Broad assessment of bioactivity of a collection of spiroindane pyrrolidines through “cell painting”

Article abstract of DOI:10.1016/j.bmc.2020.115547

A collection of small molecules has been synthesized by composing photo-cycloaddition, C–H functionalization, and N-capping strategies. Multidimensional biological fingerprints of molecules comprising this collection have been recorded as changes in cell

Full text of DOI:10.1016/j.bmc.2020.115547

Bioorganic&MedicinalChemistry28(2020)115547
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
Broad assessment of bioactivity of a collection of spiroindane pyrrolidines
through “cell painting”
⁎
Manvendra Singha, Nathan Garzaa, Zachary Pearsona, Justin Douglasa,b, Zarko Boskovica,
a Department of Medicinal Chemistry, University of Kansas, Lawrence, 66045 KS, United States
b Nuclear Magnetic Resonance Laboratory, University of Kansas, Lawrence, 66045 KS, United States
A R T I C L E I N F O
A B S T R A C T
Keywords:
A collection of small molecules has been synthesized by composing photo-cycloaddition, C–H functionalization,
and N-capping strategies. Multidimensional biological fingerprints of molecules comprising this collection have
been recorded as changes in cell and organelle morphology. This untargeted, phenotypic approach allowed for a
broad assessment of biological activity to be determined. Reproducibility and the magnitude of measured fin-
gerprints revealed activity of several treatments. Reactive functional groups, such as imines, dominated the
observed activity. Two non-reactive candidate compounds with distinct bioactivity fingerprints were identified,
as well.
Photochemistry
Biological activity
Complexity
Fingerprints
1. Introduction
and analyzed quantitatively to create a numerical “fingerprint” of a
cell’s response.7 This process is called “cell painting.” Correlating the
Small organic molecules possess unique features that allow them
both to probe and to modulate biological systems.1 The structural fea-
tures that inscribe particular biological activities into the small mole-
cules are in general unknown, but the diversity of biologically active
molecules is at least as broad as the range of biological activities ob-
servable at the organismal, cellular and biochemical levels combined.
To find new lead compounds, as opposed to hits in a specific assay, the
challenge for medicinal chemists is to reach new regions of chemical
space and learn about the biological activities that may be provoked as
quickly and efficiently as possible.2 Both naturally-evolved and syn-
thetically-created products suggest that the complexity of their struc-
tures plays an important role in endowing them with the potency and
specificity required to reliably perturb intricate functions of the biolo-
gical systems.3,4 Diverse chemical origins of molecules (i.e. the se-
quence of chemical reactions that gave rise to them, either through a
biosynthetic gene cluster, or a combinatorial synthesis scheme) corre-
late best with the uniqueness of their protein-binding profiles.5 These
two notions taken together charge chemists with creating a variety of
reaction sequences that produce small molecules with complex struc-
tures. Image-based cell profiling techniques offer one way to assess
quantitatively and reproducibly, multiple markers of the biological
activity of a newly-synthesized small molecule.6 Perturbations induced
by a small molecule lead to changes in cell morphology that can be
captured by staining with fluorescent dyes, imaged microscopically,
chemical-induced changes observed for a set of reference compounds
having known but diverse biological activities generates a database that
can be used to guide the screening of completely new compounds to
uncover those whose “biological activity profiles” are similar, at least at
the cellular level, to the reference compounds. These broad observa-
tions allow one to screen multiple new chemical entities for multiple,
disparate, potential biological activities in one experiment. An out-
standing attribute of many biologically-active natural products that is
not always reflected among synthetic compounds is their stereo-
chemical richness. To approximate this richness in the synthesis of
completely new scaffolds, we have taken advantage of the high-energy
intermediates involved in photochemical processes. Ultraviolet light
infuses small molecules with enough energy to radically reconfigure
their often flat, two-dimensional carbon skeletons.8 Many synthetically
useful transformations have been achieved by gaining control of
chemistry that follows the absorption of a photon.9 Complex and
“unusual” scaffolds that are generated in this fashion can be further
elaborated by the application of modern chemistry that relies on the
activation of carbon-hydrogen bonds.10 Our overall synthetic strategy
consists of combinations of four basic transformations: (1) photo-
cycloaddition between phenylpyrroline and electron-poor terminal
olefins; (2) electrophilic aromatic C-H functionalization with an acti-
vated thianthrene; (3) palladium-catalyzed C-C bond formation; (4) N-
capping alkylation or carbamylation reactions.
⁎ Corresponding author.
E-mail address: zarko@ku.edu (Z. Boskovic).
https://doi.org/10.1016/j.bmc.2020.115547
Received 13 March 2020; Received in revised form 23 April 2020; Accepted 2 May 2020
Availableonline19May2020
0968-0896/©2020ElsevierLtd.Allrightsreserved.

M. Singh, et al.
Bioorganic&MedicinalChemistry28(2020)115547
core spiroindane scaffolds (three diastereomeric pairs) were generated
in a photocycloaddition reaction between phenylpyrrolinium per-
chlorate and acrylonitrile (1 and 14), acrylamide (10 and 22), or me-
thyl acrylate (13 and 25). This was achieved by irradiating 14 mM
solutions of starting material in acetonitrile with a 500 W Hg-Xe lamp
whose output was filtered with a band pass filter permissive to wave-
lengths in the range 255–305 nm. Diastereomers obtained in these
photocycloadditions were separated by column chromatography. Fur-
ther, site-selective C–H functionalization of the aromatic ring of aryl
spiroindane scaffold was achieved by the introduction of tetra-
fluorothianthrene. Spiroindane pyrrolidine substrates were exposed to
trifluoroacylated tetrafluorothianthrene-S-oxide and tetrafluoroboric
acid. The products were chromatographically isolated as tetra-
fluoroborate salts. Trifluoroacetyl groups that were introduced in
thianthrenation reaction were removed with sodium borohydride in
methanol to yield compounds 2, 4, 6, 9, 15, 17, 19, and 26. Compounds
2–5, and 15–18 were prepared from the corresponding thianthrenium
salts and p-methoxyphenylboronic acid, or 3-thienylboronic acid, cat-
alyzed by tetrakis (triphenylphosphine) palladium(0) or bis(diphenyl-
phosphino) ferrocene palladium(II) chloride (10–20 mol%), and basi-
fied with tribasic potassium phosphate. Similarly, compounds 6–9,
19–21, and 26 were prepared from thianthrenium salts and propargyl
alcohol, or propargyl urea in reactions catalyzed by tetrakis (triphe-
nylphosphine) palladium(0) at 10 mol%, in the presence of diisopro-
pylamine as a base, and copper(I) iodide. Compounds 11 and 23 were
prepared under reductive amination conditions with pyrrolidines and
2,3-dimethoxybenzaldehyde in the presence of sodium triacetoxybor-
ohydride, whereas compounds 12 and 24 were prepared from the
spirocyclic amine scaffold with 2-methoxyethyl chloroformate and
triethylamine in dichloromethane. Complete description of the synth-
eses, isolation, and characterization of the compound collection can be
found in the Supporting Information document.
2.2. Cell painting
Culture: U-2 OS (ATCC, HTB-96) cells were cultured in T-75 flasks
(Greiner, 658170) under McCoy’s 5a medium (HIMedia, AT057), sup-
plemented with 10% Fetal Bovine Serum (FBS) and 1% Penicillin/
Streptomycin solution, and incubated at 37 °C in a 5% CO2 atmosphere.
Upon reaching 80% confluence, cells were washed, trypsinized (VWR,
M143), and removed. Flasks were reseeded with one fifth of the cell
count for each passage. All cell plates reported in this paper were
composed of U-2 OS from incubation lengths of five or fewer passages.
These optical-bottomed, 384 black welled plates (Corning, 6569) were
seeded with U-2 OS at a density of 1,000 cells per well and allowed to
incubate for 24 h to ensure adherance. Compound handling: Compound
plates (Thermo Scientific, AB-1056) were constructed using an initial
aliqout of a 50 mM compound solution solubilized in DMSO. Each
compound aliquot was then threefold serially diluted in DMSO to yield
six concentrations ranging from 50 mM to 205 μM. This process was
repeated for each of the twenty-seven test compounds and the starting
phenylpyrrolinium perchlorate. Forty-one vehicle treatments were in-
cluded in the plate design as DMSO containing wells with no added
compound. Staurosporine diluted to 500 μM in DMSO was included on
the plate as a positive control. A second plate composed of known toxic
Fig. 1. Twenty-seven-member compound library. Structure of each compound
(bottom table) can be obtained as a combination of rows from the top table.
First column describes the relative configuration between pyrrolidine nitrogen
and the group R1, and columns 2, 3, and 4 specify the substitution of the
spiroindane scaffold at the top of the figure.
To illustrate the complete concept outlined above, we describe
herein a strategy for rapidly synthesizing a collection of 27 new com-
pounds that contain a spirocyclic amine scaffold, together with the
preliminary insights to the biological activities at the cellular level, vis-
a-vis 13 selected reference compounds having diverse biological ac-
tivities. Those that show interesting biological activities may serve as
starting points for further elaboration for studying structure–activity
relationships and for expanded biological testing. Multidimensional
fingerprinting of the synthesized library of compounds led us to identify
four bioactive compounds with different “cell painting” profiles com-
pared to the reference set.
compounds was constructed using
a maximum concentration on
10 mM, and was threefold serially diluted to a minimum concentration
of 509 nM, except as noted in the Supplemental Material. This control
plate also contained 190 wells filled with only DMSO to serve as vehicle
treatment. Compound treatment: Subsequent to a 24 h incubation
period, untreated cell plates and the compound plate were each fitted
into a Library Copier and a sterile Multi-Blot Replicator (V&P Scientific)
was used to transfer a 100 nL volume from each well of the 384 well
compound plate into the culture media (50 μL) of a corresponding well
on the 384 well cell plate. These cell plates were reintroduced to a 37
°C, 5% CO2 atmosphere for the duration of the 24-h treatment period.
2. Materials and methods
2.1. Chemistry
A small library of 27 racemic compounds (Fig. 1) was synthesized
from a single starting molecule – phenylpyrrolinium perchlorate. Six
2

M. Singh, et al.
Bioorganic&MedicinalChemistry28(2020)115547
Fig. 2. Mariano photocycloaddition between flat phenylpyrrolinium perchlorate and electron-poor olefins produces spirocyclic scaffolds of significantly increased
complexity.
Staining: Upon completion of the treatment period, cell plates had their
media replaced with RPMI media (Sigma, R8755) without phenol red
(10% FBS, 1% Penicillin/Streptomycin solution), containing 500 nM
Mitotracker Deep Red (Thermo Fisher, M22426) in DMSO and 60 μg/
mL Wheat Germ Agglutinin/Alexa Flour 594 (Thermo Fisher, W11262)
in deionized water. The cell plates were incubated for an additional
30 min at 37 °C in 5% CO2. The cells were then fixed in their wells with
4% formaldehyde (Electron Microscopy Sciences, 157110-S) for 20 min
at room temperature, protected from light. Following a wash in Hank’s
Buffered Saline Solution (HBSS), cells were permeablized with a 15 min
incubation with a 0.1% solution of Triton X-100 (VWR, M143) in HBSS
at room temperature, protected from light. After permeabilization, cells
were washed in HBSS. The second staining solution composed of 1%
BSA (VWR, 0332) in HBSS, 0.025 mL/mL Phalloidin/Alexa Flour 594
(Thermo Fisher, A12381), 100 μg/mL Concanavalin/Alexa Flour 488
(Thermo Fisher, 11252), 5 μg/mL Hoechst 33342 (Thermo Fisher,
H3570), and 3 μM Syto 14 (Thermo Fisher, S7576) was then added and
allowed to incubate for 30 min at room temperature, protected from
light. A final HBSS wash concluded the staining process and the cell
plates were sealed with aluminum sealing tape and imaged.
Experiments were repeated 3 times with cells from different passages.
Imaging: Sealed plates were imaged using an ImageXpress Micro XL
microscope and MetaXpress image acquisition software. Five wave-
length channels, DAPI, GFP, Cy3, Cy5, and TxRed, were used to capture
fluorescent images of each well at 10× magnification, and with 2 × 2
binning. Each well was divided into four quadrants in order to capture
the entire well at the selected magnification. Whole plate imaging was
preceded by an initial focusing step using the DAPI channel. Images
were stored on a local hard drive to be analyzed by Cell Profiler.11
3. Results & discussion
3.1. Synthesis of the compound collection
We used a photochemical reaction developed by Mariano12 to pre-
pare spiroindane pyrrolidine scaffolds in one step by pairing a single
readily available starting material, 2-phenyl-1-pyrrolinium per-
chlorate,13 with 3 electron-deficient olefins – acrylonitrile, acrylamide,
and methyl acrylate. Excitation of pyrrolinium salts is achieved by ir-
radiation with ultraviolet light from a 500 W Hg-Xe lamp, filtered with
a band pass filter (transmissive for wavelengths 255–305 nm), over 1 h
in acetonitrile at 14 mM concentration. The fate of the phenylpyrroli-
nium excited state in the presence of electron-poor olefins is best ex-
plained by an initial [π2 + π2] arene-olefin cycloaddition to produce a
bicyclic diene. Quaternary carbon–carbon bond migration (1,2-re-
arrangement) by cyclobutane ring expansion in concert with deproto-
nation generates the spirocyclic amine system and rearomatizes ben-
zene. Alternatively, the mechanism may invoke the formation of a
singlet biradical intermediate, followed by an electron-transferring
oxidation of olefin (formation of a radical cation) and reduction of N-
centered radical to amine.12 Recombination of radicals, and elimination
leads again to the final spirocyclic amine product. Regardless of these
intriguing mechanistic underpinnings, the reaction reliably produces a
chromatographically separable mixture of syn- and anti-benzospir-
ocyclic amines (Fig. 2). Obtained NMR data matches the reported NMR
of the syn- and anti-spiroamino nitriles and spiroamino esters.14 Re-
lative configuration of spiroamino amides was determined by dehy-
drative conversion of primary amide to nitriles that we inadvertently
accomplished15 during the thianthrenation reaction (in the presence of
trifluoroacetic anhydride). All the secondary amines are most easily
visualized with a ninhydrin stain upon resolving on a thin-layer chro-
matography plate.
2.3. Data analysis
Intuitive grasp of the increase in structural complexity provided by
this transformation is confirmed by the two complexity descriptors
which we used to measure it. Both Bertz16 and Böttcher17 complexity
indices increase by about 30% in this reaction (sum of the complexity
indices of starting material compared to the product complexity index).
Qualitatively, exploration of the insufficiently studied chemical space of
compounds containing spirocyclic amines18 provides additional moti-
vation for the development of small collections featuring this structural
motif. Hinting to the utility that photochemistry may hold as a method
for generating structurally diverse compound collections, the photo-
cycloaddition reaction with acrylonitrile afforded a novel, completely
dearomatized 2,4-meta-cycloadduct, compound 27, in addition to syn-
and anti-amino nitriles. We verified its structure through the X-ray
diffraction pattern analysis (Fig. 3; CCDC: 1976540) and included it as
part of the collection for bioactivity assessment. Diversification of the
aromatic portion of the spirocyclic scaffolds necessitated that either a
proper substituent is carried through the photochemical step, or that
selective carbon-hydrogen functionalization protocol is implemented
after the photo-cycloaddition. The effect of phenylpyrroline
Cell Profiler output comprises extracted feature values from 5 image
sets for each of four sites of each of 384-wells of compound- or vehicle-
treated cells. Features are initially measured per object (cell, nucleus, or
cytoplasm) and can be averaged by Cell Profiler to give per-site values,
that we averaged subsequently. We first calculated the average feature
values for DMSO-treated cells (190 wells on the control plate, 41 on the
test plate) together with the variability in these measurements, which
was represented as the standard deviation of the measurements. We
then subtracted the DMSO-derived averages from the treatment con-
ditions and we divided this number with the standard deviation to
obtain the Z-scaled value for each feature. From the Z-scaled values we
computed the magnitudes of feature vectors by taking their norms, and
we computed correlations between replicates, or between the active
compounds (including controls) by taking the inner products of the
appropriate feature vectors and dividing with the products of their
lengths (cosine similarity). The magnitude of feature vectors is a mea-
sure of difference between the compound and vehicle treatments. The
correlation between replicates or between the active compounds is a
measure of similarity of those conditions.
3

M. Singh, et al.
Bioorganic&MedicinalChemistry28(2020)115547
Fig. 3. Photochemical reactions produce complex
molecules. Crystal structure of compound 27, dihy-
drosemibullvalene derivative. CCDC: 1976540.
Fig. 5. Diversification of the aromatic portion of spiroindene pyrrolidine scaf-
folds has been accomplished through C-H functionalization. Regiochemistry of
the substitution was determined from HMBC correlations from the aromatic
singlet to the quaternary carbon.
electron-withdrawing groups on alkene partners in the Mariano pho-
tocycloaddition, this reductive step precluded the use of groups that can
be reduced by sodium borohydride (e.g. esters, ketones). Interestingly,
the attempted thianthrenation reaction on primary amides, produced
nitrile products through an eliminative process that follows O-acylation
with trifluoroacetic anhydride, without the need for additional base
Fig. 4. (A) Substitution of phenylpyrrolinium strongly impacts the electronic
nature of the phenyl ring as evidenced by changes in the absorption spectra; (B)
Tentatively assigned structures obtained from various substituted phe-
nylpyrrolinium perchlorates and acrylonitrile.
substitutions is apparent in the changes in absorption spectra in UV
region of the spectrum (Fig. 4A), and these changes manifested them-
seleves as different reactivity of substituted phenylpyrrolinium salts in
the photochemical reaction. For example, p-methoxy substituted phe-
nylpyrrolinium undergoes a [π2 + π2] cycloaddition with the imine
functional group instead of the benzene ring.12 This leads to the for-
mation of azetidine and no C–C bond migration during the rear-
omatization event. Bromo substituted substrates surprisingly undergo a
radical conjugate addition to acrylonitrile from the 5-position of 1-
pyrroline (Fig. 4B).
common in these dehydrations.15
.
To achieve scaffold elaboration on the secondary amine, we relied
on a well-precedented reductive amination with aromatic aldehydes,21
and the carbamate formation with a chloroformate (Fig. 6A). Em-
ploying these two conditions on spiroindane pyrrolidine amides furn-
ished tertiary amines and carbamates, the latter existing at room tem-
perature as a mixture of rotamers (Fig. 6B).22 In the NOE experiment,
the selective irradiation of the characteristic methoxy peak of com-
pound 12 at 3.47 ppm, produces a negative peak corresponding to the
irradiated signal itself, as well as a peak at 3.12 ppm in the same phase.
This confirms that signals are equilibrating at a rate faster than the
NMR NOE relaxation.
Strategically, in this diversity-oriented synthetic campaign, the
sensitivity of the substrate towards substitution necessitated con-
sidering the carbon–hydrogen bond functionalization at a post-photo-
chemistry stage to introduce groups at the aromatic portion of spior-
indane pyrrolidines. The presence of four electronically similar C–H
bonds called for a mild reagent for electrophilic aromatic substitution,
such that it would be capable of distinguishing between minor differ-
ences in nucleophilicity of these positions. Reactivity and kinetics of
thianthrene cation-radical perchlorate with aromatics have been ex-
plored by Shine in the 1970s.19 Those studies showed that electron-rich
aromatics are preferred substrates and that the electrophilic aromatic
substitution takes place para to the directing group. Ritter recently re-
ported a modified version of this SEAr reaction by employing tetra-
fluorothianthrene S-oxide and trifluoroacetic anhydride in the presence
of tetrafluoroboric acid.20 The thianthrene radical cation generated
under these conditions reacts regioselectively with a wider variety of
aromatic substrates compared to Shine’s set-up. We used these condi-
tions to introduce a chemical handle into syn- and anti-amino nitriles
and to further elaborate these pseudo-halides through palladium-cata-
lyzed C–C bond forming reactions, e.g. Sonogashira and Suzuki (Fig. 5).
As expected, trifluoroacetic anhydride, used for the activation of S-
oxide, rapidly reacts with free secondary amines in our substrates be-
fore the sluggish aromatic substitution reaction can take place. This
necessitated a reductive removal of trifluoroacetamide with sodium
borohydride, as methanolysis attempts were unsuccessful. In the con-
text of initial diversification of the library through the variation of
The attempt at reductive amination of the syn-amino amide
Fig. 6. Secondary amines were alkylated in reductive amination reaction or
carbamylated with a chloroformate. Carbamates exist as two rotamers in ap-
proximately equal amounts.
4

M. Singh, et al.
Bioorganic&MedicinalChemistry28(2020)115547
Fig. 7. (A) Syn-amino amide formes cyclic aminal
under the reductive amination conditions; (B)
Important NOE correlations for assignment of the
aminal configuration.
interested in understanding connection between structural complexity
of small molecules and their biological activity. Against that backdrop,
we designed a compound collection centered around spirocyclic scaf-
fold of spiroindane pyrrolidines (vide supra). To quickly learn about the
biological activity of molecules in this collection, we performed an
experiment in which changes in cell morphology upon compound
treatment are observed under a microscope after organelle-specific
staining with fluorescent dyes (i.e. we performed a “cell painting” ex-
periment).24 We chose fast-growing U-2 OS cells for this assessment due
to their exaggerated morphology, which lends itself to easily observable
organellar staining. Performing these experiments in 384-well plates
allowed us to test a range of compound concentrations (descending
from 100 μM to 0.41 μM for test compounds) to ensure that we were not
missing activities that are observable only at high or low concentra-
tions.25 These experiments can create a broad fingerprint of biological
activity of newly-synthesized compounds. Similarity between finger-
prints can indicate a shared mechanism of action and provide a clue
about which targets or, more broadly, pathways are modulated by the
action of the small molecule. While the power that multidimensional
measurements bring to rapid assessment of biological activity of newly-
synthesized compounds is undeniable, the obstacles to wider adoption
of this approach are centered around (1) the reproducibility of weak
fingerprints, (2) the assignment of actual phenotypic meaning of the
measured changes, and (3) the uncertainty with respect to the granu-
larity of such measurements.
Other research groups have started defining positive controls for
this type of experiment, and we adopted 13 compounds to use as the
benchmark measurements here, as well.6,24 Ten concentrations and 3
repeated treatments (biological replicates) with control compounds
helped us establish a reliable, quantitative understanding about what
constitutes a bona fide change in cell and organelle morphology that can
be measured in this experiment (Fig. 8). In a departure from previous
methods of cell painting analyses, we opted against averaging the re-
plicates, instead defining the reproducibility of the fingerprint as an
average correlation for each pair of replicates. This approach led us to
discard one of the biological replicates for the newly-synthesized
compounds as the correlation with other 2 experiments was low. Upon
closer inspection, we learned that the cells grew to greater density in
that condition – a fact that hints to one major source of “instability” of
observed fingerprints. Overall, the reproducibility among the control
compounds was higher than for test compounds. Cosine similarity be-
tween replicates 1 and 2 was 0.5731, between 2 and 3 it was 0.6428,
and between 1 and 3 0.5907. The same measure for reproducibility of
fingerprints of the test compounds is lower, 0.2189. This is expected
from a collection where the majority of treatments causes no change in
cell morphology compared to DMSO. As a general rule, the magnitude
of the fingerprint (defined as the length of 1969-dimensional vector of
Z-scaled features) correlates with its reproducibility (defined as the
inner product between fingerprint vectors from different replicates). As
can be seen in Fig. 9, values of fingerprint magnitude under 50 are very
unlikely to reach level of reproducibility of 0.7. The same figure shows
that even positive controls can have weak fingerprints, likely at low
concentrations where the effects of these bioactive compounds do not
manifest themselves as changes in organelle appearance.
Fig. 8. Reproducibility of the “cell painting”-derived bioactivity fingerprints.
Control compounds produce strong fingerprints (compare the y-axes), and
measured features correlate well between replicates.
produced a stable pentacyclic aminal (compound 11, Fig. 7A) that re-
sisted reduction with sodium triacetoxyborohydride. Configuration as-
signment of the aminal carbon has been determined in NOESY experi-
ment (Fig. 7B). Ab initio modeling suggests the exo-arrangement to be
about 5.5 kcal/mol more stable than the endo-one. This cyclization was
not observed for the anti-isomer, which further suggests that proximity
of the amide nitrogen and transiently formed iminium is critical for
cyclization.
The tangible outcome of the synthetic campaign described in pre-
vious paragraphs is a collection of small molecules in 5–20 mg quan-
tities (Fig. 1) based around a common spiroindane pyrrolidine scaffold.
Computed structural properties of the collection place it in a favorable
property space for biological activity profiling (Supporting Information:
Cheminformatics). The average molecular weight of the collection is
302.28 g/mol, AlogP23 is 2.42, and the median number of hydrogen-
bond donors and acceptors are 3 and 1, respectively. The mean polar
surface area of molecules in the collection is 56.57 Å2. A typical re-
presentative of the collection has one aromatic ring and one rotatable
bond – both structural descriptors testifying to the relative rigidity of
the molecules in this collection. Structural alerts triggered by the
compounds in the collection are due to the presence of alkyne or ester
functional groups.
3.2. Analysis of biological fingerprints
There are three main applications for “cell painting” in the context
of newly-synthesized compound collections: (1) determination of raw
bioactivity, which can be defined as a significant and reproducible
We embarked on the experiments described in this paper with a goal
of tailoring the technique of “cell painting” to fit our needs for assessing
our biological-assays-naïve compound collection. Our research group is
5

M. Singh, et al.
Bioorganic&MedicinalChemistry28(2020)115547
Fig. 9. Reproducibility and the magnitude of fin-
gerprints allow for the identification of novel
bioactive compounds from this compound collec-
tion. Each circle represents a “cell painting” finger-
print of one compound at a given concentration.
Reproducibility shown on the abscissa is obtained by
averaging correlations between replicates for each
treatment. The magnitudes of fingerprints for each
treatment were averaged and plotted on the ordi-
nate. In addition to the chemically reactive phe-
nylpyrroline and compound 27, compounds 1 and 4
produced moderately strong and reproducible
changes in cell morphology.
Fig. 10. Active compounds can be successfully grouped by cosine similarities of fingerprints quantifying the morphology changes they cause in cells.
change in appearance of organelles compared to DMSO treatment, (2)
grouping of compounds based on similarity of their fingerprints which
in a small collection would correspond to structure–activity relation-
ships, and (3) assigning the mechanism-of-action of “active” treatments
by similarity to the known bioactive compounds. This experiment re-
vealed to us that the most active compound turned out to be the sole
starting material for the collection synthesis! The precursor phe-
nylpyrroline showed the strongest and most reproducible fingerprints
in at least 2 concentrations (100 μM and 33 μM). A likely explanantion
for this activity is the reactive imine functional group. Strengthening
this explanation is the observed activity of compound 27, a dihy-
drosemibullvalene meta-cycloadduct, that also contains the same
functional group. Previously reported imines and aziridines originating
from different scaffolds were also common “active” compounds found
in this type of experiment.26 Based on the similarity of their finger-
prints, phenylpyrroline and compounds 27 and 4 cluster together and
away from any other control compound tested in this experiment
(Fig. 10). Compounds 1 and 4 were the only non-imine containing
compounds having moderately strong and reproducible fingerprints.
These compounds are both syn-spirocyclic amino nitriles and have no
6

M. Singh, et al.
Bioorganic&MedicinalChemistry28(2020)115547
Fig. 11. Images of vehicle-treated cells and cells
treated with compounds 1, 4, and 27. Differences in
cell morphology are apparent even to a human eye.
Composite image of all wavelengths is shown (blue:
nucleus, green: ER, yellow: nucleoli, orange: Golgi
and plasma membrane, red: mitochondria). (For
interpretation of the references to colour in this
figure legend, the reader is referred to the web
version of this article.)
structural “red flags” associated with their structures (Supporting
Information: Cheminformatics). Importantly, the corresponding dia-
stereomers (compounds 14 and 17) are not active in this experiment,
strengthening the case for a specific interaction that produces the ac-
tivity. Based on the similarity of their fingerprints, compound 1 clusters
with fenbendazole and rotenone, whereas compounds 4, 27, and phe-
nylpyrroline all cluster together in a distinct clade (Fig. 10).
identification of starting points for the discovery of novel bioactive
substances. Creative output of synthetic chemistry coupled with rapid
insights that can be obtained from imaging cell populations treated
with newly-synthesized compounds is a fertile ground for discovery of
bioactive substances with novel mechanisms of action. Widespread
adoption of this bioactivity assessment approach will benefit from a
more standardized set of measured features. Unambiguous naming
conventions would facilitate comparisons between fingerprints gener-
ated at different labs and originating from different experiments.
Additionally, ascribing biological meaning to the observed changes
would provide a more satisfying aspect to this biological annotation
experiment. Connecting the compound structural features to the ob-
served biological fingerprints has the potential to expedite structure–-
activity relationship studies and the identification of novel bioactive
compounds.
A qualitative interpretation of images of cells treated with com-
pounds 1, 4, and 27 reveals drastic changes compared to vehicle
treatment (Fig. 11). Each treatment was hallmarked by an increase in
both mitochondrial and endoplasmic reticulum staining compared to
that of vehicle-treated cells. Cytoplasmic swelling was commonly ob-
served and severe membrane blebbing was abundant. This may indicate
loss of the ability to control intracellular osmotic pressure. Nuclear
fragmentation was clear and frequent. Furthermore, apoptotic bodies
were numerous in the extracellular space for all the compounds that
show increased toxicity. (See Supporting Information for complete data
on toxicity of all control and test compounds.) Vehicle-treated cells are
characterized by well-defined nuclei dotted with numerous nucleoli.
Mitochondrial and ER stain intensities are balanced with that of cyto-
plasm stain. Cellular shape is defined by polygonal borders typical of U-
2 OS cancer line. Imine-containing compound 27 (at 33 μM) produces a
very strong ER and mitochondrial staining, and cells and nuclei are
swollen. Compound 1 at 100 μM causes similar amplification of the
mitochondrial staining. Membrane blebbing advanced with nuclear
fragmentation, and misshapen cells are abundant with a few apoptotic
bodies observed. Effects of compound 4 at 100 μM are similar to those
of compound 1 with abnormally shaped cells, strong ER staining, and
disappearance of numerous nucleoli present in normal U-2 OS cells.
One of the obstacles to the wider adoption of “cell painting” (vide
supra) concerns the uncertainty with respect to the granularity of these
fingerprints, i.e. what are the mechanisms that this technique can re-
liably distinguish? It seems apparent from the accumulated experience
in this area that the compounds that target cytoskeleton will un-
surprisingly produce the strongest fingerprints. In this study, we also
observed that cytotoxic compounds seem to have strong fingerprints
(they are obviously bioactive) and these fingerprints do not all seem to
coalesce around common morphological changes. For instance, the
imines (phenylpyrroline and compound 27) were cytotoxic, but this
cytotoxicity produced a fingeprint different from the cytotoxicity of
vincristine or taxol (Fig. 10). The mechanism of action of compounds 1
and 4 is difficult to precisely assign at this point, and is a topic of on-
going work in our laboratory. As a first step in this process, computa-
tional protein target prediction algorithm27 suggested acetylcholine
esterase and β-secretase as potential protein targets of these com-
pounds.
Annotating the biological-assays-naïve compound collection with
“cell painting” has provided us with the immediate experimental in-
sights into the bioactivities of its members. We will use these insights to
guide the improvements in synthetic chemistry to prepare the collec-
tion, and to define the exact mode of action of the active compounds.
5. Data statement
Raw images and Cell Profiler extracted features per-object are
available upon request. Processed per-well files, analysis scripts, code to
generate images, and compound.sdf files with annotated NMR shifts are
available on the laboratory’s GitHub repository associated with this
paper: https://github.com/boskovicgroup/spiroindane_pyrrolidines.
Declaration of Competing Interest
None.
Acknowledgments
We thank the University of Kansas New Faculty General Research
Fund, General Research Fund, and Protein Structure–Function COBRE
for financially supporting this project. We acknowledge Dr. Victor Day
for X-ray diffraction experiments, and the NSF-MRI grant CHE-0923449
that was used to purchase the X-ray diffractometer and software used in
this study. We thank Ms. Ankita Ramprasad, Ms. Annu Anna Thomas,
and Mr. Matthew McCurry for their help with reproducing previously
reported syntheses. We thank KU’s NMR core facilities for acquiring
characterization data that made conclusions of this publication
stronger. We are grateful to Prof. Robert Hanzlik for improving the
quality of the manuscript, and to Prof. Patrick S. Mariano (University of
New Mexico) for the helpful e-mail exchange regarding the structure of
compound 27.
4. Conclusions
We have described a synthetic chemistry strategy to prepare a col-
lection of 27 molecules occupying new chemical space, and we gener-
ated their bioactivity fingerprints through “cell painting.” From these
measurements we identified two compounds that could be classified as
bioactive. This project demonstrated the feasibility of experimental
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in the
online version, at https://doi.org/10.1016/j.bmc.2020.115547.
7

M. Singh, et al.
Bioorganic&MedicinalChemistry28(2020)115547
References
3-aminopropyl carbanion equivalent in the synthesis of substituted 1-pyrrolines: 2-
phenyl-1-pyrroline. Org Synth. 2003;75:215.
14. Mariano PS, Leone-Bay A. Photocycloadditions of electron poor olefins to 2-aryl-1-
pyrrolinium salts. Tetrahedron Lett. 1980;21:4581–4584.
1. Schreiber SL. Chemical genetics resulting from a passion for synthetic organic
chemistry. Bioorg Med Chem. 1998;6:1127–1152.
15. Bruckner R. Carboxylic compounds, nitriles, and their interconversion. Organ Mech:
React Stereochem Synth. 2010:321–338.
2. Mignani S, Huber S, Tomas H, Rodrigues J, Majoral J-P. Compound high-quality
criteria: a new vision to guide the development of drugs, current situation. Drug
Discov Today. 2016;21:573–584.
16. Bertz SH. The first general index of molecular complexity. J Am Chem Soc.
1981;103:3599–3601.
3. Selzer P, Roth H-J, Ertl P, Schuffenhauer A. Complex molecules: do they add value?
Curr Opin In Chem Biol. 2005;9:310–316.
17. Böttcher T. An additive definition of molecular complexity. J Chem Inf Model.
2016;56:462–470.
4. Méndez-Lucio O, Medina-Franco JL. The many roles of molecular complexity in drug
discovery. Drug Discov Today. 2017;22:120–126.
18. Müller G, Berkenbosch T, Benningshof JC, Stumpfe D, Bajorath J. Charting biologi-
cally relevant spirocyclic compound space. Chem–A Eur J. 2017;23:703–710.
19. Shine HJ, Silber JJ. Ion radicals. xxii. reaction of thianthrenium perchlorate
5. Clemons PA, Bodycombe NE, Carrinski HA, et al. Small molecules of different origins
have distinct distributions of structural complexity that correlate with protein-
binding profiles. Proc Nat Acad Sci. 2010;107:18787–18792.
.+
(C12H8S2 ClO4-) with aromatics. J Organ Chem. 1971;36:2923–2926.
20. Berger F, Plutschack MB, Riegger J, et al. Site-selective and versatile aromatic C–H
functionalization by thianthrenation. Nature. 2019;567:223.
6. Pahl A, Sievers S. The cell painting assay as a screening tool for the discovery of
bioactivities in new chemical matter. Syst Chem Biol, Springer. 2019:115–126.
7. Bray M-A, Singh S, Han H, et al. Cell painting, a high-content image-based assay for
morphological profiling using multiplexed fluorescent dyes. Nat Protoc.
2016;11:1757.
21. Abdel-Magid AF, Carson KG, Harris BD, Maryanoff CA, Shah RD. Reductive amina-
tion of aldehydes and ketones with sodium triacetoxyborohydride. Studies on direct
and indirect reductive amination procedures. J Organ Chem. 1996;61:3849–3862.
22. Hu DX, Grice P, Ley SV. Rotamers or diastereomers? An overlooked NMR solution. J
Organ Chem. 2012;77:5198–5202.
8. Karkas MD, Porco Jr JA, Stephenson CR. Photochemical approaches to complex
chemotypes: applications in natural product synthesis. Chem Rev.
2016;116:9683–9747.
23. Ghose AK, Viswanadhan VN, Wendoloski JJ. Prediction of hydrophobic (lipophilic)
properties of small organic molecules using fragmental methods: an analysis of AlogP
and ClogP methods. J Phys Chem A. 1998;102:3762–3772.
9. Hoffmann N. Photochemical reactions of aromatic compounds and the concept of the
photon as a traceless reagent. Photochem Photobiol Sci. 2012;11:1613–1641.
10. Godula K, Sames D. Ch bond functionalization in complex organic synthesis. Science.
2006;312:67–72.
24. Gustafsdottir SM, Ljosa V, Sokolnicki KL, et al. Multiplex cytological profiling assay
to measure diverse cellular states, PloS One 8 (12).
25. Perlman ZE, Slack MD, Feng Y, et al. Multidimensional drug profiling by automated
microscopy. Science. 2004;306:1194–1198.
11. Carpenter AE, Jones TR, Lamprecht MR, et al. Cellprofiler: image analysis software
for identifying and quantifying cell phenotypes. Genome Biol. 2006;7:R100.
12. Stavinoha JL, Mariano PS. Electron-transfer photochemistry of iminium salts. olefin
photoadditions to 2-phenyl-1-pyrrolinium perchlorate. J Am Chem Soc.
1981;103:3136–3148.
26. Gerry CJ, Hua BK, Wawer MJ, et al. Real-time biological annotation of synthetic
compounds. J Am Chem Soc. 2016;138:8920–8927.
27. Mervin LH, Afzal AM, Drakakis G, et al. Target prediction utilising negative bioac-
tivity data covering large chemical space. J Cheminform. 2015;7:1–16.
13. Sorgi KL, Maryanoff CA, McComsey DF, Maryanoff BE. N-vinylpyrrolidin-2-one as a
8