adaptive plasticity reduced subsequent rates of evolutionary change in a cyclic environment.
The non-plastic populations underwent more frequent selective sweeps and accumulated many more genetic changes over time, as non-plastic populations relied on genetic variation from \textit{de novo} mutations to continuously readapt to environmental changes.
The evolution of adaptive phenotypic plasticity buffered populations against environmental fluctuations, whereas repeated selective sweeps in non-plastic populations drove the accumulation of deleterious mutations and the loss of secondary beneficial traits. % via deleterious hitchhiking.
As such, adaptively plastic populations were better able to retain novel traits than their non-plastic counterparts.
In general, the evolution of adaptive phenotypic plasticity shifted evolutionary dynamics to be more similar to that of populations evolving in a static environment than to non-plastic populations evolving in an identical fluctuating environment.
\section{Materials and Methods}
\label{sec:methods}
\subsection{The Avida Digital Evolution Platform}
\label{sec:methods:avida}
% -- Define digital organisms --
Avida is a study system wherein self-replicating computer programs (digital organisms) compete for space on a finite toroidal grid \citep{ofria_avida:_2009}.
Each digital organism is defined by a linear sequence of program instructions (its genome) and a set of virtual hardware components used to interpret and express those instructions.
Genomes are expressed sequentially except when the execution of one instruction (\textit{e.g.}, a ``jump'' instruction) deterministically changes which instruction should be executed next.
Genomes are built using an instruction set that is both robust (\textit{i.e.}, any ordering of instructions is syntactically valid, though not necessarily meaningful) and Turing Complete (\textit{i.e.}, able to represent any computable function, though not necessarily in an efficient manner).
The instruction set includes operations for basic computations, flow control (\textit{e.g.}, conditional logic and looping), input, output, and self-replication.
% The virtual hardware set includes components such as a central processing unit (CPU) for executing instructions, registers to store values, buffers for inputs and outputs, and memory stacks \citep{ofria_avida:_2009}.
% -- Define reproduction & mutation --
Organisms in Avida reproduce asexually by copying their genome instruction-by-instruction and then dividing.
However, copy operations are imperfect and can result in single-instruction substitution mutations in an offspring's genome.
For this work, we configured copy operations to err at a rate of one expected mutation for every 400 instructions copied (\textit{i.e.}, a per-instruction error rate of 0.0025).
We held individual genomes at a fixed length of 100 instructions (the default genome size in Avida); that is, we did not include insertion and deletion mutations.
We used fixed-length genomes to control for treatment-specific conditions resulting in the evolution of substantially different genome sizes \citep{supplemental_material}\footnote{
We repeated our experiments without genome size restrictions and observed qualitatively similar results (see supplemental material, \citealt{supplemental_material}).
}, which could, on its own, drive differences in evolutionary outcomes among experimental treatments.
When an organism divides in Avida, its offspring is placed in a random location on the toroidal grid, replacing any previous occupant.
For this work, we used the default 60 by 60 grid size, which limits the maximum population size to 3600 organisms.
As such, improvements to the speed of self-replication are advantageous in the competition for space.
% -- Define fitness/metabolic rate/improving replication speed --
During evolution, organism replication rates improve in two ways: by improving genome efficiency (\textit{e.g.}, using a more compact encoding) or by accelerating the rate at which the genome is expressed (their ``metabolic rate'').
An organism's metabolic rate determines the speed at which it executes instructions in its genome.
Initially, an organism's metabolic rate is proportional to the length of its genome, but that rate is adjusted as it completes designated functions, such as performing Boolean logic functions \citep{ofria_avida:_2009}.
In this way, we can reward or punish particular phenotypic traits (\textit{i.e.}, Boolean logic functions).
\subsubsection{Phenotypic plasticity in Avida}
\label{sec:methods:avida:plasticity}
% -- Phenotypes & Phenotypic plasticity --
% - What is a phenotype in Avida?
In this work, we measure a digital organism's phenotype as the set of Boolean logic functions that it performs (\textit{i.e.} expresses) in a given environment.
Sensory instructions in the Avida instruction set allow organisms to detect how performing a particular function would affect their metabolic rate (see supplemental material for more details, \citealt{supplemental_material}).
We define a phenotypically plastic organism as one that uses sensory information to alter which functions it performs based on the environment.
% -- Adaptive plasticity & non-adaptive plasticity --
Phenotypic plasticity in Avida can be adaptive or non-adaptive for a given set of environments.
Adaptive plasticity shifts net function expression closer to the optimum for the given environments.
Non-adaptive plasticity changes function expression in either a neutral or deleterious way.
In this work, optimal plasticity toggles functions to always perfectly match the set of rewarded functions for the given set of environments.
\subsection{Experimental design}
\label{sec:methods:experiment}
\begin{figure}[h!]
\centering
\includegraphics[width=1\textwidth]{media-experimental-design.pdf}
\caption{\small
\textbf{Overview of experimental design.}
The three plots in panel (a) show the environmental conditions used in every experiment and whether they reward or punish each base function.
The two plots in (b) show the additional functions added in phases 2B and 2C. %Optional: phase 2A is not shown as it adds no additional functions
%All novel tasks in phase 2B confer a 10\% metabolic reward, while executing the poisonous task in phase 2C causes a 10\% metabolic punishment (bars not drawn to scale).
Panel (c) shows treatment differences and experimental phases.
Treatments are listed on the left, with each treatment specifying its environmental configuration and whether sensors are functional.
We conducted three independent two-phase experiments, each described on the right.
Phases 2B and 2C are textured to match their function definitions in panel (b).
Phase one is repeated for \textit{each} experiment with 100 replicate populations per treatment per experiment.
For each replicate at the end of phase one, we used an organism of the most abundant genotype to found the second phase population.
All STATIC and NON-PLASTIC populations move on to phase two, but PLASTIC populations only continue to the second phase if their most abundant genotype exhibits optimal plasticity.
Metrics are recorded only in phase two.
}
\label{fig:experimental-design}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% OUTLINE
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Overview of experimental design - with diagram (not a subsubsection?)
% - Focused around diagram
% Environment implementations
% Specifications for Phase 1
% specifications for Phase 2A
% specifications for Phase 2B
% specifications for Phase 2C
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% -- Treatments --
We conducted three independent experiments using Avida to investigate how the evolution of adaptive plasticity influences evolutionary outcomes in fluctuating environments.
For each experiment, we compared the evolutionary outcomes of populations evolved under three treatments (Figure \ref{fig:experimental-design}):
(1) a \textbf{PLASTIC} treatment where the environment fluctuates, and digital organisms can use sensory instructions to differentiate between environmental states;
(2) a \textbf{NON-PLASTIC} treatment with identical environment fluctuations, but where sensory instructions are disabled;
and (3) a \textbf{STATIC} control where organisms evolve in a constant environment.
% -- Experiment phases --
Each experiment was divided into two phases that each lasted for 200,000 updates\footnote{
One update in Avida is the amount of time required for the average organism to execute 30 instructions.
See \citep{ofria_avida:_2009} for more details.
} of evolution (Figure~\ref{fig:experimental-design}), which is equivalent to approximately 30,000 to 40,000 generations.
% Phase one of each experiment was an acclimation phase.
In phase one of each experiment, we evolved plastic, non-plastic, and control organisms for use in phase two.
In phase two, we founded new populations with these evolved organisms and examined their subsequent evolution under given combinations of treatment and experimental conditions.
During phase two, we tracked and saved each population's evolutionary history as well as saving the full final population.
Phase one was for preconditioning only; all comparisons between treatments were performed on phase two data.
\subsubsection{Environments}
\label{sec:methods:experiment:environments}
% ----- ENVIRONMENTS -----
% traits_set_a <- c("not", "and", "or")
% traits_set_b <- c("nand", "ornot", "andnot")
% -- Environment definitions --
% In all of our experiments, populations evolve in either static or fluctuating environments.
We constructed three experimental environmental conditions, abbreviated hereafter as ``ENV-A'', ``ENV-B'', and ``ENV-ALL''.
In ENV-A, organisms are rewarded for expressing the NOT, AND, and OR Boolean logic functions, and organisms are punished for expressing the NAND, AND-NOT, and OR-NOT functions.
ENV-B is the reverse of ENV-A; that is, in ENV-B, organisms are rewarded for expressing the NAND, AND-NOT, and OR-NOT functions and are punished for expressing the NOT, AND, and OR functions.
In ENV-ALL, organisms are rewarded for expressing each of the NOT, AND, OR, NAND, AND-NOT, and OR-NOT functions.
In all environmental conditions (ENV-A, ENV-B, and ENV-ALL), a rewarded function expressed by an organism doubles their metabolic rate, allowing them to execute twice as many instructions in the same amount of time.
A punished function halves an organism's metabolic rate.
Each Boolean logic function is a non-trivial trait to evolve, as they each require the coordination of multiple genetic instructions to express~\citep{lenski_evolutionary_2003}.
% Figure \ref{fig:experimental-design} describes these environments based on whether each of six Boolean logic tasks (NOT, NAND, AND, OR-NOT, OR, and AND-NOT) is rewarded or punished.
% -- Treatment-specific environment details --
In both the PLASTIC and NON-PLASTIC treatments, the environment cycles between equal-length periods of ENV-A and ENV-B.
Each of these periods persist for 100 updates (approximately 15 to 20 generations).
Thus, populations experience a total of 1,000 full periods of ENV-A interlaced with 1,000 full periods of ENV-B during each experimental phase.
Previous work has shown that this rate of change reliably allows for the evolution of adaptive phenotypic plasticity in Avida~\citep{clune_investigating_2007,lalejini_evolutionary_2016}.
% -- Sensory instructions + control flow and controlling the capacity for plasticity --
Organisms in the PLASTIC treatments differentiate between ENV-A and ENV-B by executing one of six sensory instructions, each associated with a particular Boolean logic function; these sensory instructions detect whether their associated function is currently rewarded or punished.
By using sensory information in combination with execution flow-control instructions, organisms can conditionally perform different functions depending on the current environmental conditions.
\subsubsection{Experiment Phase 1 -- Environment preconditioning}
\label{sec:methods:experiment:phase-one}
For each treatment, we founded 100 independent populations from a common ancestral strain capable only of self-replication.
At the end of phase one, we identified the most abundant (\textit{i.e.}, dominant) genotype and sampled an organism with that genotype from each replicate population to found a new population for phase two.
For the PLASTIC treatment, we needed to ensure that our observations during the second phase of each experiment reflected the evolutionary consequences of adaptive plasticity.
To do so, measured the plasticity of each PLASTIC-treatment population's dominant genotype by independently testing that genotype in each of ENV-A and ENV-B and recording the phenotype expressed in each environment.
We discarded PLASTIC-treatment phase one populations if the dominant genotype did not exhibit optimal plasticity (as defined in Section~\ref{sec:methods:avida:plasticity}), which ensured that PLASTIC-treatment phase two populations were founded with an optimally plastic organism.
% For the PLASTIC treatment, we measure the plasticity of each population's dominant genotype by independently testing the given genotype in each of ENV-A and ENV-B and recording the phenotype expressed in each environment.
% We discard PLASTIC-treatment phase one populations if the dominant genotype does not exhibit optimal plasticity (as defined in Section \ref{sec:methods:avida:plasticity}).
% This approach ensures that measurements taken on PLASTIC-treatment populations during the second phase of each experiment reflect the evolutionary consequences of adaptive plasticity.
\subsubsection{Experiment Phase 2A -- Evolutionary change rate}
\label{sec:methods:exp:evolutionary-change-rate}
We conducted experiment phase 2A to test for differences in evolutionary change---the accumulation of genetic and phenotypic changes---among populations evolving under each of our three treatment conditions (PLASTIC, NON-PLASTIC, and STATIC).
Phase 2A continued exactly as phase one, except we tracked the rates of evolutionary change in each of the PLASTIC-, NON-PLASTIC-, and STATIC-treatment populations.
Specifically, we quantified evolutionary change using four metrics (each described in Table~\ref{tab:metrics-definitions}):
(1) coalescence event count,
(2) mutation count,
(3) phenotypic volatility,
and (4) mutational robustness.
While environmental conditions during phases one and 2A are identical, these phases are distinct in how populations are founded: each phase one population is founded with a common ancestor capable only of self-replication, whereas each phase two population is founded with an organism that was evolved in its treatment-specific conditions.
Thus, during phase two, phylogenies are rooted with an ancestor that is well-adapted to its treatment conditions, which, in turn, ensures that our observations can exclude dynamics associated with initial adaptation.
\subsubsection{Experiment Phase 2B -- Novel function evolution}
\label{sec:methods:exp:novel-task-evolution}
We conducted experiment phase 2B to quantify the extent to which organisms evolved under PLASTIC-, NON-PLASTIC-, and STATIC-treatment conditions were able to acquire and retain novel functions.
Phase 2B extended the conditions of phase one by adding 71 novel Boolean logic functions~\citep{ofria_avida:_2009}, which were always rewarded in all treatments.
The original six phase one functions (NOT, NAND, AND, OR-NOT, OR, and AND-NOT; hereafter called ``base'' functions) continued to be rewarded or punished according to the particular treatment conditions.
% @CAO: line below can be restored (and reworded) if readers need help with the intuition.
%As such, in fluctuating environments, the six base tasks continued to fluctuate, but the additional 71 tasks were always rewarded; in static environments, performing any of the 77 logic tasks was always beneficial.
An organism's metabolic rate was increased by \novelTraitsReward\ for each novel function that it expressed (limited to one reward per function).
This reward provided a selective pressure to evolve these functions, but their benefits did not overwhelm the 100\% metabolic rate increase conferred by rewarded base functions.
As such, populations in the PLASTIC and NON-PLASTIC treatments could not easily escape environmental fluctuations by abandoning the fluctuating base functions.
During this experiment, we used three metrics to quantify novel function acquisition and retention in evolving populations (each described in Table~\ref{tab:metrics-definitions}):
% During this experiment, we tracked the extent to which populations evolving under each treatment were capable of acquiring and retaining novel tasks.
% Specifically, we used three metrics (each described in Table~\ref{tab:metrics-definitions}):
(1) final novel function count,
(2) novel function discovery,
and (3) novel function loss.
\subsubsection{Experiment Phase 2C -- Deleterious instruction accumulation}
\label{sec:methods:exp:deleterious-instruction-accumulation}
We conducted experiment phase 2C to quantify the extent to which organisms evolved under PLASTIC-, NON-PLASTIC-, and STATIC-treatment conditions acquired deleterious instructions via mutation.
Phase 2C extended the instruction set of phase one with a \code{deleterious} instruction.
When an organism executes a \code{deleterious} instruction, it performs a ``deleterious'' function, which reduces the organism's metabolic rate (and thus reproductive success) but does not otherwise alter the organism's behavior.
We imposed a 10\% penalty each time an organism performed the deleterious function, making the \code{deleterious} instruction explicitly harmful to execute.
We did not limit the number of times that an organism could perform the deleterious function, and as such, organisms could perform the deleterious function as many times as they executed the \code{deleterious} instruction.
We tracked the number of times each organism along the dominant lineage performed the deleterious function.
Specifically, we used two metrics (each described in Table~\ref{tab:metrics-definitions}):
(1) final deleterious function count and (2) deleterious function acquisition count.
\subsection{Experimental analyses}
\newcommand*{\thead}[1]{\multicolumn{1}{c}{\bfseries #1}}
\setlength{\tabcolsep}{16pt}
\renewcommand{\arraystretch}{1.5}
\begin{table}[ht]
\centering
\rowcolors{2}{gray!25}{white}
\begin{tabularx}{\linewidth}{lX} % p{10cm}
\rowcolor{gray!50}
\hline
\thead{Metric} & \thead{Description} \\
\hline
\SweepsMetricName & \SweepsMetricDesc \\
\MutationCountMetricName & \MutationCountMetricDesc \\
\PhenotypicVolatilityMetricName & \PhenotypicVolatilityMetricDesc \\
\MutationalStabilityMetricName & \MutationalStabilityMetricDesc \\
\TaskPerformanceMetricName & \TaskPerformanceMetricDesc \\
\TaskDiscoveryMetricName & \TaskDiscoveryMetricDesc \\
\TaskLossMetricName & \TaskLossMetricDesc \\
\FinalPoisonMetricName & \FinalPoisonMetricDesc \\
\LineagePoisonMetricName & \LineagePoisonMetricDesc \\
\hline
\end{tabularx}
\caption{\textbf{Metric descriptions.}}
\label{tab:metrics-definitions}
\end{table}
% naive environment = what organism experiencing
% alternative environment = what organism is not currently experiencing
% - How we extracted representative lineages -
For each of our experiments, we tracked and analyzed the phylogenetic histories of evolving populations during phase two and generated a set of summary statistics (Table~\ref{tab:metrics-definitions}).
For each phylogenetic history, we counted the number of times that the most recent common ancestor for the population shifted and used this value as the number of coalescence events.
Next, at the final time point, we identified the most abundant genotype in the evolved population, and chose a \textit{representative organism} from that genotype for further analysis.
% We then isolated the lineage from the founding organism to the representative organism, which we used as the representative lineage for further analysis.
% We used the lineage from the founding organism to the representative organism as the \textit{representative lineage} for further analysis.
We used the lineage from the founding organism to the representative organism to summarize the evolutionary pathway of a population.
This lineage represents the majority of the evolutionary history from the given population as long as the entire population traces back to the lineage in recent history.
We manually inspected evolved phylogenies and found no evidence that any of our experimental treatments supported long-term coexistence.
As such, analyses of the representative lineages reflect the majority of evolutionary history for a given population.
% To be confident that each representative lineage reflected the majority of the evolutionary history from a given populuation at the end of the ex
% [These lineages represent the majority of the evolutionary history from a given population only if there aren't members of the population that haven't diverged too far in the past / only if single the entire population traces back to that lineage in recent history (i.e., no long-term coexistence)].
% We manually inspected evolved phylogenies and found no evidence that any of our experimental treatments supported long-term coexistence.
% As such, each of the representative lineages reflect the majority of evolutionary history from a given population at the end of our experiment.
% -- PHENOTYPIC PROFILE --
% DEFINE PHENOTYPIC PROFILE - metric definitions will use this.
Some of our metrics (Table \ref{tab:metrics-definitions}) required us to measure genotype-by-environment interactions.
Importantly, in the fluctuating environments, we needed to differentiate phenotypic changes that were caused by mutations from those that were caused by environmental changes.
To accomplish this, we produced organisms with the given focal genotype, measured their phenotype in each environmental condition, and aggregated the resulting phenotypes to create a \textit{phenotypic profile}.
Although organisms with different genotypes may express the same set of functions across environmental conditions, their phenotypic profiles may not necessarily be the same.
For example, an organism that expresses NOT in ENV-A and NAND in ENV-B has a distinct phenotypic profile from one that expresses NAND in ENV-A and NOT in ENV-B.
Because phenotypic profiles encapsulate function expression across all relevant environmental conditions (ENV-A and ENV-B), a change in phenotypic profile from parent to offspring indicates a mutationally-induced phenotypic change.
% -- GENETIC MANIPULATION / MUTATIONAL NEIGHBORHOOD --
% How do we produce mutants for analyses?
While most analyses employed here are retrospective metrics applied to lineages, digital evolution allows precise manipulations on individual organisms and genomes.
Mutational robustness uses this technique when looking at the possible mutations on a representative genotype.
Genomes in Avida are linear sequences of instructions, and as such, possible mutations can be simulated by substituting other instructions at the desired site.
Indeed, the mutational robustness of a genotype examines all one-step mutations (\textit{i.e.}, each mutation where exactly one instruction is substituted).
This allows us to disentangle whether the frequency of mutationally-induced phenotypic changes observed along a lineage is a consequence of evolved genetic architectures versus the result of population dynamics that actively skewed the distribution of organisms along the lineage; that is, are genomes organized such that mutations are more likely to induce phenotypic changes, or are organisms with phenotypes different from their parents more likely to be successful and thus appear along the representative lineage?
% the results of the lineage metrics are a consequence of evolved genetic architectures versus ongoing population dynamics that are actively skewing the distribution of organisms along each lineage. %or the results of selection.
%otherwise.
%purely the results of selection.
\subsection{Statistical analyses}
Across all of our experiments, we differentiated between sample distributions using non-parametric statistical tests.
For each major analysis, we first performed a Kruskal-Wallis test \citep{kruskal_use_1952} to
determine if there were significant differences in results from the PLASTIC, NON-PLASTIC, and STATIC treatments (significance level $\alpha=0.05$).
If so, we applied a Wilcoxon rank-sum test \citep{kotz_individual_1992} to distinguish between pairs of treatments.
We applied Bonferroni corrections for multiple comparisons \citep{rice_analyzing_1989} where appropriate.
\subsection{Software availability}
We conducted our experiments using a modified version of the Avida software, which is open source and freely available on \href{https://github.com/amlalejini/evolutionary-consequences-of-plasticity}{GitHub}~\citep{supplemental_material}.
Modifications to Avida included an improved phylogeny tracking system that enabled us to track coalescence events and the addition of custom sensory instructions specific to our experiments.
We used Python for data processing, and we conducted all statistical analyses using R version 4~\citep{r_core_team_r_v4}.
We used the tidyverse collection of R packages~\citep{r_tidyverse_2019} to wrangle data, and we used the following R packages for analysis, graphing, and visualization:
ggplot2~\citep{R-ggplot2},
cowplot~\citep{R-cowplot},
Color Brewer~\citep{harrower_colorbrewerorg_2003,R-Brewer_2014},
rstatix~\citep{R-rstatix},
ggsignif~\citep{R-ggsignif},
scales~\citep{R-scales},
Hmisc~\citep{R-Hmisc},
fmsb~\citep{R-fmsb},
and boot~\citep{R-boot}.
We used R markdown~\citep{rmarkdown} and bookdown~\citep{R-bookdown} to generate web-enabled supplemental material.
All of the source code for our experiments and analyses, including configuration files and guides for replication, can be found in our supplemental material, which is hosted on \href{https://github.com/amlalejini/evolutionary-consequences-of-plasticity}{GitHub}~\citep{supplemental_material}.
Additionally, our experimental data is available on the Open Science Framework at \url{https://osf.io/sav2c/} \citep{osf_data}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% INSTRUCTIONS:
% This section may be divided by subheadings. Footnotes should not be used and must be transferred to the main text.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Results}
\subsection{Adaptive phenotypic plasticity slows evolutionary change in fluctuating environments}
\begin{figure}[h!]
\centering
\includegraphics[width=1\textwidth]{media-evolutionary-change-magnitude-panel.pdf}
\caption{\small
\textbf{Magnitude of evolutionary change.}
Raincloud plots~\citep{allen_raincloud_2019} of
(a) coalescence event count,
(b) mutation count,
and (c) phenotypic volatility.
See Table~\ref{tab:metrics-definitions} for descriptions of each metric.
Each plot is annotated with statistically significant comparisons (Bonferroni-corrected pairwise Wilcoxon rank-sum tests).
Note that adaptive phenotypic plasticity evolved in \evolutionaryChangeRatePlasticReps\ of \evolutionaryChangeRateReplicates\ replicates from the PLASTIC treatment during phase one of this experiment; we used this more limited group to found \evolutionaryChangeRatePlasticReps\ phase-two PLASTIC replicates from which we report these PLASTIC data.
}
\label{fig:evolutionary-dynamics-magnitude}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Results to report (2021-02-08 experiment)
% ----- GENERATIONS -----
% - average generations elapsed (of a population)
% - NON-PLASTIC (median: 41768.65) > PLASTIC (median: 31697.65) > STATIC (median: 30839.75)
% condition mean sd
% <chr> <dbl> <dbl>
% 1 NON-PLASTIC 41090. 2702.
% 2 PLASTIC 31016. 2615.
% 3 STATIC 30002. 3011.
%
% ----- SWEEPS -----
% - coalescence events (total)
% - NON-PLASTIC (median: 663.5) > ( PLASTIC (median: 45.5) ~~ STATIC (median: 45) )
% - Average number of generations between coalescence events (gens / sweeps)
% - ( PLASTIC (median: 685.001780758557) ~~ STATIC (median: 693.676265008576) ) > NON-PLASTIC (median: 62.0184902295191)
%
% ----- PHENOTYPIC VOLATILITY -----
% - phenotypic volatility (total)
% - i.e., total number of times phenotypes change along lineages
% - NON-PLASTIC (median: 1868) > PLASTIC (median: 2) > STATIC (median: 0)
% - phenotypic volatility / lineage length
% - i.e., how often do genomic changes reflect changes in phenotype?
% - NON-PLASTIC (median: 0.437) > PLASTIC (median: 0.0022) > STATIC (median: 0)
% - phenotypic volatility / generations
% - i.e., per-offspring rate of phenotypic changes
% - NON-PLASTIC (median: 0.0447) > PLASTIC (median: 6.33e-05) > STATIC (median: 0)
%
% ----- MUTATION ACCUMULATION -----
% - mutation accumulation (total)
% - NON-PLASTIC (median: 4657.5) > STATIC (median: 1100) > PLASTIC (median: 998.5)
% - mutation accumulation / lineage length
% - NON-PLASTIC (median: 1.10048311715591) > STATIC (median: 1.03794597464116) > PLASTIC (median: 1.0328599144651)
% - mutation accumulation / generation
% - NON-PLASTIC (median: 0.11) > STATIC (median: 0.0368) > PLASTIC (median: 0.0319)
%
% ----- MUTATIONAL EFFECTS -----
% - fraction of mutational steps that alter (aggregate) phenotype
% - NON-PLASTIC (mean: 0.434007, CI [0.4242, 0.4406]) > PLASTIC (mean: 0.002717008, 0.0020, 0.0035) > STATIC (mean: 0.0006788834, CI [0.0004, 0.0009])
% - fraction of phenotype-altering mutation steps that alter unexpressed phenotype (PLASTIC condition only)
% - mean: 0.8247126 CI [0.7443, 0.8994]
% - fraction of mutations that affect unexpressed phenotype that are deleterious (PLASTIC only)
% - mean: 0.5172414 CI [0.4402, 0.5977]
% - fraction of mutations that affect unexpressed phenotype that are beneficial (PLASTIC only)
% - mean: 0.4827586 CI [0.4046, 0.5598]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% -- Magnitude of evolutionary change --
% - Selective sweeps
% - Mutation accumulation
% - Phenotypic volatility
In experimental phase 2A,
we tested whether adaptive phenotypic plasticity constrained or promoted subsequent evolutionary change in a fluctuating environment.
First, we compared the total amount of evolutionary change in populations evolved under the PLASTIC, NON-PLASTIC, and STATIC treatments as measured by coalescence event count, mutation count, and phenotypic volatility (Figure~\ref{fig:evolutionary-dynamics-magnitude}).
According to each of these metrics, NON-PLASTIC populations experienced a larger magnitude of evolutionary change than either PLASTIC or STATIC populations.
We observed significantly higher coalescence event counts in NON-PLASTIC populations than in PLASTIC or STATIC populations (Figure~\ref{fig:evolutionary-dynamics-magnitude}\hyperref[fig:evolutionary-dynamics-magnitude]{a}).
NON-PLASTIC lineages had significantly higher mutation counts (Figure~\ref{fig:evolutionary-dynamics-magnitude}b) and phenotypic volatility than PLASTIC or STATIC lineages (Figure~\ref{fig:evolutionary-dynamics-magnitude}c).
% -- Elapsed generations --
Changing environments have been shown to increase generational turnover (\textit{i.e.}, how rapidly generations elapse) in Avida populations~\citep{canino-koning_evolution_2016}, which could explain why we observe a larger magnitude of evolutionary change at the end of 200,000 updates of evolution in NON-PLASTIC populations.
Indeed, we found that significantly more generations of evolution elapsed in NON-PLASTIC populations (mean of $41090\pm2702$ std. dev.) than in PLASTIC (mean of $31016\pm2615$ std. dev.) or STATIC (mean of $30002\pm3011$ std. dev.) populations during phase 2A (corrected Wilcoxon rank-sum tests, p $<10^{-4}$).
% -- Rate of evolutionary change => intuition --
To evaluate whether increased generational turnover explains the greater magnitude of evolutionary change in NON-PLASTIC populations, we examined the average number of generations between coalescence events and the realized mutational robustness of lineages (Table \ref{tab:metrics-definitions}).
A coalescence event indicates a selective sweep, which is a hallmark of adaptive evolutionary change.
Realized mutational robustness measures the frequency that mutations cause phenotypic changes along a lineage.
We expect that static conditions should favor fit lineages with high realized mutational robustness that no longer undergo rapid adaptive change and hence do not trigger frequent coalescence events.
Under fluctuating conditions, however, lineages must be composed of plastic organisms if they are to maintain both high fitness and realized mutational robustness.
Without plasticity, we expect fluctuating conditions to produce lineages with low realized mutational robustness and frequent coalescence events as populations must continually acquire and fix mutations to readapt to the environment.
\begin{figure}[h!]
\centering
\includegraphics[width=0.66\textwidth]{media-evolutionary-change-pace-panel.pdf}
\caption{\small
\textbf{Pace of evolutionary change.}
Raincloud plots of
(a) average number of generations between coalescence events,
and (b) realized mutational robustness (Table \ref{tab:metrics-definitions}).
Each plot is annotated with statistically significant comparisons (Bonferroni-corrected pairwise Wilcoxon rank-sum tests).
}
\label{fig:evolutionary-dynamics-rate}
\end{figure}
% -- Rate of evolutionary change => findings --
On average, significantly fewer generations elapsed between coalescence events in NON-PLASTIC populations than in either PLASTIC or STATIC populations (Figure \ref{fig:evolutionary-dynamics-rate}a).
We also found that both STATIC and PLASTIC lineages exhibited higher realized mutational robustness relative to that of NON-PLASTIC lineages (Figure \ref{fig:evolutionary-dynamics-rate}b); that is, mutations observed along NON-PLASTIC lineages more often caused phenotypic changes in offspring.
Overall, our results indicate that NON-PLASTIC populations underwent more rapid (and thus a greater amount of) evolutionary change than either PLASTIC or STATIC populations.
% -- Mutational stability in static and plastic lineages --
While both STATIC and PLASTIC lineages exhibited high realized mutational robustness, we found that STATIC lineages exhibited higher realized robustness than PLASTIC lineages (Figure \ref{fig:evolutionary-dynamics-rate}b).
Overall, there were rare instances of mutations that caused a change in phenotypic profile across all PLASTIC lineages.
Of these mutations, we found that over 80\% (83 out of 102) of changes to phenotypic profiles were cryptic.
That is, the mutations affected traits that would not have been expressed in the environment that the organism was born into but would have been expressed had the environment changed.
\begin{figure}[ht!]
\centering
\includegraphics[width=0.33\textwidth]{media-mutational-robustness.pdf}
\caption{\small
\textbf{Mutational robustness.}
Raincloud plot of mutational robustness of each representative genotype (Table \ref{tab:metrics-definitions}).
The plot is annotated with statistically significant comparisons (Bonferroni-corrected pairwise Wilcoxon rank-sum tests).
}
\label{fig:mutational-robustness}
\end{figure}
% -- mutational landscaping --
% Thus far, our analyses have focused on dominant lineage.
% What about mutations off the line of descent?
% Mutational stability result is NOT due to random mutations being more likely to induce phenotypic change.
% KEY: Motivation for looking at mutational neighborhoods needs to made very clear here
% - Why not look at mutational neighborhoods for the other experiments?
Given that NON-PLASTIC lineages exhibited the lowest realized mutational robustness of our three experimental treatments, we sought to determine if this effect was driven by differences in evolved genetic architectures.
Specifically, did the NON-PLASTIC genetic architectures evolve such that mutations were more likely to result in phenotypic change?
Such a mutational bias would trade off descendant fitness in the same environment in exchange for a chance of increasing descendant fitness in alternate environments.
This strategy would be an example of diversifying bet-hedging (\textit{i.e.}, reducing expected mean fitness to lower variance in fitness) \citep{childs2010evolutionary}.
Alternatively, the lower realized mutational robustness in NON-PLASTIC lineages could be due to survivorship bias, as we measured realized mutational robustness as the fraction of mutations observed along \textit{successful} lineages that caused a phenotypic change.
That is, our measure of realized mutational robustness concentrates only on mutations that appear in the representative lineage (\textit{i.e.}, ``survivors'' of selection), ignoring mutations that did not.
Thus, lower realized mutational robustness in NON-PLASTIC lineages could simply be due to phenotype-altering mutations being more frequently favored by selection.
%(\textit{i.e.}, the lineages of extant organisms)
We analyzed the mutational robustness of representative genotypes by calculating the fraction of single-instruction mutations that change the phenotypic profile.
We found that mutations to representative genotypes on NON-PLASTIC lineages are \textit{less} likely to result in a phenotypic change than mutations to comparable genotypes on either STATIC or PLASTIC lineages (Figure \ref{fig:mutational-robustness}).
These data provide evidence against NON-PLASTIC lineages engaging in a mutation-driven bet-hedging strategy, and instead, are consistent with the hypothesis that lower realized mutational robustness in the NON-PLASTIC treatment was due to survivorship bias.
% @AML: Maybe one more sentence elaborating on implication of survivorship bias?
% -- in general, PLASTIC and STATIC more similar than NON-PLASTIC --
In general, adaptive plasticity stabilized PLASTIC-treatment populations against environmental fluctuations, and their evolutionary dynamics more closely resembled those of populations evolving in a static environment.
We observed no significant difference in the number and frequency of coalescence events in PLASTIC and STATIC populations.
We did, however, observe small, but statistically significant, differences in each of the following metrics: elapsed generations, mutation counts, phenotypic volatility, realized mutational robustness, and mutational robustness between PLASTIC and STATIC populations.
We expect that these differences are a result of plastic organisms needing to simultaneously maintain both function and function-regulation machinery, resulting in more genetic components that can be broken by mutations; moreover, many of these components are under relaxed selection in periods between environmental changes.
Overall, these differences were not substantial enough to play an obvious role in any of the dynamics we analyzed, but could be examined further in the future.
\subsection{Adaptively plastic populations retain more novel functions than non-plastic populations in fluctuating environments}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Results to report (2021-01-31)
% - Number of plastic replicates
% - Final dominant genotype # novel traits
% - non-plastic < (plastic == static)
% - Final population (1% threshold):
% - non-plastic < plastic < static
% - Final population (1% threshold) discovered:
% - non-plastic > (plastic ~~ static)
% - Lineage functions discovered
% - non-plastic > static ~>(nosig) plastic
% - Lineage functions discovered / step
% - (static ~~ plastic) > non-plastic
% - Lineage functions lost
% - non-plastic > static > plastic
% - Lineage functions lost / step
% - non-plastic > static > plastic
% - functions discovered per generation(?)
% - NON-PLASTIC [0.00014358046266055] ~~ STATIC [0.00015363220504867] > PLASTIC [0.000117695011124939]
% - functions lost per generation(?)
% - NON-PLASTIC [0.0022026054610079] > STATIC [0.000161396283669756] > PLASTIC [6.25141973661864e-05]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[h!]
\centering
\includegraphics[width=0.9\textwidth]{media-complex-traits-magnitude-panel.pdf}
\caption{\small
\textbf{Novel function evolution.}
Raincloud plots of
(a) final novel function count,
(b) novel function discovery,
and (c) novel function loss.
See Table \ref{tab:metrics-definitions} for descriptions of each metric.
Each plot is annotated with statistically significant comparisons (Bonferroni-corrected pairwise Wilcoxon rank-sum tests).
Note that adaptive phenotypic plasticity evolved in \novelTraitsPlasticReps\ of \novelTraitsReplicates\ replicates from the PLASTIC treatment during phase one of this experiment; we used this more limited group to seed the resulting \novelTraitsPlasticReps\ phase-two PLASTIC replicates.
}
\label{fig:complex-traits-magnitude}
\end{figure}
% -- What did we test? --
We have so far shown that adaptive plasticity constrains the rate of evolutionary change in fluctuating environments.
However, it is unclear how this dynamic influences the evolution of novel functions.
Based on their relative rates of evolutionary change, we might expect NON-PLASTIC-treatment populations to evolve more novel functions than PLASTIC-treatment populations.
But, how much of the evolutionary change in NON-PLASTIC populations is useful for exploring novel regions of the fitness landscape versus continually rediscovering the same regions?
% - Magnitude of exploration/exploitation -
To answer this question, we quantified the number of novel functions performed by a representative organism in the final population of each replicate.
We found that both PLASTIC and STATIC populations had significantly higher final function counts than NON-PLASTIC populations at the end of the experiment (Figure~\ref{fig:complex-traits-magnitude}a).
The final novel function count in PLASTIC and STATIC lineages could be higher than that of the NON-PLASTIC lineages for several non-mutually exclusive reasons.
One possibility is that PLASTIC and STATIC lineages could be exploring a larger area of the fitness landscape when compared to NON-PLASTIC lineages.
Another possibility is that the propensity of the NON-PLASTIC lineages to maintain novel traits could be significantly lower than PLASTIC or STATIC lineages.
When we looked at the total sum of novel functions discovered by each of the PLASTIC, STATIC, and NON-PLASTIC lineages, we found that NON-PLASTIC lineages generally explored a larger area of the fitness landscape (Figure~\ref{fig:complex-traits-magnitude}b).
Although the NON-PLASTIC lineages discovered more novel functions, those lineages also exhibited significantly higher novel function loss when compared to PLASTIC and STATIC lineages (Figure~\ref{fig:complex-traits-magnitude}c).
\begin{figure}[h!]
\centering
\includegraphics[width=0.66\textwidth]{media-complex-traits-pace-panel.pdf}
\caption{\small
\textbf{Rate of novel function evolution.}
Raincloud plots of
(a) novel function discovery frequency
and (b) novel function loss frequency.
Each plot is annotated with statistically significant comparisons (Bonferroni-corrected pairwise Wilcoxon rank-sum tests).
}
\label{fig:complex-traits-rate}
\end{figure}
% - Rates of exploration/exploitation -
A larger number of generations elapsed in NON-PLASTIC populations than in PLASTIC or STATIC populations during our experiment \citep{supplemental_material}.
Are NON-PLASTIC lineages discovering and losing novel functions more frequently than PLASTIC or STATIC lineages, or are our observations a result of differences in generational turnover?
To answer this question, we converted the metrics of novel function discovery and novel function loss to rates by dividing each metric by the number of elapsed generations along the associated representative lineages.
We found no significant difference in the frequency of novel function discovery between NON-PLASTIC and STATIC lineages, and we found that PLASTIC lineages had a lower frequency of novel function discovery than STATIC lineages (Figure~\ref{fig:complex-traits-rate}a).
Therefore, we cannot reject the possibility that the larger magnitude of function discovery in NON-PLASTIC lineages was driven by a larger number of elapsed generations.
NON-PLASTIC lineages had a higher frequency of function loss than either PLASTIC or STATIC lineages, and PLASTIC lineages tended to have a lower frequency of novel function loss than STATIC lineages (Figure~\ref{fig:complex-traits-rate}b).
% - Characterizing trait loss -
Next, we examined the frequency at which novel function loss along lineages co-occurred with the loss or gain of any of the six base functions.
Across all NON-PLASTIC representative lineages, over 97\% (10998 out of 11229) of instances of novel function loss co-occurred with a simultaneous change in base function profile.
In contrast, across all PLASTIC and STATIC dominant lineages, we observed that approximately 20\% (29 out of 142) and 2\% (13 out of 631), respectively, of instances of novel function loss co-occurred with a simultaneous change in base function profile.
As such, the losses of novel functions in NON-PLASTIC lineages appear to be primarily due to hitchhiking or epistatic effects where a mutation that knocks out a maladaptive base function (after the environment changes) also knocks out a beneficial novel function.
\subsection{Lineages without plasticity that evolve in fluctuating environments express more deleterious functions}
%%%%%%
% 2021-02-05 - Results
% - Number of offspring on lineage where hitchhiker instruction execution increases (i.e., instances of hitchhiking)
% - PLASTIC ~~ STATIC < NON-PLASTIC
% - Hitchhiker instruction increases / offspring on lineage
% - PLASTIC ~~ STATIC < NON-PLASTIC
% - What fraction of mutations that increase hitchhiker instruction execution co-occur with base trait changes?
% - NON-PLASTIC > PLASTIC ~~ STATIC
% - What about unexpressed vs expressed trait changes in plastic populations? (plastic only)
% - Not much hitchhiking. Did not find evidence that hitchiking occurring as cryptic variation in unexpressed phenotype.
%%%%%%
Phenotypic plasticity allows for genetic variation to accumulate in genomic regions that are unexpressed, which could lead to the fixation of deleterious instructions in PLASTIC populations.
However, in NON-PLASTIC lineages, we observe a higher rate of novel function loss, indicating that they may be more susceptible to deleterious mutations (Figure \ref{fig:complex-traits-rate}b).
Therefore, in experiment phase 2C, we tested whether adaptive phenotypic plasticity can increase the incidence of deleterious function performance.
Specifically, we added an instruction that triggered an explicitly deleterious function and measured the number of times it was executed.
Each execution of the \code{deleterious} instruction reduces an organism's fitness by 10\%.
At the beginning of phase 2C, the \code{deleterious} instruction is not present in the population, as it was not part of the instruction set during phase one of evolution.
Accordingly, if a \code{deleterious} instruction fixes in a population, it must be the result of evolutionary dynamics during phase 2C, including cryptic variation, hitchhiking, or as a result of sign epistasis where a \code{deleterious} instruction knocks out an even more maladaptive trait.
% sign epistasis; two things that are deleterious that are less deleterious in combination than you would expect form individual effects
% sign epistasis
\begin{figure}[ht!]
\centering
\includegraphics[width=1.0\textwidth]{media-poison-accumulation-panel.pdf}
\caption{\small
\textbf{Deleterious instruction accumulation.}
Raincloud plots of
(a) deleterious function acquisition,
(b) deleterious function acquisition frequency,
and (c) the proportion of mutations that increase deleterious function expression along a lineage that co-occur with a change in phenotypic profile.
Each plot is annotated with statistically significant comparisons (Bonferroni-corrected pairwise Wilcoxon rank-sum tests).
Note that adaptive phenotypic plasticity evolved in \deleteriousHitchhikingPlasticReps\ of \deleteriousHitchhikingReplicates\ replicates from the PLASTIC treatment during phase one of this experiment; we used this more limited group to seed the \deleteriousHitchhikingPlasticReps\ phase-two PLASTIC replicates.
}
\label{fig:deleterious-hitchhiking}
\end{figure}
% -- Instruction execution by final dominant & along lineage --
At the end of our experiment, no representative organisms from the PLASTIC or STATIC treatments performed the deleterious function under any environmental condition; however, representative organisms in 14\% of replicates of the NON-PLASTIC treatment performed the deleterious function at least once.
NON-PLASTIC lineages contained significantly more mutations that conferred the deleterious function as compared to PLASTIC or STATIC lineages (Figure \ref{fig:deleterious-hitchhiking}a), and these mutations occurred at a significantly higher frequency in NON-PLASTIC lineages (Figure \ref{fig:deleterious-hitchhiking}b).
% Additionally, we did not observe
% This result does not change when we normalize [what?] by the number of generations represented in the given lineage (Figure \ref{fig:deleterious-hitchhiking}b).
% -- When/where does hitchhiking take place? --
Next, we measured how often mutations that increased deleterious function performance co-occurred with changes to the base function profile within representative lineages.
A \code{deleterious} instruction can fix in a population by having a beneficial effect that outweighs its inherent cost (\textit{e.g.}, knocking out a punished function) or through linkage with a secondary beneficial mutation at another site within the genome.
Across all NON-PLASTIC representative lineages, we found that approximately 49\% (956 out of 1916) of mutations that increased deleterious function expression co-occurred with a change in the base function profile (Figure \ref{fig:deleterious-hitchhiking}c).
In all representative lineages from the PLASTIC treatment, only 18 mutations increased deleterious function expression, and none co-occurred with a change in base function profile (Figure \ref{fig:deleterious-hitchhiking}c).
Likewise, only 58 mutations increased deleterious function performance in all representative lineages from the STATIC treatment, and none co-occurred with a change in base function profile (Figure \ref{fig:deleterious-hitchhiking}c).
We did not find compelling evidence that the few mutations that increased deleterious function expression occurred as cryptic variation in PLASTIC lineages.
We repeated this experiment with 3\% and 30\% metabolic rate penalties associated with the deleterious function, which produced results that were consistent with those reported here \citep{supplemental_material}.
% GUIDELINES:
% This section may be divided by subheadings. Discussions should cover the key findings of the study: discuss any prior research related to the subject to place the novelty of the discovery in the appropriate context, discuss the potential shortcomings and limitations on their interpretations, discuss their integration into the current understanding of the problem and how this advances the current views, speculate on the future direction of the research, and freely postulate theories that could be tested in the future.
\section{Discussion}
In this work, we used evolving populations of digital organisms to determine how adaptive phenotypic plasticity alters subsequent evolutionary dynamics and influences evolutionary outcomes in fluctuating environments.
Specifically, we compared lineages of adaptively plastic organisms in fluctuating environments to both non-plastic organisms in those same environments and other non-plastic organisms in static environments.
\subsection{Evolutionary change}
% -- Adaptively plastic populations underwent less evolutionary change than non-plastic populations --
We found strong evidence that adaptive plasticity slows evolutionary change in fluctuating environments.
Adaptively plastic populations experienced fewer coalescence events and fewer total genetic changes relative to non-plastic populations evolving under identical environmental conditions (Figure \ref{fig:evolutionary-dynamics-magnitude}).
Whereas non-plastic populations relied on \textit{de novo} mutations to adapt to each environmental fluctuation, plastic populations leveraged sensory instructions to regulate function performance.
Indeed, in fluctuating environments, selection pressures toggle after each environmental change.
We hypothesize that in non-plastic populations such toggling would repeatedly drive the fixation of mutations that align an organism's phenotypic profile to the new conditions.
This hypothesis is supported by the increased frequency of coalescence events in these populations (Figure \ref{fig:evolutionary-dynamics-rate}a) as well as increased rates of genetic and phenotypic changes observed along the lineages of non-plastic organisms.
% - Mutational neighborhood results
Representative lineages in the non-plastic treatment experienced lower realized mutational robustness than plastic and static lineages (Figure~\ref{fig:evolutionary-dynamics-rate}b).
We reasoned that this lower realized mutational robustness was due to non-plastic populations evolving a bet-hedging strategy where mutations are more likely to modify the phenotypic profile.
However, when we switched from measuring the realized mutational robustness of representative lineages to measuring the mutational robustness of representative genotypes (\textit{i.e.}, what fraction of one-step mutants change the phenotypic profile), we observed that non-plastic genotypes exhibited the highest mutational robustness of all three treatments (Figure~\ref{fig:mutational-robustness}).
This result runs contrary to both our expectations and the results of other fluctuating environment studies in Avida~\citep{canino-koning_fluctuating_2019}.
\cite{canino-koning_fluctuating_2019} found that mutational robustness is negatively correlated with the number of function-encoding sites in the genome.
In our work, most plastic and static genotypes encode all six base functions, while most non-plastic genotypes only encode functions from one environment; this results in fewer function-encoding sites, which may increase mutational robustness in non-plastic genotypes (relative to plastic and static genotypes). %relative to genotypes from other treatments.
Regardless of the cause, this higher mutational robustness in non-plastic organisms indicates that bet-hedging is not driving the low realized mutational robustness observed in non-plastic lineages.
Thus, we expect the lower realized mutational robustness in non-plastic lineages to be driven by survivorship bias.
Because non-plastic lineages must rely on mutations to adapt to environmental changes, phenotype-altering mutations are often highly advantageous, and their selection decreases the realized mutational robustness of successful lineages.
% -- in context of previous digital evolution work --
To our knowledge, this study is the first in-depth empirical investigation into how the \textit{de novo} evolution of adaptive plasticity shifts the course of subsequent evolution in a cyclic environment.
The relative rates of evolutionary change that we observed in non-plastic populations, however, are consistent with results from previous digital evolution studies.
For example, \cite{dolson_interpreting_2020} showed that non-plastic populations that were evolved in cyclically changing environments exhibited higher phenotypic volatility and accumulated more mutations than that of populations evolved under static conditions.
Furthermore, \cite{lalejini_evolutionary_2016} visually inspected the evolutionary histories of non-plastic organisms evolved in fluctuating environments, observing that mutations along successful lineages readily switched the set of traits expressed by offspring.
%\cite{canino-koning_evolution_2016} also observed that genomes evolved in harsh cyclic environments often contained vestigial fragments of genetic material adapted to prior environments.
% -- In context of conventional evolutionary theory: evo response => f(selection, variation) --
Our results are also consistent with conventional evolutionary theory.
A trait's evolutionary response to selection depends on the strength of directional selection and on the amount of genetic variation for selection to act upon \citep{lande_measurement_1983,zimmer_evolution_2013}.
In our experiments, non-plastic populations repeatedly experienced strong directional selection to toggle which functions were expressed after each environmental change.
As such, retrospective analyses of successful lineages revealed rapid evolutionary responses (that is, high rates of genetic and phenotypic changes).
Evolved adaptive plasticity shielded populations from strong directional selection when the environment changed by eliminating the need for a rapid evolutionary response to toggle function expression.
Indeed, both theoretical and empirical studies have shown that adaptive plasticity can constrain evolutionary change by weakening directional selection on evolving populations \citep{price_role_2003,paenke_influence_2007,ghalambor_non-adaptive_2015}.
% TODO, consider: Expectations that could arise if we took into consideration lineages off the line of descent.
\subsection{The evolution and maintenance of novel functions}
% -- Exploration + Exploitation --
In fluctuating environments, non-plastic populations explored a larger area of the fitness landscape than adaptively plastic populations (Figure \ref{fig:complex-traits-magnitude}b).
However, adaptively plastic populations better exploited the fitness landscape, retaining a greater number of novel functions than non-plastic populations evolving under identical environmental conditions (Figure \ref{fig:complex-traits-magnitude}a).
In our experiment, novel functions were less important to survival than the fluctuating base functions.
In non-plastic populations, when a mutation changes a base function to better align with current environmental conditions, its benefit will often outweigh the cost of losing one or more novel functions.
Indeed, we found that along non-plastic representative lineages, 97\% of the mutations associated with novel function loss co-occurred with phenotypic changes that helped offspring adapt to current environmental conditions.
% -- Changing environments promote evolutionary change --
Previous studies have shown that transitory environmental changes can improve overall fitness landscape exploration in evolving populations of non-plastic digital organisms \citep{nahum_improved_2017}.
Similarly, changing environments have been shown to increase the rate of evolutionary adaptation in simulated network models \citep{kashtan2007varying}.
In our system, however, we found that \textit{repeated} fluctuations reduced the ability of non-plastic populations to maintain and exploit functions; that said, we did find that repeated fluctuations may improve overall function discovery by increasing generational turnover.
Consistent with our findings, \cite{canino-koning_fluctuating_2019} found that non-plastic populations of digital organisms evolving in a cyclic environment maintained fewer novel traits than populations evolving in static environments.
% -- Plastic rescue, stabilizing effect of plasticity --
Our results suggest that adaptive phenotypic plasticity can improve the potential for populations to exploit novel resources by stabilizing them against stressful environmental changes.
The stability that we observe may also lend some support to the hypothesis that phenotypic plasticity can rescue populations from extinction under changing environmental conditions \citep{chevin_adaptation_2010}.
% -- relevance to genes as followers hypothesis --.
Our data do not necessarily provide evidence for or against the genes as followers hypothesis.
The genes as followers hypothesis focuses on contexts where plastic populations experience novel or abnormally stressful environmental change.
However, in our system, environmental changes were cyclic (not novel), and no single environmental change was \textit{abnormally} stressful.
Further, the introduction of novel functions during the second phase of the experiment merely added static opportunities for fitness improvement.
This addition did not change the meaning of existing environmental cues, nor did it require those cues to be used in new ways.
\subsection{The accumulation of deleterious alleles}
% -- Overview of results --
% - More accumulation in non-plastic
% - no evidence for cryptic variation housing poison
% - plastic ~~ static
We found that non-plastic lineages that evolved in a fluctuating environment exhibited both greater totals and higher rates of deleterious function acquisition than that of adaptively plastic lineages (Figure \ref{fig:deleterious-hitchhiking}).
There are several, non-mutually exclusive possibilities that could explain the fixation of explicitly deleterious instructions: random genetic drift, deleterious hitchhiking, epistatic effects, and cryptic variation (in plastic organisms).
We find it unlikely that random genetic drift explains our observations.
Each time an organism expresses a \code{deleterious} instruction, the organism incurs a 10\% penalty to their replication rate, which results in strong purifying selection against mutations that cause offspring to execute \code{deleterious} instructions.
In asexual populations without horizontal gene transfer, all co-occurring mutations are linked.
As such, deleterious mutations linked with a stronger beneficial mutation (\textit{i.e.}, a driver) can sometimes ``hitchhike'' to fixation \citep{smith_hitch-hiking_1974,van_den_bergh_experimental_2018,buskirk_hitchhiking_2017}.
Natural selection normally prevents deleterious mutations from reaching high frequencies, as such mutants are outcompeted.