diff --git a/src/Fable.Transforms/Python/Fable2Python.Transforms.fs b/src/Fable.Transforms/Python/Fable2Python.Transforms.fs
index c445abfde4..180e9855e1 100644
--- a/src/Fable.Transforms/Python/Fable2Python.Transforms.fs
+++ b/src/Fable.Transforms/Python/Fable2Python.Transforms.fs
@@ -382,6 +382,14 @@ let transformCast (com: IPythonCompiler) (ctx: Context) t e : Expression * State
             let value, stmts = com.TransformAsExpr(ctx, e)
             libCall com ctx None "core" "int32" [ value ], stmts
         | _ -> com.TransformAsExpr(ctx, e)
+    // The other integer widths are wrapper classes too, so `uint32 (x: byte)` needs the same explicit constructor call
+    | Fable.Number((Int8 | UInt8 | Int16 | UInt16 | UInt32) as kind, _), _ ->
+        match e.Type with
+        | Fable.Number(sourceKind, _) when sourceKind = kind -> com.TransformAsExpr(ctx, e)
+        | Fable.Number _ ->
+            let value, stmts = com.TransformAsExpr(ctx, e)
+            libCall com ctx None "core" (Py.Replacements.getIntTypeName kind) [ value ], stmts
+        | _ -> com.TransformAsExpr(ctx, e)
     // Float64 is a plain Python `float`, and the same reasoning applies:
     // `Replacements.toFloat` emits a bare `TypeCast` for every numeric source that is
     // not bigint/decimal/int64, so `float (x: float32)` would otherwise stay a
diff --git a/tests/Python/TestConvert.fs b/tests/Python/TestConvert.fs
index 781b711289..a77a50fcaf 100644
--- a/tests/Python/TestConvert.fs
+++ b/tests/Python/TestConvert.fs
@@ -418,6 +418,26 @@ let private toIntFromInt16 (x: int16) = int x
 let private toFloatFromSingle (x: float32) = float x
 let private toFloatFromByte (x: byte) = float x
 let private toFloatFromInt (x: int) = float x
+let private toInt16FromSByte (x: sbyte) = int16 x
+let private toUInt16FromByte (x: byte) = uint16 x
+let private toUInt32FromByte (x: byte) = uint32 x
+let private toUInt32FromUInt16 (x: uint16) = uint32 x
+
+[<Fact>]
+let ``test Widening between integer wrappers uses the target width`` () =
+    // Compare as int so wrapper equality cannot truncate the expected value to the source width.
+    toInt16FromSByte 100y <<< 1 |> int |> equal 200
+    toInt16FromSByte -100y <<< 1 |> int |> equal -200
+    toUInt16FromByte 200uy <<< 1 |> int |> equal 400
+    toUInt32FromByte 200uy <<< 1 |> int |> equal 400
+    toUInt32FromUInt16 60000us <<< 1 |> int |> equal 120000
+
+[<Fact>]
+let ``test Widening between integer wrappers has the target runtime type`` () =
+    box (toInt16FromSByte 5y) :? int16 |> equal true
+    box (toUInt16FromByte 5uy) :? uint16 |> equal true
+    box (toUInt32FromByte 5uy) :? uint32 |> equal true
+    box (toUInt32FromUInt16 5us) :? uint32 |> equal true
 
 [<Fact>]
 let ``test Widening to int does not keep the source width`` () =