1
A NOVEL LABORATORY ACTIVITY FOR TEACHING ABOUT THE EVOLUTION OF
2
MULTICELLULARITY
3
4
William C. Ratcliff1, Alison Raney2, Sam Westreich2, Sehoya Cotner2*
5
6
7
1
Ecology, Evolution and Behavior, University of Minnesota, Minneapolis, MN, 55108 USA
2
8
Biology Program, University of Minnesota, Minneapolis, MN, 55108 USA
9
10
*Correspondence: Sehoya Cotner, harri054@umn.edu, phone: 612-626-2385 fax: 612-626-
11
7823
12
1
13
A NOVEL LABORATORY ACTIVITY FOR TEACHING ABOUT THE EVOLUTION OF
14
MULTICELLULARITY
15
16
Abstract
17
The evolution of complexity remains one of the most challenging topics in biology to teach
18
effectively. Here we present a novel laboratory activity, modeled on a recent experimental
19
breakthrough, in which students experimentally evolve simple multicellularity using single-celled
20
yeast (Saccharomyces cerevisiae). By simply selecting for faster settling through liquid media,
21
yeast evolve to form snowflake-shaped multicelled clusters that continue to evolve as
22
multicellular individuals. This paper presents core experimental and curriculum tools, including
23
discussion topics and assessment instruments, and provides suggestions for teacher
24
customization. Pre and post-lab assessment demonstrates that this lab effectively teaches
25
fundamental concepts about the transition to multicellularity. Yeast strains, the student
26
laboratory manual and an introductory presentation are available free of charge.
27
28
Keywords: multicellularity, evolution, complexity, authentic, experimental evolution, curriculum
29
30
2
31
Introduction
32
A growing body of literature documents a range of strategies for teaching about evolution (see
33
Wei, et al., 2012, and references therein), including direct experimentation with live organisms
34
(e.g., Drosophila a la Plunkett and Yampolsky, 2010), role-playing (e.g., Reichert et al., 2011)
35
and computer simulation (e.g., Abraham et al., 2012). Financial and temporal constraints
36
typically limit hands-on experimentation to a few generations, suitable only for microevolutionary
37
questions. As a result, large-scale phenotypic change (macroevolution) is taught only via
38
lecture, computer simulations or other abstract representation. The origin of multicellularity from
39
unicellular ancestors was a major transition in evolution (Maynard Smith and Szathmary, 1995).
40
Despite its fundamental importance in the evolution of large, complex organisms, the evolution
41
of multicellularity has never (to our knowledge) been incorporated into student labs or active-
42
learning exercises. This omission is unfortunate, for copious work (Simurda, 2012, Armbruster
43
et al., 2009, Bailey et al., 2012; also see Bransford, 1999 for review) demonstrates the value of
44
using hands-on methodology to promote deeper understanding and internalization.
45
46
Ratcliff et al. (2012) carried out a novel experiment to evolve simple multicellularity in the lab,
47
starting with single-celled microbes. The authors created an environment that favored strains
48
that evolve to form clusters of cells (the first step in the transition to multicellularity) by
49
subjecting baker’s yeast (Saccharomyces cerevisiae) to daily selection for fast settling through
50
liquid media. Within just a few weeks, yeast that formed snowflake-shaped clusters of cells
51
evolved and displaced their single-celled ancestors. “Snowflake” yeast display several hallmarks
52
of multicellularity, including juvenile and adult life stages, determinate growth, and a rudimentary
53
cellular division of labor utilizing programmed cell death.
54
55
Here we describe a new laboratory exercise, modeled on the experiment of Ratcliff et al. (2012).
56
Using simple procedures and inexpensive materials, we have developed and tested a three-
57
week lab that involves students in cutting-edge research and encourages them to think critically
58
about the evolution of multicellularity.
59
60
Methods
61
In Spring 2012, over 300 university students enrolled in either Animal Diversity or General
62
Zoology—organismal biology courses with an introductory-biology prerequisite—participated in
63
an “Evolution of Multicellularity” laboratory exercise. Students were divided into 16 sections of
3
64
~18 students. Undergraduate teaching-assistants oversaw the practical aspects of the lab
65
activities.
66
67
The laboratory exercise
68
The complete written lab protocol (for students) and introductory PowerPoint presentation are
69
available at: http://www.cbs.umn.edu/bioprog/staff/cotner/
70
71
The following materials are necessary, and with the exception of an autoclave and shaking
72
incubator, relatively simple and inexpensive to obtain:
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
 Unicellular yeast (strain Y55)*
 Snowflake yeast, evolved after 3 weeks of selection (strain Y55_wk3)*
 YPD media (recipe in Box)
 Test tubes for cell culture (25 mm X 150 mm)
 Test tube racks
 1.5 mL microcentrifuge tubes
 100 µl micropipettors
 Micropipette tips
 Serological pipettes – (5 mL)
 Serological pipettes – (25 mL)
 Bulbs for the serological pipettes
 22x22 mm coverslips
 Microscope slides (plain)
 Microscope slides (concave)
 Autoclave
 Shaking incubator
 Compound microscope
 Rulers
*Yeast strains can be obtained free of charge by contacting S. Cotner (harri054@umn.edu).
93
The lab procedure spans three weeks, with daily settling selection and transfer to fresh media.
94
Each day, students perform a round of settling selection, and then transfer surviving yeast to
95
fresh media (see Figure 1 for an overview). Students start with either unicellular yeast (strain
96
Y55), or a snowflake strain that previously evolved during 3 weeks of daily transfer (Y55_wk3)
97
using the “faster settling selection” regime (Figure 1a). Using both strains allows the students to
98
examine 6 weeks of evolutionary time in only three weeks, or approximately 300 generations of
99
yeast.
100
101
Using unicellular yeast (str. Y55), students will select only for faster settling, favoring strains that
102
evolve to form clusters. Using snowflake yeast (str. Y55_wk3), students will examine the ability
103
of selection to act on differences among clusters, selecting for either faster or slower settling.
4
104
105
***************************************************************************************************
106
BOX. PROTOCOLS
107
108
109
YPD media preparation
Ingredients:
110
o
20 g dextrose
111
o
20 g peptone
112
o
10 g yeast extract
113
 Dissolve the above ingredients into 1 L water.
114
 Bring the final volume up to 2 L.
115
 Using the 25 mL serological pipette, aliquot 5 mL YPD into each 25 mm X 150 mm test tube,
116
and place into test tube racks.
117
 Cap tubes.
118
 Autoclave to sterilize.
119
120
Initial tube inoculation
121
 Obtain three test tubes containing 5 mL YPD.
122
 Inoculate one tube with 100 µl of the Y55 stock culture, and two tubes with 100 µl of
123
124
125
Y55_wk3.
 Label each tube with the inoculation date, strain, and selection scheme (i.e., faster or slower
settling). Incubate overnight (30ºC, shaking).
126
127
128
129
130
Settling selection (days 2+). See Figure 1 for illustrations.
 Obtain three test tubes containing 5 mL YPD. Label tubes with transfer number, strain, and
selection scheme.
 Obtain a 5 mL serological pipette with bulb.
131
132
Selection for faster settling
133
 Suck up 5 mL of the turbid yeast culture into a 5 mL serological pipette.
134
 Allow the pipette to stand upright in a glass beaker or pipette rack for 10 minutes. Make sure
135
136
the media is not leaking out. If the pipette leaks, seal the bottom with a piece of parafilm.
 Transfer the bottom 0.5 mL to sterile YPD.
137
5
138
Selection for slower settling
139
Same as above, but after 10 minutes of settling, discard the bottom 4.5 mL of media in the
140
pipette, transferring the top 0.5 mL to sterile YPD.
141
142
Repeat the settling selection ~20 times, viewing samples under the microscope at weekly
143
intervals to track progress. Labs that only meet once or twice a week will require careful
144
planning: daily transfers can be accomplished by allowing separate lab sections to transfer the
145
same populations. Instructors, TAs and students may be required to perform transfers when no
146
class is scheduled to meet. When daily selection cannot be performed (such as during a
147
weekend), place the selection lines in the refrigerator.
148
***************************************************************************************************
149
150
151
152
Figure 1. Settling selection protocols illustrated. Selection regime for faster settling (A), or
153
slower settling (B). Snowflake yeast evolve faster settling by evolving larger size. Shown in (C)
154
is an isolate taken from 14 transfers, and in (D) 60 transfers. The cluster in (D) is larger both
6
155
because there are more cells per cluster, but also because the size of individual cells has
156
increased.
157
158
Analyses to perform on the last lab
159
By the last lab, each selection line should have undergone about 20 transfers, or ~150
160
generations. As time permits, have the students measure the following traits on both the
161
evolved lines and the ancestors used on day 1 (regrown from stock cultures):
162
1) Cells per cluster. Faster settling mainly results from the evolution of larger clusters.
163
To measure cells per cluster, students first need to dilute each population grown in
164
YPD 10-fold into nonsterile water, adding 100 µl culture to 900 µl water in a 1.5 mL
165
microcentrifuge tube. Mix by inversion, then place 5 µl onto a standard slide with
166
coverslip, and view on a compound scope. Have each student or group of students
167
find the largest cluster on a slide and count the number of cells in it.
168
2) Cell size. Larger clusters of Y55_wk3 will sometimes evolve as a consequence of an
169
increase in the size of individual cells. Think of this as building a larger house not by
170
changing the design, but simply by using larger bricks (Figure 1 C,D). Have students
171
note the size of cells in the evolved lines relative to the single-celled ancestor.
172
3) Settling speed. Suck up 5 mL of each of the five populations (two inoculum strains
173
plus three selection lines) into a serological pipette, stand upright for 5 minutes.
174
Using a ruler, measure the height of the cell pellet that forms at the bottom of the
175
pipette. This measurement is a proxy for settling rate.
176
177
Measures 1 and 3 can be reported to the teacher and graphed for the whole class, allowing
178
students to examine the distribution of maximum cluster size and settling speed for their evolved
179
lines, relative to their starting lines, across multiple independently evolving replicate populations.
180
181
Additional exploration
182
Expanding on the above selection schemes is possible and encouraged, but depends on time,
183
student interests and abilities, and equipment availability. Additional projects could focus on
184
other ecological scenarios that could favor the evolution of cell clusters, such as evasion of
185
small-mouthed predators (e.g., Daphnia or Paramecium), protection from UV exposure or
186
desiccation, or increased resistance to antibiotics. Students can investigate various hypotheses
187
about size-related advantages using simple, inexpensive and readily available supplies (e.g.,
188
UV lamps and live zooplankton). Students with access to a fluorescent compound microscope
7
189
can explore the evolution of programmed cell death (apoptosis) in snowflake yeast through
190
live/dead staining, following protocols described in Ratcliff et al (2012).
191
192
Possible Discussion Topics
193
Certain themes may arise during post-lab discussions or written reflections, including the
194
following:
195
196
 What is experimental evolution?
o
Experimental evolution involves testing hypothesis about evolutionary processes
197
in controlled laboratory or field settings. By observing evolutionary change in an
198
organism, such as yeast over generations, students can view evolution in action
199
in an ecologically relevant context. With organisms that have short generation
200
times, experimental evolution allows students to see morphological change
201
occurring over relatively short timescales.
202
203
 Did multicellularity arise more than once?
o
Multicellularity has evolved at least 25 times independently (Grosberg and
204
Strathmann, 2007), resulting in multicellular organisms as diverse as animals,
205
plants, seaweed and fungi. The purpose of this exercise is not to suggest that
206
settling selection led to the evolution of multicellularity in any of these groups.
207
Instead, the goal is to demonstrate that simple environmental constraints and
208
relatively few generations could have been sufficient for the evolution of a key
209
step in these transitions.
210
211
 What are benefits and costs of multicellularity?
o
Early multicellular organisms likely benefit from their larger size, resisting things
212
like predation, toxin or UV exposure, or desiccation. More derived multicellular
213
organisms benefit from cooperation among component cells. This division of
214
labor can allow multicellular organisms to perform some tasks more efficiently,
215
and can facilitate the evolution of complex traits (like differentiated tissues) that
216
provide novel multicellular-level functionality.
217
o
Multicellularity is not without its costs, however. Multicellular organisms require
218
more resources and take longer to mature than unicellular organisms. Further,
219
their complexity makes them more vulnerable to disruption. For example,
220
conflicts among cells within an organism, such as cancer, can lead to dissolution
221
and death of the organism; cancer is a problem not faced by single-celled
222
organisms.
8
223
224
 Is multicellularity reversible?
o
In principle, all multicellular organisms should be able to evolve back to
225
unicellularity. It may be difficult to imagine complex multicellular organisms like
226
vertebrate animals evolving back to unicellularity, because individual cells are so
227
specialized and integrated into the multicellular whole. However, whether or not
228
increased multicellular complexity makes it more challenging to evolve back to
229
unicellularity remains an open question. The HeLa cell line presents an
230
interesting example of the evolution of unicellularity from a human ancestor.
231
Cervical cancer cells isolated from Henrietta Lacks have been grown in the lab
232
since 1951, and are so widely used that their collective biomass is estimated to
233
be about 50 million tons. HeLa cells are exceptionally adapted to life as a single-
234
celled microbe, and have become a major contaminant of laboratory cultures
235
(Skloot, 2011). Their success as a unicellular organism has even led to a formal
236
description as a new species of microbe, named Helacyton gartleri (Van Valen,
237
1991).
238
239
 What is the distinction between a multicellular organism and a multi-celled cluster?
o
Scientists have not reached consensus on what constitutes a multicellular
240
organism. Like other prominent concepts in biology (e.g., what counts as alive,
241
what species are, etc.), determining if an organism is multicellular is trivial at the
242
extremes of the spectrum (e.g. E. coli and blue whales), but challenging for
243
intermediate states. A key question is: when does a cluster of cells stop being a
244
group of single-celled organisms, and become one multicellular organism?
245
246
One way to answer this is to determine if single cells are still 'individuals', or if
247
they are 'parts' that work together for the benefit of the multicellular individual,
248
like skin cells in animals. Put into evolutionary terms, once the Darwinian fitness
249
(survival and reproduction) of single cells within the cluster is less important than
250
the contribution of single cells to the fitness of the cluster, then the group of cells
251
can be considered to be a multicellular organism. Ratcliff et al (2012) show that
252
single cells in snowflake yeast evolve to commit cellular suicide (apoptosis), to
253
act as 'break points' within the multicellular cluster. This reduces the fitness of
254
single cells (they die), but increases the fitness of the cluster, allowing it to
255
regulate the number and size of propagules they produce. So by this definition,
9
256
snowflake yeast (at least those that have evolved apoptosis) can be considered
257
simple multicellular organisms.
258
259
Assessment
260
In order to determine the level of background knowledge demonstrated by the students,
261
individual assessments were given to each student before the laboratory exercise (see
262
Appendix 1). Nine weeks after the laboratory exercise, the students were each given a similar
263
post-exercise assessment, designed to test their retention of the material. Statistics were
264
calculated only for students that completed both pre and post-exercise assessments. Matched-
265
pairs t-tests were performed for each question, with error from multiple comparisons controlled
266
with a Bonferroni correction.
267
268
Results & Discussion
269
Students demonstrated a noticeable improvement in the understanding of the underlying
270
concepts behind the evolution of multicellularity. Students were noticeably better at explaining
271
various concepts--such as cellular division of labor, cell-to-cell communication, the differences
272
between specificity and differentiation, and how multicellularity can be more efficient--on the
273
post-exercise assessment, in comparison with the pre-exercise assessment. Students were also
274
able to articulate specific benefits and weaknesses of multicellularity on the post-exercise
275
assessment.
276
277
Substantial improvement was seen for the question in which students identified that
278
“multicellularity arose several times, in several independent lineages (Figure 2);” the pre-lab
279
average was 66.2% correct and the post-lab average was 79.7% correct (t147 = 2.84; p<0.01).
280
This finding indicates that not only did most of the students understand that multicellularity
281
evolved in multiple lineages, but that the exercise was able to emphasize this idea. Students
282
also showed noticeable improvement in listing advantages of multicellularity: students were able
283
to list an average of 2.68 logical benefits on the pre-lab quiz, and 2.89 benefits after the
284
exercise and discussion (89.9% and 96.3% of the maximum number of correct answers,
285
respectively. t147= 4.67; p<0.001). Similarly, prior to the lab students could list an average of
286
2.07 possible disadvantages of multicellularity, compared to a post-lab average of 2.53 (69%
287
and 84.3% of the maximum number of correct answers, respectively t147= 6.12; p<0.001).
288
10
289
290
Figure 2. In a pre- and post-lab assessment, students showed improvement in their
291
understanding of the origins of multicellularity (q2), and could better articulate possible benefits
292
(q3) and disadvantages (q4) of multicellularity over single-celled existence. Error bars are the
293
SEM, asterisks indicate statistical significance below p=0.05 (corrected for multiple
294
comparisons). Students did not show significant improvement in their understanding which
295
groups of organisms are multicellular (question 1). See Appendix 1 for a full list of assessment
296
instruments.
297
298
In the pre-exercise assessment, the most commonly described benefits of multicellularity were
299
physical characteristics of the organism, such as an improved ability to gather food, adapt to
300
novel environments, and survive after the death of individual cells. In the post-exercise
301
assessment, students were more likely to discuss benefits such as division of labor, specificity
302
of cell tasks, and physical differentiation of cells increasing the overall energy efficiency of the
303
individual. Similarly, students tended to list physical qualities (higher risk of predation, reduced
304
surface area to volume ratio) as disadvantages of multicellularity on the pre-exercise
305
assessment, while they favored innate qualities related to the evolution of multicellularity (higher
306
total energy costs, conflicts among cells, greater fragility due to higher complexity) on the post-
307
exercise exam. This change demonstrates a shift in thinking from merely observing the physical
11
308
effects of a specific quality to considering the greater implications of evolving into
309
multicellularity, as well as the associated benefits and the obstacles that must be overcome by
310
the organism.
311
312
These trends are especially promising, as the post-exercise assessment was not given to the
313
students immediately after completion of the laboratory exercise, but nine weeks later. Thus,
314
students demonstrated that they were able to master and apply the core concepts from the
315
laboratory, and were able to retain these concepts and apply them to similar problems more
316
than two months afterwards.
317
318
Recommendations
319
We encourage our colleagues to modify this lab activity to suit their curricular needs. However,
320
we can make a few recommendations based on our experiences. Brief introductory materials,
321
discussing the basics of yeast biology and the advent of multicellularity, were sufficient for
322
engaging students in a discussion of the experiment and associated predictions. Clear
323
directions, tailored from the Ratcliff et al (2012) experiment, should be given to both the
324
students and the teachers/staff. These directions should include a day-to-day plan for the first
325
week to ten days, specifically instructing teaching staff in how to refresh the growth media and
326
how to maintain the selection regime. After the first week, the process—growth, selection, and
327
inoculation of new media—repeats itself.
328
329
We recommend sharing a week-by-week layout with the students, outlining the plan for the
330
duration of the exercise. Student lab materials should contain ample writing and drawing space,
331
so that the students can describe what they are seeing on the slides (prepared from small
332
amounts of yeast from both the unicellular to multicellular selection lines, and the divergent
333
selection lines) in words and pictures, and collect appropriate data. Whenever possible, we
334
recommend that data for the entire section is pooled, allowing the students to examine the
335
evolution of snowflake yeast traits in multiple independently evolving lineages.
336
337
Conclusion
338
Yeast, with their short generation time and large population sizes, are an ideal model for
339
teaching about evolution and natural selection. This lab activity not only allows the students to
340
explore the process of evolution through the origins of multicellularity, but can also lead to
341
thoughtful discussion of (1) how multicellularity arose more than once, (2) how multicellularity
12
342
incurs costs and is, as predicted, reversible; and (3) the distinction between multicellularity and
343
colonial unicellular organisms. Additionally, students become familiar with materials (e.g.
344
micropipettes and growth media) that are used in advanced teaching and research laboratories.
345
346
Hands-on activities that are authentic (i.e., using real organisms and legitimate unexplained
347
phenomena) are ideal for clarifying misconceptions and affecting conceptual change (e.g.,
348
pertaining to evolutionary biology; see Chinsamy and Plagányi, 2008). Few topics in
349
evolutionary biology challenge educators like the origins of complexity. The popularity of non-
350
scientific explanations, such as Intelligent Design and other forms of creationism (Scott and
351
Matzke, 2007), attests to our limited ability to address complexity in a scientifically meaningful
352
way. Thus, we are especially attracted to this demonstration of a legitimate, intellectually
353
available mechanism for the origins of multicellularity.
354
355
Acknowledgements
356
This work was supported by the National Science Foundation, grant no. DEB-1051115.
357
13
358
Literature Cited
359
Abraham, J.K., Perez, K.E., Downey, N., Herron, J.C. & Meir, E., 2012, Short Lesson Plan
360
Associated with Increased Acceptance of Evolutionary Theory and Potential Change in Three
361
Alternate Conceptions of Macroevolution in Undergraduate Students, CBE-Life Sciences
362
Education, 11(2), pp. 152-64.
363
364
Armbruster, P., Patel, M., Johnson, E. & Weiss, M., 2009, Active learning and student-centered
365
pedagogy improve student attitudes and performance in introductory biology, CBE-Life
366
Sciences Education, 8(3), pp. 203-13.
367
368
Bailey, C.P., Minderhout, V. & Loertscher, J., 2012, Learning transferable skills in large lecture
369
halls: Implementing a POGIL approach in biochemistry, Biochemistry and Molecular Biology
370
Education.
371
372
Chinsamy, A. and Plagányi, E. 2008. Accepting evolution. Evolution, 62, pp. 248-254.
373
Grosberg, R.K. & Strathmann, R.R., 2007, The evolution of multicellularity: a minor major
374
transition? Annu. Rev. Ecol. Evol. Syst., 38, pp. 621-54.
375
376
Plunkett, A.D. & Yampolsky, L.Y., 2010, When a Fly Has to Fly to Reproduce: Selection Against
377
Conditional Recessive Lethals in Drosophila, The American Biology Teacher, 72(1), pp. 12-5.
378
379
Ratcliff, W.C., Denison, R.F., Borrello, M. & Travisano, M., 2012, Experimental evolution of
380
multicellularity, Proceedings of the National Academy of Sciences, 109(5), pp. 1595-600.
381
382
Riechert, S.E., Leander, R.N. & Lenhart, S.M., 2011, A Role-Playing Exercise That
383
Demonstrates the Process of Evolution by Natural Selection: Caching Squirrels in a World of
384
Pilferers, The American Biology Teacher, 73(4), pp. 208-12.
385
386
Scott, E. and Matzke, N. 2007, Biological design in science classrooms. Proc. Natl. Acad. Sci.
387
USA, 104 Suppl. 1, pp. 8669-8676.
388
14
389
Simurda, M.C., 2012, Does the Transition to an Active-Learning Environment for the
390
Introductory Course Reduce Students’ Overall Knowledge of the Various Disciplines in
391
Biology? Journal of Microbiology & Biology Education, 13(1).
392
393
Skloot, Rebecca. The Immortal Life of Henrietta Lacks. 2011. London, England.
394
395
Van Valen, LM. 1991, HeLa, a new microbial species. Evolutionary Theory 10: 71-74
396
Wei, C.A., Beardsley, P.M. & Labov, J.B., 2012, Evolution Education across the Life Sciences:
397
Making Biology Education Make Sense, CBE-Life Sciences Education, 11(1), pp. 10-6.
398
399
15
400
Appendix 1- Questions used in the assessment
401
402
1. Of the following groups, which are exclusively multicellular (circle any/all that apply)?
403
404
a) Animals
405
b) Land plants
406
c) Fungi
407
d) Bacteria
408
e) Algae
409
410
2. Which of the following best characterizes our current understanding of the evolution of
411
multicellularity?
412
413
a) Multicellularity arose several times, but only in a single lineage
414
b) Multicellularity arose once, and gave rise to all animals
415
c) Multicellularity arose several times, in several independent lineages
416
d) Multicellularity arose once, un an organism like present-day Volvox
417
e) Multicellularity arose once, in an organism much like present-day Baker’s yeast
418
419
3. List three possible advantages of multicellularity.*
420
421
4. List three possible disadvantages of multicellularity.*
422
423
While not included in our first assessments, we recommend these questions for use in your
424
class:
425
426
427
5. What distinguishes a multicellular organism from a clump of cells or a colony of
unicellular organisms?
428
429
430
6. Describe a plausible scenario by which multicellular organisms could evolve from
unicellular ancestors. Pictures are encouraged!
431
432
* In the post-lab assessment, these questions were changed to “What are some possible
433
advantages [for Question 3] or disadvantages [for Question 4] of multicellularity (list as many as
16
434
you can think of).” We scored both pre- and post-lab quizzes out of a maximum of three correct
435
responses. Because some students listed more than three advantages/disadvantages in the
436
post lab, this estimate is conservative, likely underestimating improvement.
17