J.K. Eaton, et al.
Bioorganic&MedicinalChemistryLetters30(2020)127538
selectivity for ferroptosis induction, as the electrophilic sites are at
different positions in these compounds than they are in the active ni-
trile-oxide form of ML210.
featuring RSL3.2,10,15 Our observations do, however, demonstrate that
GPX4-targeting warheads and scaffolds are interchangeable in certain
instances. On the basis of this discovery, we explored the activity of the
nitroisoxazole warhead of ML210 in conjunction with the core scaffolds
of several chloroacetamide-based GPX4 inhibitors (Fig. 3A).
To investigate this possibility, we synthesized an analog of ML162
wherein the chloroacetamide group is replaced with a nitrolic acid
warhead (40) (Fig. 3B), which reacts with nucleophiles via a nitrile-
but with diminished potency relative to ML162 and propiolamide
analog 8. The different position of the electrophilic site of 40 likely
contributes its lessened potency; the active nitrile-oxide species extends
the position of the electrophilic site relative to the core scaffold by one
carbon atom compared to the chloroacetamide in ML162 and changes
the orientation relative to the propiolamide of 8. However, despite the
diminished potency of 40, this result demonstrates that it is possible for
an ML162 analog to target GPX4 through a nitrile-oxide electrophile.
The electrophilic site in 40 is positioned similarly to the hypothetical
site in 32 upon cellular activation of the nitroisoxazole. This parallel
suggests that failure to unmask the nitrile-oxide electrophile may un-
derlie the inactivity of 32 and possibly other nitroisoxazoles.
based GPX4 inhibitors, including ML162, RSL3, DPI19,2,25 and
DPI132,25 (32–36, 38) (Fig. 3A). Assessment of these nitroisoxazoles
revealed that most cannot induce ferroptosis at the highest tested
concentrations (Fig. 3A), suggesting that this warhead is not generally
DPI19 scaffold functionalized with a nitroisoxazole warhead, can in-
duce ferroptotic cell death (Fig. 3A).
Cellular transformation of the nitroisoxazole group of ML210 into
its active nitrile-oxide form requires two main steps: ring-opening hy-
drolysis of the nitroisoxazole and subsequent dehydration of the re-
sulting α-nitroketoxime (Supplementary Fig. 4A).10 It remains unclear
whether enzymatic or purely chemical processes underlie these trans-
formations, but our results indicate that there may be structural re-
quirements distal to the nitroisoxazole that determine whether this
prodrug mechanism is operative. We also note that we were unable to
use standard hydrolysis conditions10,26 to convert certain nitroisox-
azoles into their corresponding α-nitroketoxime forms, including 32–34
(Supplementary Fig. 4B, C). It may be the case that this initial activating
transformation is also not achievable within live cells.
Our studies reveal that the nitroisoxazole warhead cannot ne-
cessarily substitute for other electrophiles in GPX4 inhibitors. This
distinction is likely due to specific structural requirements for both
multistep nitroisoxazole activation as well as for GPX4-targeting ability.
However, we hypothesized that other structurally distinct nitroisox-
azoles could target GPX4 without being appended to scaffolds of known
GPX4 inhibitors. To understand the ability of nitroisoxazoles to target
GPX4 more generally, we screened 30 nitroisoxazole-containing com-
pounds in cell viability and fer-1 rescue experiments (Supplementary
Table 1). Most lacked the ability to induce ferroptosis, but we identified
several compounds inducing cell death that was rescued by fer-1 co-
treatment. This collection includes two compounds that are close
structural analogs of ML210, 41 and 42 (Supplementary Table 1). In
Another mechanistic hypothesis for inactive nitroisoxazole analogs
is that the requisite transformations to the nitrile-oxide electrophile
occur, but the resulting nitrile oxides are unable to target GPX4. One
possibility may be that for certain inhibitor scaffolds, the nitrile-oxide
group is not accessible to the selenocysteine residue for covalent
modification. This caveat may explain why chloroacetamide (15) and
propiolamide (16) analogs of ML210 exhibit diminished potency and
Fig. 4. Characterization of low-MW nitroisoxazole
GPX4 inhibitors. (A) Chemical structures of 43 and
44. (B) 43 and 44 induce ferroptosis in LOX-IMVI
cells. Data are plotted as mean
s.e.m. of n = 2
biological experiments performed in duplicate. See
also Supplementary Table 1. (C) Assessment of 44
(100 µM, 16 h) proteome-wide reactivity reveals la-
beling of a major target protein at ~38 kDa. (D)
Treatment of LOX-IMVI cells with 44 (100 µM, 16 h)
enables pulldown of GPX4. (E) Compound 44 cova-
lently binds GAPDH.
5